A de-identified composite from those conversations captures the problem. A financially sophisticated professional had accumulated dozens of mutual funds across multiple platforms: active SIPs alongside forgotten one-time investments, overlapping exposures, no clean way to decide what to keep, exit, or ignore. The returns were fine. The management was chaos.

India is not short of financial products. It is short of someone - or something - coordinating them. The missing loop is insight → action → trust: understand the household, help it make and complete a suitable decision, remember what happened, and show up again when circumstances change.

That loop has never closed at scale because the labour required to close it costs more than most households are worth to serve. AI attacks exactly that cost. But cheaper intelligence does not automatically produce trust, permission to act, or a durable business - and the mistake most “AI wealth” pitches make is assuming it does.

Part I - Why the loop never closed

India’s investing population has expanded faster than its advice infrastructure: roughly 13.9 crore unique securities investors and over 21 crore demat accounts in early 2026, against around 1,000 registered investment advisers. Banks and brokers fill some of the gap, but they distribute products or service wealthy relationships - very little of that supply continuously coordinates an ordinary household’s financial life.

The economics explain why. Good advice is not one portfolio recommendation a year; it is reconstruction, explanation, suitability checks, execution, monitoring, and intervention when a workflow breaks. That work does not scale down - as the wallet shrinks, revenue per household falls faster than the effort to serve it. So advisers move upmarket, platforms simplify, and the household coordinates the rest.

Robo-advisers digitised the portfolio, not the adviser. Fifteen years on, two independent survivors: Wealthfront’s FY2026 filing shows $94.1B in platform assets, 74% of revenue from cash management against 25% from advisory, and Betterment has been stuck near $1.3B since 2021. Our read: at the category’s best outcome, advice earns a quarter of the revenue - valuable, but not the business. Advice is a wedge; spread is a business.

What AI changes is not the model. It is the price of the wedge. Robo 1.0 paid roughly $650 to acquire an account; Cleo built a $300M revenue business at around $11, because the product itself does the acquiring. Two orders of magnitude - and at that price three segments become servable instead of one.

Part II - One market, three definitions of wealth

The right segmentation is not by AUM band but by what wealth means to the user - because the definition of wealth determines the value lever, the value lever determines the business model, and the business model determines what kind of company you are underwriting. Run that across the pyramid and five demographic tiers collapse into exactly three plays.

Play 1 - Supply aggregation: the AI-native Anand Rathi. At the top - roughly 19,900 UHNWIs and 8.7 lakh millionaire households - the user is not underserved but expensively served, and happily so. Wealth means preservation, succession, access, and the relationship manager already owns the trust. Alpha pitched against an engaged RM is a losing proposition, so own the RM instead.

The incumbent template’s economics are public: Anand Rathi Wealth runs 417 RMs managing ₹249 crore each - roughly ₹3 crore of revenue per RM, 33 client families per RM, at ~34% PAT margins. The AI proposition reduces to hours saved, workflows automated, tasks outsourced, all rolling into ARR per FTE. The US has validated the shape at Savvy Wealth and Farther. The twist for India is downward: margin expansion is the entry, but if AI takes an RM from 33 families to 100+, the ₹50 lakh-₹2 crore emerging-HNI tier becomes reachable at premium quality.

Play 2 - Demand aggregation: trust routing for the never-served. At the base, wealth is not a portfolio. It is a credit score, a loan, gold, insurance, a savings account. This user arrives with ambiguous intent - no deadline, no portfolio to diagnose, no job they can name. Acquisition cannot begin with a diagnosis, because there is nothing to diagnose.

So the wedge must meet their definition of wealth, and the best-proven one is the credit score: 183 million Indians self-monitor their CIBIL score, 75% of them outside the metros. Paisabazaar built its engine on 60 million free credit-score users who now drive 70-75% of loan disbursals as repeat customers; Jar proved the gold version at 35 million users, over 95% first-time savers, profitable. The AI-native version is conversational - PAN in, score out, then a surface where the user asks what wealth means for them, in their own language. Value capture is demand aggregation: accumulate trust, then route it into loans, insurance, gold, eventually funds. Our read: the hard part is not the routing. It is holding the trust while you do it.

Play 3 - Execution delegation: closing the behavior gap. In the middle sits the user this piece opened with, and the standard framing needs one correction. Her binding constraint is not advice. It is execution. She watches the content, holds the opinions, already knows the diagnosis, and still cannot act. The 43-fund portfolio nobody exits. The SIP nobody rebalances.

The cost is quantified: Axis Mutual Fund found equity funds returned 18.7% CAGR while the average investor in those same funds earned 13.2% - a 5.5-point behavior gap from timing errors, churn and panic exits.

That gap is the product. An AI agent here does not need to beat the market; it needs to beat the user’s own execution. Execution alpha, not information alpha - a categorically easier claim to deliver and to regulate. The ladder is a sequence of delegation proof points: diagnose, recommend, execute-with-approval, standing mandate.

Two overlays. The NRI diaspora - 3.7 crore overseas Indians - is a context, not a tier: remittance, India tax filing, inherited property, parent care. It attaches to any play. And household mobility connects the tiers over time - today’s gold saver is 2032’s mass-affluent professional, whose parents’ succession event is the largest wealth transfer in the family’s history. The platform that owns the transition between definitions of wealth captures the migration every static segmentation misses. That is what a household-level financial graph is ultimately for.

Part III - When the user actually shows up

Segmentation tells you who to build for. It does not tell you when they are reachable, and in wealth, that second question decides acquisition entirely. Households do not switch providers because a better product exists. They switch when something happens.

Mapping those moments against the pyramid produces a sharper picture than any AUM band:

Three things fall out.

The middle switches on life events, not returns. None of those triggers is a financial-product event; all are calendar events a system holding a household graph can anticipate. The graph is usually defended on retention grounds. It is really an acquisition asset - you do not need to outbid Groww on keywords if you know a user’s second child is due.

Every trigger at the base is promotional. Cashback, a bank push, social proof, an employer nudge. Not one is a product-quality trigger, so the channel that demonstrably acquires the never-served user has the worst retention economics. That is the real argument for the credit-score wedge: it is the only non-promotional door into that tier.

Play 3 is two form factors. The emerging-HNI user wants a scheduled expert call; the mass-affluent user wants autopay. Same constraint, different cost structure - and that discontinuity is where a Play 3 company either finds software economics or discovers it has built a services business.

Part IV - What you would have to believe

Three plays are three companies, so one scoreboard will not do. Each measures a different word in the loop: Play 1, whether insight got cheap enough to produce at scale; Play 3, whether insight converts to action; Play 2, whether action accumulates into trust rather than spending it.

What would prove us wrong. Three things, in ascending order of damage. If households per RM does not move within twenty-four months, Play 1 is a services business wearing software pricing - and note that Savvy and Farther both publish AUM growth and conspicuously not revenue-per-advisor deltas, so the productivity claim is unproven even in the market that pioneered it. If Play 3 users stall permanently at execute-with-approval and never grant a standing mandate, the delegation ladder has a missing rung and the AUM relationship that pays for everything never arrives. And the largest risk sits under the whole thesis: it assumes households eventually punish misaligned advice. Indian consumers have tolerated mis-selling for thirty years without moving their money. If alignment turns out to be something users say they want and never pay for, the independent's moat is a preference, not a structure - and the brokers win on distribution exactly as they always have.

Part V - What it looks like when it works

July 28th, 11 pm. A product manager uploads her Form 16. Ninety seconds later her return is drafted - the least of it. The agent has also noticed ESOP capital gains that change her advance-tax position, ₹3.2 lakh idling at 3% against a personal loan at 14%, and two ELSS funds bought for a deduction she no longer claims. It files, then asks one question: your loan is costing you ₹34,000 a year more than your idle cash earns - want me to set up the prepayment? That second permission is the whole business.

The claim. A user’s father is hospitalized. The agent already holds the family’s policies. Within the hour: pre-authorization drafted, exclusions checked, the room-rent sub-limit flagged before the family chooses a room - the clause that quietly shreds thousands of claims. When the insurer pays 71%, it drafts the escalation with the policy language cited. One founder told us about a user who, facing surgery with no one to call, fed his policy into an LLM just to understand his coverage - and came out feeling more secure than any agent had made him feel. Security manufactured at the moment of need is the deepest trust event in consumer finance.

Note what is absent from both: a stock tip. The AI never predicts the market. Everything it does is deterministic or near-deterministic - leakage, structure, protection, process, productivity.

Part VI - Where new company formation opens up

Part III mapped when a household changes provider. There is a second trigger, and it is the more valuable one: when a household changes what it holds. A distributor switch moves share between existing players. A form-factor switch creates a category.

Those transitions are legible per tier. The top of the pyramid moves into private credit, global custody, AIFs and estate structures - it is buying access and orchestration. The middle moves from single SIPs into goal baskets, tax-aware rebalancing, PMS-lite and loans against securities - it is buying transparency plus human backup. The base moves from a savings account or FD into a micro-SIP, a liquid fund, gold - it is buying confidence, defaults, and a frictionless first step.

Five companies sit on those transitions.

The first four are applications. The fifth is what they all leak into, and it is the only one whose asset is permission rather than product. A broker owns an account, an insurer a policy, a tax platform a filing history. The household graph is the first thing in Indian finance that can affordably own the household - and a competitor cannot copy a permission it was never granted.

Which is also the answer to the obvious objection, that Groww or CRED will simply ship all of this. They will bolt on AI, and they start with distribution independents cannot buy. But transaction platforms carry a structural conflict into advisory: their revenue is activity, and honest advice frequently means less of it. An AI adviser owned by a broker is a chatbot with a conflict of interest.

Can one company run two plays? The question every founder reading this will ask, and our own argument forces it: the Play 1 endgame - an RM serving 100 families instead of 33 - is Play 3’s customer, reached from the supply side. So the plays converge downward on the emerging-HNI tier from opposite directions, one carrying a human and one trying not to. Our read is that the convergence is real but the sequencing is not optional. Play 2 and Play 3 can share a company, because they share a consent model and a graph; a user acquired on a credit score can be graduated into execution delegation on the same permissions. Play 1 cannot be bolted on, because acquiring RMs is a recruiting business with a balance sheet, not a product motion. Anyone pitching all three at seed is pitching a holding company.

The professional we opened with does not need a better dashboard telling her she owns 43 funds. She needs something that knows which seventeen overlap, is allowed to exit them, remembers why it did, and is there in eighteen months when her father is hospitalised, or her portfolio crosses ₹50 lakh. That is one system, not three products.

The winning companies will not be the ones that generate the most recommendations. They will be the ones a household - or a relationship manager - trusts to remember, to act, and sometimes to leave things alone.

India AI News Roundup

The most important AI developments shaping India’s market structure this month.

Sarvam’s board approves a $74M Series B extension led by NVIDIA at a $1.51B valuation

RBI’s draft Model Risk Management guidance puts every AI model under a board-owned framework

HDFC Bank ships ‘Neev’, an in-house enterprise GenAI platform, with wealth management named as a use case

Emergent becomes a unicorn with a $130M Series C led by Creaegis, thirteen months after launch

IndiaAI Mission picks 20 sovereign models - 12 LLMs and 8 SLMs - with 93 lakh GPU hours sanctioned

Startup Signals

Early AI companies from India we are watching because they point to where the market may be forming.

Oolka: An agent that works your credit file, not your questions

Oolka deploys AI agents that act on a user’s credit profile rather than reporting on it - closing dormant accounts, disputing bureau errors, negotiating rates with lenders on the user’s behalf. Founded 2024 by a former Meesho executive; $14M Series A led by Accel at ~₹730 crore post, six million users, roughly a hundred regional languages.

Why it matters: every credit app in India tells you what is wrong and almost none fixes it. If the first experience of financial help is a completed action rather than a diagnosis, the trust ladder starts a rung higher than the category assumed.

GoodScore: The same wedge, sold as a subscription

GoodScore attacks the identical entry point with the opposite model - AI-led guidance, expert calls, a lending marketplace underneath. $13M Series A led by Peak XV, five million-plus users, ₹6,200 crore of loans under management, and ₹99 a month.

Why it matters: one monetises the action, the other the advice. A base-of-pyramid platform charging directly for guidance is the sharpest live test of whether trust can be sold, or only brokered.

thrive.money: Read-only diagnosis, one authorised action

thrive.money is pre-launch and waitlist-only, and still the crispest statement of Play 3 we have read: connect read-only, surface what your money quietly costs you, execute the single action you choose. The homepage leads with “investment adviser registration pending” rather than a growth number.

Why it matters: sequencing the licence before the user base is expensive and slow, and the only order that works if the endpoint is a standing mandate.

Draconic: Conviction as the product, not execution

Draconic runs specialised agents across price action, options flow and sentiment, resolving into a documented reason for every trade. India-built, globally sold across thirty-plus markets, backed by WEH Ventures.

Why it matters: it states plainly that it is not a registered adviser, and that constraint shapes the product rather than merely disclaiming it. Selling reasoning the user acts on themselves is the ceiling on what an unlicensed AI wealth product can capture.

FinStocks AI: Strategies you can read before you run them

FinStocks AI turns a prompt into a strategy the user can read, edit, and mock-test, then run on their own broker account. Custody never leaves the broker; the stop stays with the user. Early, no disclosed funding.

Why it matters: the bet is that the answer to black-box scepticism is legibility rather than track record. Whether a readable strategy survives its first drawdown is the question the whole category owes an answer to.

Two things are happening inside India’s consumer platforms at once, and they pull in opposite directions.

The first is compression. Support, cataloging, onboarding, QA, analytics, coding, creative production, invoice review, hiring screens, warranty triage, and internal operations are getting cheaper. Work that used to require a queue, a dashboard, a team, or a vendor can now be delegated to a model, an agent, or a small internal team. Every workflow that is repetitive, measurable, and trapped inside a bad interface is being rethought.

The second is expansion. AI is not just making marketplaces cheaper to run. It is pulling new users, sellers, products, categories, and transactions into the market. Voice can reach consumers who never became fluent in e-commerce search. Catalog agents can onboard sellers who never had clean product data. Recommendation systems can create demand in categories where the user does not know what to ask for. AI-generated creative can make millions of micro-campaigns economically possible.

The old marketplace question was: can you aggregate enough demand and supply to make the flywheel spin? The new marketplace question is: can you convert messy intent, messy supply, and messy offline reality into a governed transaction better than anyone else?

But the most important shift is subtler than both. The consumer interface is not converging on a single universal chat box. It is fragmenting by category behaviour. For transactional needs, the user wants the answer. For expressive categories, the user wants to shop. That distinction sounds small. It decides whether an AI interface becomes the marketplace or another failed experiment sitting alongside search, feed, and filters.

The tempting thesis is that AI agents become the new consumer front door and marketplaces get reduced to commodity fulfillment layers. That will be true in some categories. It will be false in others.

The AI Marketplace Matrix

The easiest mistake is to treat marketplaces as one category. A food-delivery marketplace, a fashion marketplace, a quick-commerce app, an eyewear retailer, a travel aggregator, a services marketplace, and a grocery platform may all look like consumer internet businesses but structurally they are different machines.

The useful map has two axes:

• Does the user want a definitive answer or an exploratory journey?

• Is the transaction mostly informational or physically constrained?

Each quadrant has a different AI endgame. In answer-seeking categories, AI can collapse the funnel. The agent asks the right questions, reasons through the options, and presents a small set of choices. The old browse/search interface loses power because the user did not want to browse in the first place.

In exploratory categories, AI cannot simply compress the journey into three recommendations. The act of shopping is part of the product. The user wants to compare, wander, refine taste, react visually, and feel agency. In these categories, AI has to enrich discovery without killing the pleasure of exploration.

In physically constrained categories, the interface matters, but the operating system matters more. Availability, density, ETA accuracy, substitutions, returns, trust, price, freshness, support, and fulfilment determine whether the transaction actually works. AI can change the front door. It cannot wish the delivery network, supply quality, or local operating memory into existence.

What Marketplaces Actually Are

A marketplace is four businesses hidden under one brand.

First is the demand engine: acquisition, activation, search, discovery, personalisation, conversion, cross-sell, support, and habit formation. This is the visible layer. It is also where consumer AI gets most of the attention.

Second is the supply engine: seller onboarding, merchant tooling, catalog ingestion, content generation, pricing, inventory quality, ad spend, availability, reliability, and retention. This layer is less visible, but it is often where liquidity is actually created.

Third is the market-making core: ranking, matching, recommendations, trust, fraud, logistics, pricing, demand shaping, movement planning, forecasting, and fulfillment orchestration. This is the compounding layer. It decides whether the marketplace is merely a directory or a machine that can create transactions.

Fourth is the operating system: customer support, finance, audit, warranty, hiring, vendor reconciliation, ERP/WMS integrations, warehouse workflows, store workflows, analytics, and developer productivity. This is where AI releases margin, but also where the ugly truth of enterprise systems shows up.

AI touches all four. But it does not touch them evenly. On the demand side, AI changes the interface from browse to intent. On the supply side, AI changes the bottleneck from seller acquisition to seller readiness. In the market-making core, AI changes matching from static rules to adaptive models. In the operating system, AI changes marketplaces from human-process businesses into agent-supervised execution systems.

The best companies are asking which layer controls the transaction, and where can AI move that control point?

Demand: Chat Is Not The Universal Commerce UI

The first wave of commerce chatbots taught the wrong lesson. Many did not fail because the model was too weak. They failed because chat was the wrong interface for the category.

Chat works when the user wants a definitive answer. Find me a flight under this budget. Reorder this medicine. Which policy covers this claim? Which product should I buy for this use case? The user has a job and wants resolution. Chat is weaker when the user enjoys the act of browsing. If someone is buying a saree, a t-shirt, a frame, a lamp, a lipstick, or a piece of decor, three perfect recommendations may still be unsatisfying. The point is not only to complete the transaction. The point is to explore taste. That means the future of AI commerce is not simply conversational shopping. It is interface-market fit.

AI discovery matters most where the user is still forming intent. Fashion, beauty, eyewear, home decor, travel, and other low-repeat or high-variety categories can benefit meaningfully from AI-assisted discovery. The consumer does not merely want a faster search box. They need help forming the choice without losing the pleasure of choosing. A food-delivery marketplace is a useful counterexample. In high-frequency, logistics-heavy categories, the moat remains availability, reliability, density, pricing, substitutions, support, and fulfillment. The interface can change; the operating system still has to work.

Voice is part of the same story. In markets where the interface itself has been the bottleneck, voice can unlock users who were never fully served by text-first commerce. But voice is not one product. In production, it often becomes a system of choices: ASR, reasoning, TTS, latency, language coverage, tool use, escalation, and cost. Cascaded voice systems win where control and composability matter. Speech-to-speech models will matter where emotion, latency, and natural turn-taking are the product.

So the AI threat is not one generic shopping agent. It is category-specific. In some markets, the AI front door becomes the marketplace. In others, it becomes a thin layer on top of the real moat.

Supply: The Catalog Becomes The API

Most marketplace conversations over-index on demand because consumers are visible. But many of the hardest marketplace bottlenecks are supply-side. Can the seller onboard without a human ops team? Can messy catalog files become structured SKUs Can bad images be repaired or rejected? Can the platform infer attributes, categories, styles, variants, compliance requirements, and substitutions from incomplete seller data? Can the seller understand which products to push, which ads to run, which inventory to refresh, and which demand signals matter? Can a local merchant participate in a digital marketplace without becoming a systems integrator?

This is where AI becomes less glamorous and more important. A consumer marketplace is only as good as the structured representation of its supply. If the catalog is wrong, every downstream model is wrong: search, recommendations, pricing, availability, ad targeting, personalisation, and support. The bottleneck is not just extraction. It is normalisation, enrichment, deduplication, image quality, SKU mapping, category inference, attribute completion, compliance, and continuous refresh.

The catalog is becoming the API for supply. This also changes seller retention. A seller agent that improves catalog quality, recommends content, manages ad budgets, explains demand, and reduces operational work becomes more than a tool. It becomes the seller’s interface to the marketplace. If the consumer front door is where AI gets attention, the seller cockpit may be where value quietly compounds.

Core: The Recommender Stack Is Being Rewritten

The strongest marketplaces have never been mere aggregators. They are market-making systems. They decide what to show, when to show it, which supply to allocate, how to price, how to forecast demand, how to place inventory, how to route jobs, how to detect fraud, when to subsidize, when to intervene, and how to trade off short-term conversion against long-term liquidity. AI makes that layer deeper. The architectural shift is from cascaded recommenders to generative recommenders.

The old stack split the work into retrieval, ranking, re-ranking, business rules, and objective-specific models. The emerging stack tries to collapse more of that into larger transformer-based systems, often using semantic representations of products, users, and actions as the vocabulary of recommendation. Instead of many small models passing candidates down a pipeline, the direction of travel is toward fewer, larger systems that can reason across products, users, objectives, and context. This does not mean every marketplace model becomes a generic API call. The model is not the moat. The production loop is.

The moat is the combination of proprietary marketplace data, feedback loops, local operating constraints, transaction outcomes, and the ability to measure whether a model actually improved conversion, availability, ROAS, CSAT, attach rate, fraud loss, returns, inventory turns, or fulfillment quality. The more physical the marketplace, the more important this becomes.

In a pure digital marketplace, AI can crawl, compare, recommend, and transact with fewer real-world dependencies. In physical marketplaces, every recommendation has an operational shadow. Inventory may be wrong. Delivery may fail. A product may be unavailable. A return may be expensive. A vendor may be unreliable. A store may be crowded. A customer may churn after one bad experience.

The agent can shape demand. The marketplace still has to fulfill it.

Operations: Marketplaces Become Agent-Supervised Companies

Marketplaces contain a shocking amount of internal bureaucracy. Support teams jump across old dashboards. Finance teams reconcile vendor systems, WMS, ERP, and revenue recognition. Audit teams review invoices and exception patterns. Warranty teams decide claims. Hiring teams screen applicants. Analysts generate reports. Developers write and maintain internal tools. Ops teams move information between systems that were never designed to talk to each other.

This is the labor-heavy underside of consumer internet. AI attacks it directly. Support is the obvious first wave because the workflow is measurable: cost per interaction, resolution rate, CSAT, response time, escalation rate. When token costs fall and quality improves, the ROI becomes obvious.

But support is only the beginning. Warranty decisions, fraud checks, invoice review, candidate screening, ledger reconciliation, store operations, catalog refresh, ad creative generation, campaign analysis, and developer productivity all sit inside the same pattern: high-volume knowledge work trapped between legacy systems and human queues.

The more interesting internal use case is not coding productivity by itself. It is turning non-engineering operators into builders. When business owners, category managers, analysts, ops leaders, and P&L owners can build their own tools, the organization changes shape. The gain is not only individual productivity. It is lower coordination cost. A single person can understand the KPI, inspect the data, build the workflow, test the fix, and own the outcome. That collapses the handoff between business, product, data, and engineering.

In marketplace operations, this matters enormously. The person closest to the exception often knows what needs to change but historically lacked the technical leverage to change it. AI gives that operator a builder persona. The old operating model was a human process plus software. The new one is a software process plus human escalation.

The Disruption Vectors

Five mechanics are doing the actual work.

Vector 1: The interface splits by category behavior

The app is no longer the only consumer interface. A user can speak, describe, upload an image, show a preference, ask for a bundle, or delegate a task. The marketplace that understands intent best gets the first shot at the transaction.

But the interface does not always become chat. In answer-seeking categories, the agent can collapse the journey. In taste-led categories, the agent has to enrich discovery without killing the joy of exploration. In logistics-led categories, the best AI may never be visible to the consumer at all.

The risk to incumbents is not that a model knows more about commerce. It is that the model becomes the user’s default interpreter of intent in the categories where intent is the bottleneck.

Vector 2: Supply becomes machine-readable

AI does not work on “supply.” It works on structured representations of supply. A marketplace with clean catalogs, rich attributes, good images, reliable availability, historical interaction data, and transaction feedback has a compounding advantage. A marketplace with messy PDFs, poor images, inconsistent attributes, duplicate SKUs, and vendor-specific formats has an AI tax.

This is why catalog automation is strategically underrated. The company that makes supply machine-readable can sit upstream of search, recommendations, ads, personalization, seller success, and marketplace liquidity.

Vector 3: Model strategy splits by value, volume, and maturity

The winning model stack is not simply frontier or open source. It is a router. Frontier models dominate exploration, long-tail use cases, multimodal reasoning, complex review, and high-value workflows where quality matters more than unit cost. Open-weight or self-hosted models absorb high-volume, mature, accuracy-critical workloads once the task is well understood and the economics matter.

This is why the answer to “open or closed?” depends on what you are measuring.

The count of use cases may be frontier-heavy because experimentation starts on the most capable models. Request volume may be open-heavy because mature, high-scale workflows eventually move toward cheaper or fine-tuned systems. Economic spend can sit between the two, depending on where quality still commands a premium.

That means AI infrastructure decisions will increasingly be made at the workload level, not the company level. The model is not the strategy. Allocation is the strategy.

Vector 4: ROI governance replaces AI theater

The first wave of AI adoption rewarded experimentation. The second wave will reward accounting Marketplaces are starting to ask a harder question: which AI use cases create value, which merely create activity, and which looked good only because nobody was measuring the full cost?

The cleanest ROI cases are still traditional business cases with AI inside them. ROAS improves. Cost per support interaction falls. Activation rises. Conversion improves. Fraud loss decreases. Hiring throughput increases. Warranty turnaround falls. Developer velocity improves. Team size ratios change.

The hardest cases are internal productivity tools where benefit is real but attribution is fuzzy. Coding agents, analyst copilots, and general employee automations can feel obviously useful, yet still require better measurement: shipping velocity, maintenance burden, discoverability, rework, quality, and total cost.

This is where many companies will discover that AI spend is not the same as token spend. The cost base includes GPU infrastructure, ranking models, video processing, voice, multimodal pipelines, vendor contracts, cloud commitments, self-hosting, observability, evals, safety, and people needed to maintain the system.

The CFO will not care that the demo was impressive. They will care whether the workflow creates revenue, reduces cost, reduces risk, or changes the operating leverage of the company.

Vector 5: Physical marketplaces are defended by operational memory

The most important distinction in marketplace AI is physical coupling. This is why a food-delivery marketplace behaves so differently from a low-repeat fashion or travel marketplace. In one category, AI may reshape discovery. In the other, discovery is only one small part of the transaction.

AI can reason across information markets quickly. It can compare digital goods, recommend products, summarize options, generate content, and automate coordination. But in physical-world marketplaces, the transaction is only partly informational.

The rest is local density, trust, vendor reliability, logistics, returns, inventory, payments, compliance, support, and operating history. That does not make incumbents invulnerable. It makes the battle more specific.

The AI-native challenger has to decide whether it is attacking the front door, the supply cockpit, the market-making core, or the operating system. Each requires a different product, GTM motion, data asset, and moat.

The incumbent has to decide whether it is using AI as a feature, a cost program, or a new control point. Both extremes are wrong. AI will not kill marketplaces wholesale. But marketplaces that are only traffic aggregators will be exposed.

Efficiency Tokens And Value Tokens

The most useful way to think about AI spend in marketplaces is to split it into two pools. Efficiency tokens reduce cost. Value tokens create revenue.

Efficiency tokens show up in support, code, QA, analytics, catalog processing, invoice review, hiring screens, warranty triage, and internal workflows. These use cases matter because they release margin. But over time, they become cost-sensitive. Once a workflow is stable, high-volume, and measurable, the pressure to move it to cheaper models, open-weight systems, caching, fine-tunes, or self-hosting increases.

Value tokens show up in activation, conversion, personalization, recommendations, ad performance, seller ad budget growth, demand shaping, cross-sell, category expansion, and new user unlocks. These use cases justify frontier spend longer because the upside is tied to GMV, take rate, frequency, or ad revenue.

That split explains why AI budgets can look irrational from the outside. A company may aggressively optimize token cost in support while spending heavily on multimodal discovery. It may self-host one workload and pay frontier prices for another. It may use open models for volume and closed models for judgment. It may appear model-agnostic because the real strategy is not allegiance to a model vendor. It is margin-aware routing across workflows.

This is also where the marketplace AI stack becomes investable. Every large marketplace will need cost attribution, evals, routing, quality measurement, prompt/version control, data governance, model monitoring, and business-outcome linkage across dozens or hundreds of AI workflows.

Where Potential New Company Formation Opens Up

If AI compresses parts of the old marketplace model, where do new companies get built?

We think the opportunity is in the infrastructure and workflow layers where models meet marketplace reality: messy supply, multilingual demand, physical fulfillment, ROI measurement, and high-volume operations.

The common thread is not “AI for commerce”. It is workflow ownership. If you are demand-side, prove activation, conversion, frequency, or GMV lift. If you are supply-side, prove onboarding speed, catalog quality, seller revenue, ad efficiency, or supply liquidity.

If you are core marketplace infrastructure, prove matching quality, attach rate, fraud reduction, forecasting accuracy, inventory turns, or contribution margin. If you are on the operations side, prove cost per workflow, turnaround time, accuracy, escalation rate, auditability, and labor leverage.

The startup opportunity is not evenly distributed. The most glamorous layer may not be the most defensible. The most defensible layer may not be the easiest wedge. And the biggest spend may not produce the best venture outcome if it is only cost compression with no control point.

India AI News Roundup

The most important AI developments shaping India’s market structure this week.

Google picks 20 AI startups for India Accelerator 2026 cohort

HCLTech lands a $1.1B AI-led enterprise deal, reportedly with Mercedes-Benz

Dell pushes sovereign AI storage in India with PowerStore Elite

ESDS launches Swaraj Cloud, an India-operated sovereign AI cloud platform

Nurix AI acquires Verloop.io to strengthen enterprise conversational AI platform

Startup Signals

Early AI companies from India we are watching because they point to where the market may be forming.

Observal: The system of record for AI coding agents

Observal is an open-source registry and analytics platform for AI coding agents. It tracks skills, MCP servers, hooks, prompts, sandboxes, components, agents, session traces, and usage across tools like Claude Code, Cursor, Gemini CLI, VS Code, and GitHub Copilot.

The insight is simple: AI coding agents are becoming part of the software supply chain, but most teams do not yet have a clean inventory of what agents are using, what components they depend on, or which agent workflows are actually helping. If engineering work increasingly happens through agents, then agent components need versioning, observability, and governance.

Why it matters: the next developer platform may not just manage code. It may manage the agents, tools, prompts, and reusable skills that produce the code.

Shadow Labs: Agents that execute while the call is still happening

Shadow Labs is building an AI coworker for sales and customer-facing teams, moving beyond the now-crowded “meeting summary” layer into real-time execution. The sharper wedge: the call itself becomes the prompt, and background agents can pull context, update systems, draft follow-ups, and move work forward before the conversation ends.

Why it matters: enterprise AI is shifting from passive copilots to action-taking operators. The meeting stack is a natural insertion point because it sits closest to customer intent, but the durable company here is not the best transcriber. It is the one that owns the workflow after intent is captured.

Dreabee: Creator intelligence for influencer marketing

Dreabee is building a creator intelligence platform for brands and agencies, positioning itself as a more data-driven decision layer for influencer marketing. The wedge is simple: as creator spend grows, teams need better ways to discover creators, underwrite audience quality, benchmark performance, and allocate campaign budgets.

Why it matters: influencer marketing is still too taste-led and relationship-led for the amount of money now flowing through it. The opportunity is to turn creator selection into something closer to a research, underwriting, and measurement workflow. If AI makes supply discovery cheaper, the scarce layer becomes trust, signal quality, and decisioning.

The first is deflation. Every hour of work an agent can now do - a migrated codebase, a resolved L1 ticket, a reconciled ledger - is what the market stops paying full freight for. TCS shed roughly 28,550 people in nine months while revenue per employee rose 3.6%. Persistent’s renewal TCV dropped 10.9% in a single quarter, entirely in contracts.

The second is creation, and it is the Jevons paradox doing its familiar work: when the unit cost of a capability collapses, total consumption of it rises. Cheaper, faster delivery does not shrink the market. It pulls forward demand that never had an economic path before. Infosys’s subcontractor spend grew 19.2% year-on-year, twice the rate of internal staff costs. Wipro booked $7.8B in large deals, up 45.4%, while current-quarter revenue fell 1.6%. Record bookings against declining revenue is the signature of an industry mid-transition.

So the map is not shrinking. It is being redrawn.

To see where the value goes, you first have to read the map honestly: IT services is not one market, and AI is not pressuring it evenly.

The AI Disruption Matrix for IT Services

The easiest way to understand the sector is not as one monolith, but as five layers sitting on two different gradients: defensibility and startup opportunity.

Deep product engineering stays defensible because the work is dense with tacit context. Managed services come under maximum pressure because the work is repetitive, ticketed, measurable, and increasingly agent-addressable. Data, evals, and governance are the opposite - it is not being disrupted as much as it is being created.

The market is not collapsing. The margin pool is moving. The subtle point is that the most defensible segment is not necessarily the best startup opportunity.

Deep product engineering is protected, but that same tacit-knowledge moat makes it hard for new companies to enter. Managed services are less defensible, but that is exactly why margin gets released. Data, evals, and governance are still emerging, which makes it a new-company formation zone rather than a replacement market.

What IT Services Actually Are Today

Treat IT services as one monolith, and you miss everything. It is five distinct pools, and they sit on a steep gradient of defensibility.

At the most defensible end is deep product engineering: the shops building storage firmware, virtualization platform integration, networking stacks, and complex infrastructure software for global technology companies. This work demands fluency in proprietary architectures, multi-year product roadmaps, and the institutional context that lives in engineers’ heads, not in any public training corpus.

Vertical consulting and transformation comes next: designing the claims workflow, the reconciliation flow, the prior-auth process, the manufacturing exception loop, and the telecom support path. This is roughly 8-12% of IT spend, but with a high margin.

Implementation and migration - cloud migration, ERP/CRM rollouts, modernization, testing, integration - is the mid-margin middle, around 25-30% of spend.

Managed services - L1/L2 support, ticket resolution, QA, infrastructure ops, helpdesk, process operations - is the largest pool at 40-45% of spend, and the most volume-dependent.

Data, eval, and governance - model evaluation, red-teaming, domain training data, AI assurance, agent monitoring - is smaller today but growing fastest, around 15-20% and rising.

Where Each Piece is Under Attack

The competitive pressure hits each pool from a different angle, and the angle is what determines survival.

Deep product engineering: protected by tacit knowledge

Deep product engineering faces the least pressure. An agent can help an engineer write code faster here; it cannot replace the engineer who knows how a hypervisor change cascades through a storage stack.

The moat is tacit knowledge, which is by definition the part that never made it into the model’s weights. This is the last thing to be disrupted, not the first.

Vertical consulting: squeezed by frontier labs

Vertical consulting is squeezed from an unexpected direction: the frontier labs themselves.

When OpenAI stands up a deployment arm to embed engineers and redesign enterprise workflows, it is competing for exactly this layer. Generic “AI strategy” commoditizes. Only a proprietary vertical playbook holds.

Implementation and migration: compressed by accelerators, hollowed by GCCs

Implementation is compressed by AI accelerators and hollowed by GCC insourcing.

But it does not vanish. It re-prices. Margin shifts to whoever owns reusable delivery IP: connectors, migration playbooks, testing harnesses, eval suites, agent orchestration, and integration depth.

Managed services: hit from both sides

Managed services is the most contested pool, hit simultaneously by AI agents and by GCCs pulling the work in-house.

This is where renewal compression originates. The client does not have to believe in a full AI replacement story. It only has to believe that repetitive tickets, alerts, reconciliations, support flows, and exception-handling work should no longer be priced as if every unit requires the same human effort.

Data, eval, and governance: the greenfield layer

Data, eval, and governance face almost no incumbent competition because the category is still being formed.

Neither IT giants nor GCCs fully own it yet. But as copilots become agents, every regulated enterprise needs model QA, evals, audit trails, monitoring, compliance evidence, and assurance infrastructure.

This layer is not a cost takeout story. It is a market creation story.

The Disruption Vectors

Five mechanics are doing the actual work. They cut across every pool above.

Vector 1: Agents are run-time software, not build-time software

Traditional enterprise software is a build-time asset: you spend the effort once, and it runs at near-zero marginal cost. The margin sits in the build, the same logic that lets SaaS write code once and sell it ten thousand times.

Agents invert this. The model is pre-built and increasingly commoditized. The value shows up at run-time, in the messy, per-customer act of pointing a general capability at one enterprise’s specific reality: legacy systems, access controls, compliance constraints, bad data, human approvals.

None of that compresses into a one-time build. It is continuous, contextual, local - which is to say, it looks like services.

This is also why deep product engineering is hardest to disrupt: the value there is precisely the run-time judgment that lives in the engineer. The thesis and its strongest objection are the same observation pointed in two ways.

When computing moved to the cloud, the margin did not go to whoever owned the server. It went to whoever could operate the workload against a specific stack. Agents run that play one level up.

Vector 2: Pricing splits along a bespoke/commodity line

The Valley reflex is “everything becomes outcomes-as-a-service.” The reality is more textured.

Outcome pricing wins where work is scoped, measurable, and repetitive: resolved tickets, processed claims, completed reconciliations. There it compounds. Price the outcome first in a vertical, accumulate data on what resolution actually costs, and pull away.

But in high-tech product engineering, engineering managers prefer time-and-materials because it preserves the flexibility to reallocate resources to whatever the roadmap makes urgent this week.

When the scope is genuinely uncertain, and the client is buying senior judgment, the option value of flexibility beats the predictability of a fixed price.

The endgame is not one winner. It is a split that mirrors how law and consulting firms already bill: hours for the bespoke, fixed-fee, and per-outcome for the commodified.

Vector 3: The delivery pyramid collapses into a partnership

The classic pyramid kept a delivery-manager layer between the architect and the team for a very human reason: architects do not want to be interrupted while writing the hard code, so someone else ran the people.

Agents dissolve that logic. The architect now delegates delivery to agents or runs delivery through them. The middle thins out.

What is left looks more like a law firm: a senior architect-partner who understands the system, manages the agentic team, and sells to the client. The three roles the old model deliberately separated collapse into one person with leverage.

The talent question flips from “can we hire enough engineers?” to “can we hire people who can make agents productive and own a client outcome?”

Vector 4: The bigger threat is the GCC, augmented by AI

Call AI the thing killing IT services, and you have named the weapon, not the hand.

The more direct threat is the in-house Global Capability Center. If incumbents get hollowed, it will be GCCs augmented with AI that do it, not AI alone.

The labs moving downstream confirm where the value is. In May 2026, OpenAI launched the OpenAI Deployment Company, a unit embedding Forward Deployed Engineers inside enterprises. A model company built a services arm. The model alone is not the product; deployment is.

Vector 5: Timing, and whether incumbents have time to adapt

Token costs are not zero. In the near term they are a real line item, and the implicit bet behind every “AI eats services” thesis is that the industry will not have time to evolve.

That bet is probably wrong.

No enterprise rips out its vendors overnight. Switching is slow, trust is sticky, and the binding constraint is human readiness, not model capability.

There is an enormous capability overhang right now. Most enterprises have little idea what these models can do, which means they cannot self-serve to value even with cheap tokens. That overhang is the services TAM.

But “survive” should not be read as “stay the same.” Every prior shift was labor-expanding. This is the first that is labor-substituting. The adaptable players will make it. They will just emerge as structurally different businesses.

The Five Wedges

That leaves the real founder question: if AI compresses the old services model, where does new company formation open up?

Our view is that the opportunity is not in generic “AI transformation,” but in the deployment layer where models meet enterprise reality: GCCs, regulated workflows, permissions, evals, and repeatable delivery systems.

The common thread across these wedges is specificity. The best companies will not sell AI demos or cheaper hours; they will turn messy enterprise work into governed, outcome-priced workflows, using India’s implementation depth as the wedge into global demand.

One Corollary for Founders

The most defensible layer, deep product engineering, is also the least open to new entrants. The same tacit-knowledge moat that protects it from AI protects it from challengers. Low disruption and low new-company formation are the same coin.

The disruption and the opportunity both concentrate in the commodity and greenfield layers. Which is also why we are not in the “IT services is dead” camp. The adaptable incumbents and mid-caps, with distribution, domain trust, and switching-cost protection, will ride the transition.

The new company opportunity sits beside them, in the layers they are slowest to own.

The Bottom Line

AI does not kill IT services. It collapses the delivery pyramid, splits pricing along the bespoke/commodity line, and converges the whole industry on professional-services economics.

The margin trapped in headcount-heavy managed services and implementation is being released. The firms that capture it will not be better staffing businesses, but vertically deep, partner-led, agentic delivery firms that own workflow completion rather than workflow effort, operating as the intelligence layer inside enterprise GCCs.

The $98.4B GCC ecosystem is the distribution. The agent stack is the engine. Run-time is the real estate. And India, with the domain depth, the talent cost structure, the enterprise trust, and the GCC density already in place, is holding the deed.

If you are building in AI GCC enablement, agentic managed operations, evals, and governance, or vertical AI workflows in India, we would love to have a conversation to develop these ideas further.

India AI News Roundup

The most impactful AI developments & announcements shaping India in recent weeks.

Agentic platform Codex adoption in India up 27x since January, says OpenAI

India is now the world’s #2 Consumer AI market

Bajaj Finserv sets up ~$200M fund to invest in AI ventures

Anthropic sets up AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs

OpenAI launches the OpenAI Deployment Company to help businesses build around intelligence

Startup Signals

Spotlighting brand new emerging AI startups from India every month, early and undiscovered.

FirstOps: Runtime control for enterprise agents

FirstOps is building the control plane for AI agents at work: identity, access, and audit across MCP servers, tool calls, LLM calls, skills, and subagents. The wedge is simple but structurally important. Enterprises want to increase agent usage, but security teams need a way to see what agents are doing, scope their permissions, and create audit trails before production deployment becomes acceptable.

Observal: The system of record for agent adoption

Observal is an open-source registry and observability platform for AI coding agents. It tracks agent usage across tools like Claude Code, Cursor, Gemini CLI, VS Code, and GitHub Copilot, with traces, spend visibility, adoption metrics, and agent registry primitives.

Terrastack: Spatial intelligence for Indian credit

Terrastack is building location intelligence for land and property decisions in India. Its public product gives plot-level assessments and map exploration; its April 2026 newsletter describes digital screening for Loan Against Property and branch intelligence for BFSI institutions, combining satellite imagery, government records, cadastral data, market data, and catchment-level accessibility layers.

For most of the last two years, much of the conversation about Indian AI orbited around the model layer - whether Indian labs could build frontier-grade systems, who was funding them, how the cost curves would shake out. Those questions are still live, and companies like Sarvam are answering them in their own ways. But for the much larger universe of Indian AI builders sitting outside the foundation-model race, the practical reality on the ground is simpler. The model layer is being rented. Our recent stack survey of 244 Indian builders made this concrete: 74% run on closed proprietary APIs, 9% on Western open-weights, 3% on Chinese open-weights. For most builders, the model is no longer the bottleneck. It is the input.

The more interesting question, then, is what kinds of AI companies India can uniquely produce on top of that input. The answer, increasingly, looks like a voice layer purpose-built for messy multilingual business calls that no Silicon Valley product manager has ever sat next to. It looks like a software creation platform that turns natural-language intent into a production workflow rather than a one-shot demo. It looks like a storytelling system where serialised content is written, validated, localised, and stress-tested at machine speed against millions of listener journeys.

Over the last few weeks, we sat down with three founders building exactly this kind of company: Pocket FM, Rocket, and Bolna. Different categories. Different stages. Different go-to-market motions. But across the conversations, one architectural pattern kept recurring - each of them is using AI not to produce a single impressive output, but to collapse a bottleneck inside real workflows that nobody had previously owned cleanly.

The common thread is worth stating plainly. India’s AI winners are unlikely to win by owning the model. They will win by owning the workflow where the model becomes useful - the messy, sticky, distinctly local layer where intent meets execution and where every additional unit of usage hardens the moat.

What follows is a structured profile of each company. Same shape, three times: the facts, what they do, why they are unique, the big opportunity, the risks, and our final take.

Bolna: A Voice Operating Layer for Indian Business

What does Bolna do?

Bolna is a voice AI orchestration platform purpose-built for India’s telephony complexity. Enterprises deploy voice agents from their own transcripts and FAQs, with support for 10+ Indian languages, 50+ accents, sub-500ms latency, and the kind of code-switched conversations (Hinglish, English + Tamil, Hindi → Gujarati mid-sentence) that break global voice stacks. Customers go live in days, not the weeks a legacy IVR replacement typically takes.

The product is delivered through two motions: a self-serve platform for SMBs and modern startups (plug in a number, point at a CRM, run agents within minutes), and a forward-deployed engineering team for enterprise rollouts where voice workflows need to be embedded into core operations.

Why is it unique?

Bolna’s defining choice is that it is not a single-model company. Maitreya Wagh’s (CEO) position is direct: no foundational model can meet enterprise voice needs across India. Instead, Bolna runs an orchestration layer that picks the best-fit model for the language, accent, latency budget, and use case on every call - and re-routes as better models arrive. The model is rented. The orchestration is owned.

The founders describe four reinforcing moats underneath that orchestration.

1. Scale. Volume gives Bolna access to the underlying models at unit costs that new entrants cannot match.

2. Orchestration as a product surface. Different models route to different purposes within a single call, and the voice agent connects to CRM data, email, routing logic, and downstream workflow tools. This is not a chatbot. It is an agentic workflow embedded inside a real business process.

3. Tacit knowledge from running calls at India scale. How Indian users actually interrupt. Where latency breaks the experience. When prompts need to be rewritten. When to escalate to a human. None of this lives in a paper. It accumulates only by shipping.

4. Distribution via self-serve. A user plugs in their own number, completes verification, points at their CRM, and runs an agent against their own customers within minutes. In a category that otherwise feels procurement-heavy, that frictionless onboarding does as much work as the orchestration underneath.

The most India-specific insight, and the one no global competitor will reach without years in market: background noise is a semantic risk, not just an audio problem. Noise-cancellation models biased toward English can suppress real Indian speech because the system treats non-English audio as noise. That is a product truth Bolna had to learn by running calls.

The big opportunity

India runs over a billion voice calls a day across support, recruitment, logistics, onboarding, and collections. Most of it still flows through legacy IVR or human-led workflows that are expensive and slow to scale across languages.

The cost case is striking - a typical human call costs ₹6-7 in raw spend, ₹13-14 fully loaded; Bolna’s AI calls drop as low as ₹2. But Maitreya is careful that the better argument is structural rather than cost-based: AI can fetch data from multiple systems mid-call, respond at consistent speed, and hold quality across high-volume workflows that human teams degrade on. The cost saving is a side-effect of the structural advantage.

They are also clear about where humans still win. Expensive purchases. Sensitive support. Hiring conversations. Anywhere trust is the deliverable. Voice AI shines in conversational information collection, reminders, qualification, follow-ups, and the long tail of repetitive business calls.

If Bolna becomes the orchestration layer Indian enterprises standardise on, the platform sits at the intersection of telephony, CRM, and agentic workflows. International optionality is real - 10+ countries already - but the right way to read that is as proof the orchestration logic generalises, not as a near-term GTM bet.

The risks

Three are worth holding honestly.

The first is competitive. Vapi, LiveKit, VoiceRun and others are building voice orchestration globally, and US-funded competitors will eventually arrive in India with better-capitalised playbooks. Bolna’s head-start has to convert into either default-distribution or deep integration moats before that happens.

The second is concentration. Most current customers are SMBs paying small ticket sizes. The founder has signed two large enterprises and has four more in pilot - real progress, but the path from current scale to a venture-scale outcome runs through enterprise. That motion is operationally heavier (forward-deployed engineers, custom implementations) and slower-compounding than self-serve.

The third is category-definition. Voice AI is still in flux as a category. Is it infrastructure (priced like Twilio), application software (priced like a SaaS seat), or services (priced per-call)? The choice shapes margins, GTM, and competitive set. The founders seem clear-eyed about this, but the answer will only emerge through the next eighteen months of contracts.

Our final take

The orchestration-over-single-model bet is the architectural choice we find most defensible in Indian voice AI today, and Bolna's commitment to that bet is what makes the company worth watching closely. Whether it converts into a durable position depends less on the orchestration logic itself, which is straightforwardly correct on its own terms, than on whether the four reinforcing layers underneath - scale economics, tacit operational knowledge, agentic workflow integration, and self-serve distribution - genuinely compound at the pace the early traction suggests. The next eighteen months will tell us whether the enterprise motion catches up to the self-serve, and whether the head-start translates into the kind of category-defining position that survives the eventual arrival of better-capitalised global competitors. The signals so far are encouraging. The verdict is not yet in.

Rocket: Product Creation as a Natural-Language Workflow

What does Rocket do?

Rocket.new turns a natural-language prompt into a production-ready application - landing pages, dashboards, internal tools, native mobile apps, full-stack SaaS. It is a vibe-coding platform in the same broad category as Lovable, Bolt, and Replit, but with a different architectural bet: it takes longer (~25 minutes) on the first generation in exchange for a much more complete output, and is built around the assumption that day one is the start of the product, not the end of it.

The platform combines frontier models from Anthropic, OpenAI, and Google with Rocket’s own deep learning systems, trained on a proprietary 10M Figma-to-code corpus inherited from DhiWise. A multi-agent orchestration layer breaks high-level prompts into structured sub-prompts, generates across Next.js, React, and Flutter, and merges outputs into a single coherent application. A continuous RLHF pipeline improves output quality over time.

Why is it unique?

Vishal’s framing is simple enough to repeat in one line: day one is easy; day two is where the real product begins.

Take a landing page, the canonical AI-generated artefact. A founder can spin one up in minutes today, and it will look perfectly serviceable. Then the actual work begins. Are leads coming through. Is the page ranking. Should the copy be personalised for visitors arriving from India, the Gulf, or the US. Should the team run an A/B test. App generation is the output. The actual customer wants the outcome. Almost everyone in the category is pricing the first thing. Rocket is built for the second.

This thesis was not invented post-ChatGPT. Rocket’s parent company DhiWise launched a code-generation platform in 2021, before the LLM wave, and the hard learning from that era is what shapes the current product. Code generation alone, Vishal says, does not create wide adoption. Users could ship in a day, but unless the application was reliable, maintainable, and capable of growing with them, they would forget about it within a month.

DhiWise also gave the team an engineering base most vibe-coding entrants do not possess - components built across the full software lifecycle (requirement analysis, design-to-code, end-to-end testing, code generation). That structural depth is what keeps pure AI generation from collapsing into architectural mess at scale. The first 90% of the work now happens in 10% of the effort. The remaining 10% - taste, polish, edge cases, last-mile quality - is where products are still differentiated, and it is what Vishal is building Rocket around.

The proprietary data advantage is the second half of the moat. Rocket is trained on 10 million Figma-to-code pairs accumulated through DhiWise. That dataset cannot be quickly replicated by a new entrant, and it produces UI code with higher fidelity and better architectural soundness than tools trained on generic public data.

The Surat point is also worth holding. Vishal does not frame Surat as a constraint - he frames it as focus. Strong engineering talent locally, no relocation just because the market expected serious startups to be in Bangalore, and an anti-hype philosophy that runs against the grain of an extremely loud category.

The big opportunity

The vibe-coding market is in explosive top-of-funnel growth with unclear unit economics. Most platforms are racing to ship the first generation faster; few are building the iteration, deployment, and outcome layer that determines whether users stay past the first month. Rocket’s traction signal is strong because it is built around that second layer: 30-40% MoM growth, double-digit million ARR, ~$3K annual ARPU, 50-55% gross margins, 400,000+ users across 180 countries, 85% of recent cohort apps being serious projects.

The bigger opportunity sits beyond app generation. Vishal’s stated ambition is for Rocket to become a comprehensive agentic system that builds the app, but also runs competitive research, conversion analysis, and product iteration. That is a much larger company than a code-generation tool. It is closer to a software agency rebuilt as a self-serve product surface - what Salesforce Ventures, in announcing the round, called a “full-stack AI solutioning engine.”

Geography is instructive. The US is already the largest revenue contributor at 26%, with Europe at 15-20% and India at 10%. The Palo Alto office and Salesforce Ventures relationship suggest enterprise distribution in the US is the next vector to unlock - while the engineering and product centre of gravity stays in Surat.

The risks

The category is the biggest risk. Vibe coding has become the most crowded segment of consumer AI, with billions of dollars flowing into Lovable, Bolt, Replit, Cursor, and adjacent players. Differentiation has to be defended every quarter, and the moat needs to compound faster than the next model release commoditises individual features.

The 25-minute generation time is a deliberate trade-off, not a bug. But it does create real friction in a category where competitors generate output in three minutes. Rocket has to keep proving that the additional time is repaid in quality and in downstream usability. Vishal’s bet is that day-two outcomes are what users will eventually be paying for; the question is whether they realise it before they bounce on day one.

Last-mile production is a structural challenge. Production-ready software requires hosting, databases, auth, payments, deployment, monitoring, and a long tail of edge cases. Each new layer is a potential drop-off point, and the more Rocket owns the iteration loop, the more operational surface area it inherits.

Our final take

The day-one-vs-day-two diagnosis is the part of Vishal's thinking that most clearly distinguishes Rocket from the rest of the vibe-coding category. Most platforms in this space are competing on speed of first generation, which is a feature race that the underlying models will commoditise within a year or two; Rocket is competing on completeness and second-month outcomes, which is a much harder thing to copy and a much more defensible thing to own. The 10M Figma-to-code corpus inherited from DhiWise is the kind of asset that gets harder to replicate the longer Rocket operates, and the engineering discipline visible in the multi-agent orchestration layer suggests the team can actually execute on the architectural ambition. Whether all of this converts into a category-defining position depends on whether Rocket can extend beyond app generation into the broader agentic product loop fast enough to stay ahead of category compression. The strategy is sound. And it all comes down to execution.

Pocket FM: A Closed-Loop Storytelling System

What does Pocket FM do?

Pocket FM is the global category-leader in long-form serialised audio entertainment, with operations across 20+ countries. The product is structured around episodic audio dramas - fiction, romance, thrillers, horror, self-help - distributed through a freemium model where listeners pay micro-payments to unlock new episodes.

What makes the company genuinely interesting in this edition, and what most coverage misses, is the architecture underneath. In the company’s own framing, Pocket has shifted from a content marketplace to an AI-native storytelling system. A proprietary AI Co-pilot - purpose-built for long-form storytelling and trained on billions of minutes of listener engagement and retention data - powers an end-to-end content supply chain across three coordinated stages: creation, blockbuster identification, and scale & expansion.

Crucially, even leading general-purpose models like OpenAI or Claude are not designed to do what this system does. The Co-pilot can produce hundreds of hours of coherent narrative while maintaining character consistency, emotional arcs, and core storytelling principles - the long-form coherence problem that horizontal models do not solve. It is not a generic LLM applied to fiction. It is a specialised system optimised against a different objective function entirely.

Why is it unique?

Pocket FM treats engagement as a closed-loop reward signal. Completion rate, drop-off curves, binge depth, re-listen frequency, session continuation - each is treated as an implicit reward function in reinforcement learning loops. Instead of optimising for next-token likelihood, the model is optimised toward sequences that maximise sustained engagement, tighter pacing, and stronger narrative hooks. Layered on top, the corpus of high-performing stories paired with fine-grained engagement annotations serves as a supervised fine-tuning signal that teaches the model not just how to write coherently, but how to write well in a specific genre, language, or audience context.

At sufficient volume, the system also discovers structural patterns - tension buildup, cliffhanger placement, character introduction timing, emotional cadence - that consistently correlate with retention. These patterns are not visible from smaller datasets. They emerge clearly only when you observe millions of listener journeys, and once extracted, they become priors guiding new stories. The system is not just generating coherent prose. It is generating prose inside structural rules empirically validated against audience behaviour at scale.

The localisation engine is the second piece worth pausing on. Pocket treats localisation as a system-design problem, not a translation task. Translation gets the words right; Pocket’s pipeline preserves the experience. Before any rewriting begins, the system extracts a complete map of the source story - world, characters, relationships, cultural elements - and builds a cultural transposition strategy. Adaptation then proceeds in three phases: characters first, then dependent entities like places and institutions (with graph-based grouping ensuring a renamed family cascades correctly across estates, titles, and symbols everywhere they appear), and only then a chapter rewrite that preserves tone, rhythm, and emotional equivalence. Surface details may change - food, setting, customs - but the underlying feeling stays intact.

The position on what stays human is unambiguous. Creativity itself remains entirely with the writer. The ideas, imagination, cultural instincts, emotional truth - all of it is human. The process is writer-led, with creators retaining full control and agency over every aspect of the story. The Co-pilot acts as a creative partner, supporting structure, continuity, and optimisation, while remaining fully directed by the writer. AI does not replace creativity. It compounds it.

The big opportunity

Pocket FM is operating in the largest revenue pool of the three companies and arguably one of the largest categories in consumer AI globally - entertainment. The numbers tell the story: UGC ecosystem at $50M ARR within 12 months. AI-enabled creator-led shows ~25% of total platform playtime. 80+ blockbuster IPs each generating $1M+ in revenue, with 4+ new blockbusters added every month. ~$50M ARR in Europe within 12 months. LATAM accelerating.

The Co-pilot does not just produce content. It changes the cost structure of producing hits. The company reports 4x more hits at 50x lower cost. Traditional content businesses underwrite a small number of expensive bets and absorb a high failure rate. Pocket is underwriting a much larger number of cheaper bets and using engagement signals to identify the winners early.

The terminal opportunity is to become the global default infrastructure for long-form serialised content - first audio, then potentially adjacent formats. The Co-pilot, the engagement-annotated corpus, the structural pattern discovery, and the localisation engine are all assets that travel across formats, languages, and geographies.

The risks

Three matter, in declining order of severity.

The first is regulatory. Pocket has faced lawsuits in the US over the classification of independent contractors, with creators alleging circumvention of US wage and worker-classification laws. The outcome of these cases, and the precedent they set, materially affects the cost structure of the creator economy underneath the platform. This is the single most important diligence item for any serious investor or partner.

The second is content homogenisation. AI-assisted serialised content carries a real risk that scaled-up production starts to feel formulaic - and a 25% playtime contribution from AI-enabled shows is large enough that any quality drift would be visible in retention metrics. The structural pattern discovery that powers the Co-pilot also creates the risk of producing structurally similar stories. The company’s edge will depend on whether the workflow continues to surface genuinely differentiated stories or merely scaled-up variants.

The third is competitive. Kuku FM, Audible’s expansion into long-form fiction, ByteDance’s audio plays, and Spotify’s serialised content investments are all operating in adjacent territory. Pocket has a meaningful technical and data lead today, but the moat is engagement compounding - and that compounds slowly.

Our final take

The architectural choice that distinguishes Pocket FM from the rest of the AI-native content category is the treatment of engagement as a reward signal rather than as a vanity metric, and watching how that closed-loop system performs over the next eighteen months will tell us a great deal about whether the underlying pattern generalises beyond audio entertainment to other long-form content categories. The engagement-annotated corpus, the structural pattern discovery layer, and the localisation engine are each genuinely difficult to replicate, and the early traction across India, Europe, and LATAM suggests the system is doing real work rather than benefiting from temporary tailwinds. The regulatory overhang in the US, around independent contractor classification and creator economy compliance, is the single live question that could materially change the cost structure of the platform underneath, and any honest assessment of Pocket's trajectory has to weight that question heavily. The architecture is genuinely promising.

India AI News Roundup

The most impactful AI developments & announcements shaping India in recent weeks.

Nandan Nilekani: India doesn’t need to lead the world in building the most advanced AI models

Pixxel, Sarvam join hands to build orbital data centre satellite

Yotta, Gorilla Technology expand India AI infra pact with $2.8 billion project

Krutrim announces strategic pivot to AI cloud services, reports 3x revenue growth and first net profit in FY26

AI startup Nava raises $22 million in round led by Greenoaks Capital

Startup Signals

Spotlighting brand new emerging AI startups from India every month, early and undiscovered.

OpenRound: AI-native engineering hiring assessments

Every method engineering teams use to evaluate candidates - LeetCode puzzles, multi-day take-homes, week-long work trials - was designed before AI changed how engineers ship, and most of them are now one-shotted by the same tools the candidates would use on the job. OpenRound is rebuilding the assessment workflow against that new reality. Candidates work on real codebases with full access to AI coding agents through their own CLI, and are evaluated across six dimensions including analysis, planning, judgment, execution, and AI collaboration itself. A single ninety-minute session replaces a process that typically runs three weeks and twelve interviewer-hours. Built by Fabric and already deployed at companies including Aays Analytics, OpenRound has positioned itself early in a category that is being fundamentally rewritten rather than incrementally improved.

https://www.openround.ai/

Navana: Foundational voice AI infrastructure for Indian enterprise

India’s voice AI landscape is splitting into two architectural bets - orchestration over rented models, and proprietary foundation models for Indian linguistic complexity - and Navana is the cleanest expression of the second. Its proprietary speech engine, Bodhi, is trained on hundreds of thousands of real-world conversations across 12 Indian languages and 40+ dialects, built specifically to handle code-switching, overlapping speech, and ambient noise that global voice stacks treat as edge cases. The company has aligned early with the heaviest end of Indian enterprise procurement: ISO 27001:2022, SOC 2 Type II, RBI guidelines, on-premises deployment, and full data localisation. Hello Ujjivan, its deployment with Ujjivan Small Finance Bank, now serves 1.5M+ users in 12 Indian languages. Built for the 900M+ Indians, urban and rural, that the global voice stack was not designed around.

https://navana.ai/

Atlas: AI as a junior accountant for independent firms

The accounting profession has shed 300,000+ professionals in the US since 2019, and smaller independent firms - squeezed between rising labour costs and the inability to absorb the kind of automation that the largest firms already have - are the structural beneficiaries of a workflow rewrite. Atlas, founded in 2025 by Arpit Maheshwari and Jagmal Singh (ex-CarDekho, PolicyBazaar), is building software that embeds AI across both client-facing and back-office accounting workflows, with a deliberately human-in-the-loop architecture that positions AI as a junior accountant rather than a replacement. The company reports productivity gains of more than 5x on certain workflows in early deployments. $6M seed in April 2026 co-led by Accel and Stellaris Venture Partners, targeting a $150B global market with go-to-market focused on North America.

https://www.getatlas.ai/

This time we went upstream. We surveyed senior technology leaders at 77 of India’s most consequential tech companies built in the last two decades. Nearly half of respondents are CTOs (49%), alongside Heads of AI (21%) and VPs of Engineering (19%), representing companies such as Meesho, upGrad, ixigo, Pine Labs, CRED, Jar, Dream11, etc, with a median company size of 500-2,000 employees.

This is a benchmark of what India’s foremost technology companies have already deployed, what impact they are measuring, and how they are restructuring their organisations in response.

If anyone is still debating whether India's technology industry has "adopted AI," the answer from the people who actually sign off on production deployments is unambiguous: this is not experimentation, it is an operational reality.

The Model Wars Look Different From the CTO’s Chair

In our builder & developer survey, OpenAI led production usage at 48%, with Anthropic at 25% and Google at 15%. Clear hierarchy with a clean story. The CTO-level picture is fundamentally different.

Three things jump out.

1. First, the gap has compressed. OpenAI still leads, but by single digits over Gemini and Anthropic. When you are a CTO running multiple production workloads - customer support in one stack, coding tools in another, content generation in a third - you don’t pick one vendor. You pick three. The average CTO in this survey uses 3 model vendors simultaneously. 44% use all three of OpenAI, Anthropic, and Gemini in parallel. This is the multi-model future arriving in production, not as a theoretical best practice, but as an operational necessity.

2. Second, Google Gemini’s position is far stronger among enterprises than among startups. At 68% penetration, Gemini is effectively tied with OpenAI in the CTO stack. The likely driver is distribution: Google Cloud relationships, Android ecosystem integration, and enterprise procurement pathways that simply don’t exist for early-stage builders paying with credit cards.

3. Third, and the most interesting one, Chinese open-weight models have achieved 48% penetration among India’s top CTOs. In February, we found only 3% of builders using Chinese open weights. Among CTOs at companies with 2,000+ employees, the number shoots up to 66%! Larger companies naturally end up using a wider mix of models across teams and use cases, so this indicates more about the breadth of experimentation. DeepSeek, Qwen, and their cohort are now in production at the majority of India’s large technology companies.

While Everyone Else is Split, Deeptech Chooses Chinese Models

While most of the ecosystem remains split across OpenAI, Anthropic, and Google, deeptech startups are making a very different and far more opinionated choice. 71% of deeptech companies are using Chinese open-weight models in production, making them the single most adopted category in this segment by a wide margin. This divergence is not accidental since deeptech teams are fundamentally more performance-sensitive, control-oriented and infrastructure-heavy (often building at the model/system layer rather than just consuming APIs).

Chinese open-weight models optimize exactly for these constraints. In contrast, sectors like SaaS and fintech—where speed to market and reliability matter more—continue to favor closed models from OpenAI and Anthropic.

What this signals is important: the deeper you go into the AI stack, the more the center of gravity shifts from closed APIs to open-weight ecosystems. Deeptech isn’t following the market but rather leading a structural shift in how AI systems are actually built. And increasingly, that stack is not Western by default.

Engineering Productivity is Table Stakes. The New Frontier is Revenue.

94% of CTOs report measurable productivity lift in engineering and product functions. 88% have coding AI tools live in production. This is no longer a finding but a settled fact of the Indian technology industry. The more interesting question: where is AI creating value beyond engineering efficiency?

Marketing is emerging as the second major AI beachhead at 57%. Revenue functions such as sales development, outbound, and pipeline are at 42% and climbing. Finance and legal sit at 36%, constrained by accuracy requirements and regulatory stakes.

But the headline finding is on impact measurement. When asked about the primary impact of their most important AI use-case:

Nearly half of India’s top CTOs say AI is already driving measurable revenue lift. Not “we think it might.” Not “our projections suggest.” But rather measured, reported, and in production. Among companies where AI touches >50% of workflows, 69% report revenue lift. Among companies with 5–50% penetration, only 21% do. There appears to be a threshold beyond which AI stops being a cost optimisation play and becomes a revenue driver.

The Great Re-Org Is Underway—But It’s Not What You Think

70% of CTOs say AI is actively changing their organisation structure and hiring plans. But the nature of the change is counterintuitive.

The dominant pattern is not “AI replaces headcount.” It is “AI restructures what headcount does.” 56% are reorganising their teams while continuing to hire aggressively. This is the shape of an industry that has found AI-driven leverage but hasn’t found AI-driven substitution, at least not across the board.

Where substitution is happening, it is concentrated:

Support and operations is the kill zone. 45% of CTOs report headcount reductions there, and the number scales with company size where 59% of companies with 2,000+ employees are cutting support/ops roles. Customer support is live in production at 52% of companies, and AI-powered support at scale is demonstrably cheaper per resolution than human agents.

Engineering headcount is being trimmed at 23% of companies. The framing from CTOs in follow-up conversations is that they are not hiring fewer engineers but rather fewer engineers for the same output, and reallocating the surplus into new product surface area. GTM and sales remain largely untouched at 6%. The sales org is the last fortress. Whether that holds through 2027 depends on how fast agentic outbound matures.

Even in India, Accuracy Beats Cost as the #1 AI Blocker

In a market assumed to be highly price-sensitive, the biggest constraint to scaling AI isn’t cost—it’s getting the models to be accurate and reliable enough in production

This is the signature of a market that has moved past “can we afford AI?” to “can we trust AI in production?” When your customer support bot hallucinates a refund policy that doesn’t exist, the cost per API call is irrelevant. The blocker landscape shifts dramatically by company size:

Large companies have largely solved for organisational readiness (14%) and are now stuck on the harder problems: quality at scale, cost at volume, and integration with legacy systems. Small companies face an inverted profile where security and org readiness are their primary constraints.

The regulated industry gap: 80% of Fintech CTOs and 100% of Healthcare CTOs cite data security as a scaling blocker. Compare that to 7% in E-Commerce. This is not a sentiment but a structural gap that requires purpose-built infrastructure of data residency, audit trails, model governance, and deterministic fallback systems.

The Priority Stack: Customer-Facing AI is the Consensus Bet

Over half of India’s top CTOs say their number one AI priority is shipping more customer-facing AI features. Internal productivity where most of the current deployment is concentrated is already in the rearview mirror as a primary objective. This priority intensifies with scale. 76% of CTOs at companies with 2,000+ employees are focused on customer-facing AI, versus 29% of sub-500 companies. The larger the company, the more confident the bet on AI as a product feature rather than an internal efficiency tool.

This is the shift from AI as back-office automation to AI as product surface area. It is the most consequential strategic reorientation in the dataset, and has implications for every AI startup selling to Indian enterprises. The buyer’s question is no longer “how do I make my team more productive?” but rather “how do I make my product smarter?”

The Inference Layer: Hybrid Wins

The majority of CTOs (52%) run hybrid inference combining API calls, managed platforms, and self-hosted models depending on the workload. This is the mature pattern where you use APIs for low-volume tasks, hyperscaler-managed inference for compliance-sensitive production, and self-hosted open-weights for high-volume cost-sensitive workloads.

Among 2,000+ employee companies, hybrid adoption hits 69%. The pure API-first pattern that dominated our early-stage builder survey is a stage of development, not a steady state. As companies scale, they graduate to a portfolio approach.

What All of This Means

1. The multi-model enterprise is here. 44% of CTOs run all three major vendors in parallel. Startups building model-agnostic infrastructure, such as routing layers, eval frameworks, and prompt management across providers, are building for the actual architecture of enterprise AI, not the hypothetical single-vendor world.

2. Chinese open-weights are the stealth story of Indian enterprise AI. 48% penetration overall, 66% among large companies, by far the preferred choice amongst Deeptech. This is not going to show up in API revenue figures from US model providers. But it’s reshaping the inference cost curve and the competitive dynamics of the Indian AI stack from below. The companies that help enterprises manage, fine-tune, and deploy open-weight models at scale have a large and rapidly growing addressable market.

3. Quality, not cost, is now the primary constraint. The market has flipped from “can we afford it?” to “can we trust it?” This is an opportunity for companies to build evaluation, monitoring, and quality assurance infrastructure. The companies that can guarantee output quality at scale will command pricing power that pure inference providers cannot.

4. Support/Ops is being restructured. Sales is not. Engineering is in between. The workforce impact of AI is concentrated and specific. If you are building AI tooling, the highest willingness-to-pay is in customer support automation (where ROI is already proven) and engineering productivity (where adoption is universal). Sales AI is still a future bet.

5. The shift from internal AI to product AI is the defining strategic move of 2026. 52% of all CTOs and 76% of large-company CTOs say shipping customer-facing AI is their top priority. This is the transition from AI as a cost centre to AI as a product differentiator. Startups that help enterprises embed AI into their products, not just their operations, are building for the next phase of demand.

6. Regulated industries need dedicated infrastructure. 80% of fintech and 100% of healthcare CTOs cite security as a blocker. This is not a sentiment. It is a structural gap. The companies that build data residency, model governance, and audit-trail infrastructure specifically for Indian regulated industries will find a market with high barriers to entry and sticky customers.

India AI News Roundup

The most impactful AI developments & announcements shaping India in recent weeks.

Emergent’s $100 million ARR puzzle: What counts as revenue in the AI era 

CBSE Launches AI & Computational Thinking Curriculum for Classes 3–8

Oracle layoffs: AI pivot triggers 10,000 job cuts in India

India AI Impact Summit Generated ₹67 Crore in Expo Revenue, $250B+ in Pledges

Gnani AI Raises $10M Series B for Conversational AI Platform

MeitY Selects 10 AI Startups for International Acceleration via Station F/HEC Paris

Startup Signals

Spotlighting brand new emerging AI startups from India every month, early and undiscovered.

ProactAI - Vision AI for India’s 234M CCTV Cameras

India has 234M installed CCTV cameras generating 100+ petabytes of video daily, and less than 0.1% gets analyzed. ProactAI converts this passive infrastructure into real-time intelligence using a ViT-based engine they call Bhaskara - targeting retail shrinkage, manufacturing safety, and defence surveillance. Founded by three IIT alumni (IIT BHU, IIT Bombay, NSUT) with combined 50+ years across HDFC ERGO, Octro, HT Media, and Razorpay. Three paying retail clients (including Nike/RJCorp and Davaindia) within six months of founding. Claims 98% accuracy vs. 70% from legacy CNN-based competitors at 1/10th the cost.

https://proactai.ai/

PhotonSilica - Indian Fabless Chip Design for Defence & Space

India’s semiconductor ambitions need fabless chip designers building for indigenous defence and space applications. PhotonSilica is designing mixed-signal ICs and photonic processors targeting the 28nm–130nm range, which represents roughly 80% of defence and industrial semiconductor demand. Founder Tushar has secured TSMC MPW shuttle access through Taiwan’s semiconductor diplomacy channel and is pursuing iDEX and IN-SPACe contracts where government co-funding dramatically improves seed-stage capital efficiency.

https://photonsilica.in/

Zyra - India’s First AI Insurance Expert

India has 550M+ insured lives, but most policyholders don’t understand what they’re paying for - jargon-filled documents, fragmented management, and an agent model built on commissions rather than clarity. Zyra is an AI-native consumer app that reads your policy, scores your coverage (via a proprietary “Z-Score”), identifies gaps, and provides unbiased guidance - no spam, no agent calls. The product is IRDAI-regulated and ISO 27001 certified. 10,000+ policies analyzed, ₹5,000+ crore in coverage reviewed across 50+ cities and is live on both iOS and Android.

https://heyzyra.ai/

Survey Methodology

All responses are voluntary and self-reported. Results reflect the composition of respondents and are weighted toward technology companies with active AI deployment.

We were there. And beneath the announcements, three big truths emerged. About sovereign AI, Indian IT’s future, and who actually captures value from India’s AI demand.

1. Sovereign AI Is a National Project, Not Just a Technical One

The entire summit week revolved around Sarvam AI.

Their pavilion drew more crowds than Google’s or OpenAI’s. PM Modi wore their Kaze smart glasses on camera. They launched Sarvam-30B and Sarvam-105B, mixture-of-experts models trained from scratch on domestic compute supporting all 22 scheduled Indian languages. The 105B reportedly outperformed China’s DeepSeek R1 and GLM 4.6 Air on several benchmarks. They announced partnerships with Qualcomm, HMD, and Bosch to push models to edge devices. They launched a ChatGPT-style assistant called Indus. They even shipped hardware.

It was, without question, the most important technical announcement of the week. And what makes it even more impressive is the context. Sarvam built these models with access to roughly 4,000 GPUs under the IndiaAI Mission and a core engineering team of about 15 people. Frontier labs in the US and China typically train models on clusters 10–50× larger, with teams far bigger.

India’s government backed 12 labs under the IndiaAI Mission’s Foundation Models pillar, allocating roughly ₹2,000 crore in grants to accelerate domestic model development. While Sarvam was one of them_, a few others showed promising work as well. BharatGen launched Param2, a 17B parameter model. Gnani.ai demonstrated Vachana, a multilingual voice-cloning system supporting a dozen languages. The broader effort is still early, and many of these programs are multi-year bets. But the summit made one thing clear: building frontier AI capabilities is incredibly difficult.

The goal of sovereign AI isn’t necessarily to build the single best model in the world. It is strategic capability. Nations increasingly view frontier AI the same way they view space programs, nuclear programs, or semiconductor fabs. Not every experiment succeeds. Not every research lab produces breakthroughs. But the capability itself becomes a pillar of national security and technological independence.

From that perspective, even a single success like Sarvam matters enormously. It proves that India can train large models domestically, build language systems across its own linguistic landscape, and deploy AI infrastructure without depending entirely on foreign labs.

At the same time, sovereign models will likely coexist with the best systems developed globally. The most successful AI startups in India will probably combine both worlds:
using world-class foundation models while layering India-specific intelligence on top — language, workflows, regulation, distribution, and domain expertise.

2. From Labour Arbitrage to AI Leverage: The Next Chapter of Indian IT Services

Another major theme of the summit was the wave of partnerships between global AI labs and India’s IT services giants.

TCS announced a multi-dimensional strategic partnership with OpenAI. The agreement includes ChatGPT Enterprise for hundreds of thousands of employees, Codex for software engineering workflows, and joint go-to-market for agentic AI solutions. TCS also signalled plans for a large-scale data centre buildout through its HyperVault platform. Just a day earlier, Infosys announced a strategic collaboration with Anthropic, integrating Claude models into Infosys Topaz and launching an Anthropic Centre of Excellence focused on regulated industries like telecom, financial services, and manufacturing.

On the surface, these announcements looked like traditional partnership deals. But they reflect something deeper: a structural shift in how enterprise technology services will be delivered in the AI era. For decades, the Indian IT services model was built on a simple equation. Complex enterprise work from global companies was decomposed into thousands of smaller tasks and executed by large engineering teams in India.

AI changes that equation. When frontier models can modernise legacy systems, generate documentation, automate testing, and assist with large-scale migrations, the amount of human labour required to deliver the same outcomes declines significantly. That doesn’t eliminate the need for services — but it does change their nature. Instead of providing large pools of engineering labour, IT firms increasingly become AI implementation partners: helping enterprises integrate frontier models into existing systems, workflows, and regulatory environments.

But it would be naïve to pretend this transition won’t have real consequences. At the summit, Vinod Khosla put it bluntly on one of the panels: as AI systems become capable of performing large portions of engineering, documentation, support, and compliance work, many of the tasks currently handled by IT services and BPO teams could eventually disappear. India’s IT services sector generates roughly $350B in annual exports and employs millions of engineers & BPO workers. For a country where IT services has been one of the largest engines of white-collar job creation for three decades, that shift is not just a technology story. It is an economic and workforce transition that India will have to navigate carefully.

3. India’s First AI Unicorns Are Beginning to Emerge

For the past two years, one question has quietly hung over India’s AI ecosystem: When will the first wave of Indian AI unicorns actually appear?

At the summit, the early signals began to materialise. The most striking came from the infrastructure layer. Neysa announced a $1.2B financing package - $600M in equity led by Blackstone alongside $600M in planned debt, valuing the company at $1.4B conditional on milestones. It is the largest AI-focused funding round in Indian history. Blackstone, which has backed global data centre leaders such as QTS, AirTrunk, and CoreWeave, is now making a significant bet on India’s AI infrastructure. Neysa plans to deploy 20,000+ GPUs domestically for training and inference workloads. Today, India has fewer than 60,000 GPUs deployed nationwide. Blackstone expects that number to scale more than 30× to over two million in the coming years.

The second signal came from the application layer. Emergent crossed $100M ARR in just eight months, making it one of the fastest companies globally to reach that milestone. For comparison, Slack took roughly two years, and Zoom took three. More than 6M builders across 190+ countries have already created over 7M applications on the platform. Roughly 70% of revenue comes from the US and Europe, even as the company’s engineering team remains overwhelmingly based in Bengaluru. Emergent did not train a foundation model. Instead, it rode the global vibe-coding wave, building an application-layer product that sits on top of existing models and serves developers worldwide. In doing so, it has become one of the strongest early proof points that globally relevant AI products can be built from India.

And then there is Sarvam. Following the launch of its frontier models at the summit, investor interest in the company has intensified dramatically. While Sarvam has not yet crossed unicorn status, the trajectory is clear, and it represents one of the most credible attempts to build frontier AI capability from India.

Taken together, these signals point to something important. After years of discussion about India’s potential in AI, the first wave of AI-native unicorns is now beginning to form across infrastructure, applications, and foundation models. Some are being built to power India’s own AI ecosystem, while others are building global products from Indian talent. For the first time, both of those stories are unfolding at the same moment.

India AI News Roundup

The most impactful AI developments & announcements shaping India in recent weeks.

Anthropic Valuation Eclipses All Indian IT Combined

NVIDIA deepens early-stage push into India’s AI startup ecosystem

India Earmarks $1.1 Billion for AI & Deep Tech Fund of Funds

UAE’s G42 teams up with Cerebras to deploy 8 exaflops of compute in India

Anthropic opens Bengaluru office and announces new partnerships across India

AI Impact Summit: US, China among 88 signatories to New Delhi Declaration on AI

Startup Signals

Spotlighting brand new emerging AI startups from India every month, early and undiscovered.

Von Neumann Computing - The Personal AI Computer

The thesis is simple and contrarian: cloud is wildly overpriced, and the “cloud tax” is 10x worse than the “CUDA tax.” Von Neumann Computing’s answer is JOHNAIC - a plug-and-play personal AI server that runs AI and SaaS workloads from your office or home, on a normal power socket, with no subscription. The hardware ships with an Nvidia 5070 Ti 16GB GPU, 64GB RAM, 1TB SSD, and an optional UPS, all pre-loaded with a fully open-source software stack. Customers at People+AI (Nandan Nilekani’s EkStep initiative) validated the model — they deployed production workloads on JOHNAIC boxes hosted in their own offices. Still early, still scrappy, but the edge compute + data sovereignty angle is structurally interesting as India’s AI workloads scale beyond what cloud economics can serve.

https://johnaic.com/

Eyecandy Robotics - Physical AI Characters

Three founders, all under 25. One product: Tensaur, a physical AI character designed for entertainment and companionship - essentially a robot brought to life through AI. The company was incorporated in Bangalore just weeks ago (January 2026) and immediately selected for Lightspeed’s India Ascends cohort, alongside three other deep tech teams. Founders Alqama Shaikh, Raghuvamsi Velagala, and Mankaran Singh are building at the intersection of physical robotics and foundation models - the same design space as Figure, 1X, and Unitree globally, but aimed at consumer entertainment rather than industrial use. Pre-revenue, pre-product.

https://www.eyecandyrobotics.com/

Autoloops - AI Agents for Meta Ads

Quietly building AI agents that autonomously manage and scale Meta ad campaigns for D2C and e-commerce brands. The headline claim: 47% average ROAS improvement. Autoloops sits in a fast-heating category - agentic ad ops - where AdsGency just raised $12M, Madgicx is scaling aggressively, and Meta itself is embedding its Manus AI acquisition into Ads Manager. The Indian angle matters: India has the highest density of Shopify-powered D2C brands outside the US, and performance marketing agencies here run on razor-thin margins that AI agents can structurally compress. Autoloops is pre-announcement and nearly invisible publicly - exactly the profile of a company worth tracking before the round hits the press.

https://autoloops.ai/

Respondents came from 12 industries. 53% identified as developers or engineers, 37% as founders, with the remainder split between product managers and generalists. Over half of the respondents are building at the pre-seed or stealth stage. 35% reported shipping customer-facing AI features used by more than 1,000 users.

The resulting dataset is a reasonable cross-section of India’s AI builder ecosystem, with an expected skew toward early-stage teams. While some findings align with prevailing assumptions, others meaningfully diverge. A subset of the results has direct implications for anyone building or investing in AI infrastructure for India.

OpenAI dominates the market today

When we asked which text generation APIs people actually use in production today, OpenAI still leads at 48%. Anthropic sits at 25%. Google at 15%. That's the installed base talking. OpenAI got there first, built the integrations, became the default.

But preferences are shifting to Anthropic

When we asked a different question: if you could use only one LLM provider for the next 12 months, who would you pick?

42% said Anthropic. 32% said OpenAI. 19% said Google.

This isn’t a niche preference among one builder segment. Among teams with live products serving 1,000+ users, it’s 40% Anthropic vs. 31% OpenAI. Among teams stuck in PoC purgatory, Anthropic’s share hits 53%. The forced-choice question captures where preference is moving, and preference is a leading indicator of where spend goes next.

Two things seem to drive this, which we validated from direct discussions with some of the survey respondents:

1. Claude’s perceived edge in code generation and long-context reasoning matters disproportionately when 49% of Indian builders use coding APIs as a primary workload.

2. And pricing: Claude 4.5 Sonnet sits at a price-performance point that registers in a market where 41% of builders spend under $600/month on AI infrastructure.

Anthropic seems keen on consolidating a market where they already lead in developer intent by opening up a Bengaluru office and hiring their India head.

The Great Coding Migration

The most visible shift in India’s AI stack over the past six months is in coding tools.

24% of all respondents dropped GitHub Copilot. That was the single most common tool abandoned after “our stack is stable.” When we look at what people use now, Claude Code holds 42% of the coding assistant market among Indian builders. More than Cursor and GitHub Copilot combined.

Copilot was bundled, familiar, and enterprise-approved. It still lost. Developers cite Claude Code’s multi-step execution, codebase-wide navigation, and sustained context amongst the reason for preference. Once the gap became clear, switching costs proved lower than expected.

Three adjacent migrations reinforce this trend.

• 14% deprecated LangChain in favour of native code, custom frameworks now lead orchestration (40% vs. LangChain’s 34%).

• 15% moved off GPT-5.2, largely toward Claude 4.5 and open-weight models, driven by price-performance.

• And 5% replaced Pinecone or Weaviate with PostgreSQL + pgvector or long-context approaches - small in share, but directionally meaningful.

It’s ElevenLabs vs OpenAI in Voice

Voice in India is primary interface for many interactions, both consumer and enterprise-facing. 37% of builders use voice APIs in production. The market is a tight two-way race.

Voice is where the India-specific opportunity concentrates. 22 official languages, voice-first behavior, code-mixing across languages, and low literacy in many segments. No provider has locked this down. The next breakout consumer AI product in India will probably be voice-native.

While Google Owns Image & Video Generation

It seems like the image and video generation, the race is already decided. Google holds 49% - nearly as much as the rest of the market combined.

Google's 49% share is driven by Imagen 4, Veo 3, and its distribution footprint across Android and Chrome. This is a distribution-led moat, not a model-quality moat — and it's the hardest kind to dislodge.

India is API-first, not GPU-first

74% of builders use closed, proprietary models as their dominant architecture. Only 9% run Western open-weights. 3% use Chinese open-weights like DeepSeek or Qwen.

The “open-source India” narrative doesn’t match what’s actually running in production.

83% of builders include model provider APIs as a primary stack component. 30% self-host inference. 65% of GPU cycles go to inference, not training or fine-tuning. The majority of this ecosystem is calling APIs, not training models.

60% of builders spend less than $2,400 a month. Investing in training infrastructure makes less sense every month when you can ride the capability curve by calling an API. The value creation opportunity is at the application and services layer, not the model layer.

For investors, the companies worth backing in India are not the ones with the most GPUs. They’re the ones with the deepest understanding of specific workflows, the most useful proprietary data loops, and the fastest product iteration cycles.

Fine-Tuning Is Real, But Selective

43% of respondents fine-tune their models. Higher than many would expect, but not the majority.

The rate varies by stage in an interesting way.

The growth-stage peak makes intuitive sense. These companies have found product-market fit. They have enough production data to make fine-tuning worthwhile, and enough competitive pressure to justify the investment. Pre-seed teams fine-tune at a surprisingly high rate, probably because many are building AI-native products where model customization is the product itself. Late-stage enterprises fine-tune less, likely relying on prompting, RAG, and off-the-shelf models instead.

Among fine-tuners, the techniques are practical. SFT is the most common full fine-tuning method (87 respondents). LoRA dominates parameter-efficient approaches (79). RLHF is used by 73 respondents, which is worth pausing on. That’s not just casual LoRA experiments. A meaningful subset of Indian builders are implementing reward-model pipelines.

Agentic Workflows: The Unanimous Next Bet

No result in the survey was this skewed. 62% of builders plan to add agentic workflows in the next six months. Every other capability trails far behind.

The direction is logical. Models now support production-grade agents, the use case fits India’s strength in process automation, and agents can command higher pricing by delivering outcomes rather than suggestions.

The tension is cost. Agentic workflows are also the most expensive and failure-prone pattern due to multiplying API calls, latency, and error rates. And cost is already the top blocker to adoption (43%).

The ecosystem is racing toward the costliest AI pattern at the moment, cost matters most. Whoever solves for cost per successful task (not cost per API call), wins the middleware layer.

The Non-Technical Blockers

55% said they face no significant non-technical blockers. For the other 45%, the answers matter.

Data sovereignty affects 16% of builders. That’s a structural demand signal for Indian sovereign cloud providers and for model providers willing to offer India-hosted inference.

The talent gap is specific. 13% say they have a budget but can’t find engineers who know how to evaluate and optimize AI systems. India produces plenty of ML engineers. It doesn’t produce enough who understand production AI tradeoffs: cost-quality curves, eval frameworks, latency profiling. This is a different skill set than model building, and training programs haven’t caught up.

The procurement friction at 7% is almost funny in its simplicity. Many US-based AI companies don’t offer rupee billing or GST-compliant invoicing. Indian enterprises with finance teams need clean audit trails. Trivially fixable yet remains unfixed.

What All Of This Means

1. The provider war is Anthropic’s to lose. They lead in intent, their coding tools dominate, and they’re opening locally. But intent isn’t adoption yet — OpenAI still leads in production API calls. The conversion depends on Anthropic’s enterprise sales motion, pricing, and whether they ship Indic language support and data residency fast enough.

2. Voice is underpriced. 37% of builders already use voice APIs. The market is fragmented, the demand is structural, and no provider owns it. In a 22-language, voice-first country, whoever cracks multilingual voice AI at scale will build something durable.

3. Agentic workflows will drive the next wave of enterprise value and the next wave of infrastructure pain. 62% of builders are heading there. Most aren’t ready for what it costs. The companies that solve agent economics—cost per completed task, not cost per API call—will capture the middleware layer.

4. “Boring stack” companies will outperform. pgvector over Pinecone. Custom code over LangChain. Major cloud over GPU specialists. The companies building on this stack are optimizing for what matters at scale: reliability, cost, and engineering velocity. Investors should be skeptical of tools that add abstraction without proportional value.

5. The talent bottleneck isn’t in ML engineering. It’s in AI ops. India produces people who can build models. It doesn’t produce enough who can evaluate, optimize, deploy, and monitor AI systems in production. Companies and training programs that address this specific gap have a structural opportunity.

India AI News Roundup

The most impactful AI developments & announcements shaping India in recent weeks.

‘Biggest AI Summit’: Ashwini Vaishnaw Says India AI Impact Summit Getting ‘Phenomenal Response’

Budget 2026: Tax Holiday Until 2047 for Foreign Cloud Companies Using Indian Data Centers

Anthropic Appoints Ex-Microsoft India MD to Lead Bengaluru Office

OpenAI Sam Altman plans India visit, to join Sundar Pichai, Alexandr Wang, Jensen Huang, Demis Hassabis for India AI Summit

Up to $100 billion investments may be announced at India AI Impact Summit

Startup Signals

Spotlighting brand new emerging AI startups from India every month, early and undiscovered.

Arrowhead - Voice AI Agents for BFSI Sales

Arrowhead builds human-sounding voice AI agents that handle end-to-end sales conversations for banks, NBFCs, and insurers — calls lasting up to 20 minutes without customers detecting AI. The platform supports 7 Indian languages with mid-call code-switching and claims 45% higher conversion rates than human agents. 50+ clients including Bank of Baroda, Aditya Birla Capital, Paytm. 5x ARR growth Aug–Oct 2025; every PoC has converted to live deployment. $3M seed led by Stellaris; angels include Kunal Shah (CRED) and Madhusudanan R (M2P). Stellaris estimates India’s BFSI voice AI market at $3B with <$50M penetrated.

https://arrowhead.ai/

FPV Labs - Spatial Intelligence Infrastructure for Robotics

FPV Labs is building the data infrastructure for physical AI - capturing human dexterity at scale to train general-purpose robots. The core thesis: language models scaled because the internet had trillions of tokens; robotics has no equivalent data flywheel. Simulation and teleoperation don’t produce the scale, diversity, or edge cases of real-world interaction. FPV is creating that missing layer - capturing the long tail of human physical manipulation as training data for the next generation of humanoids and spatial computing. Incubated from Lossfunk.

https://www.fpvlabs.ai/

Novyte Materials - AI-Powered Materials Discovery

Novyte uses generative AI and physics-based simulation to compress materials discovery from years to weeks. The platform targets specialty chemicals, advanced polymers, paints, adhesives, and coatings — sectors actively seeking alternative materials for performance and sustainability. Closed-loop automation links computational modelling with robotic synthesis. Incubated at ICT-NICE (Institute of Chemical Technology, Mumbai). Raised ₹4.15 Cr pre-seed led by Theia Ventures (backed by British International Investment), with angels from Venwiz and Chemvera. Currently running early pilot projects with manufacturing and aerospace partners.

https://novyte.in/

Survey Methodology

Survey conducted January 20 to February 1, 2026 through Activate’s builder community and broader network. 244 responses across founders, engineers, PMs, and operators building with AI in India. 12 industries, pre-seed through late-stage enterprise. All responses are voluntary and self-reported. Results reflect the composition of respondents and likely overindex on early-stage, technical builders within Activate’s community.

But here’s the uncomfortable truth: despite all this momentum, India still lacks a single homegrown hyper growth native-AI company clocking $100M+ ARR. And this is when the country is now ChatGPT’s second-largest market, Anthropic’s second-largest by usage, ElevenLabs’ second-largest by enterprise revenue and Perplexity’s largest by user base.

The question for 2026 isn’t whether India matters to global AI. The question is whether Indian companies can capture the value being created, or whether it concedes a lot of that to US-based frontier labs and big-tech.

Here’s a roundup of the top 10 things that mattered in the India AI landscape in 2025.

1. The IndiaAI Mission makes compute available at scale

The ₹10,371 crore IndiaAI Mission announced in 2024 moved from paper to production.

By mid-2025, the government had empanelled 34,333 GPUs across sovereign cloud providers like Yotta, E2E Networks, Neysa, Jio Platforms, and CtrlS. Access through the IndiaAI Compute Portal costs roughly ₹87.59/hour, about one-third that of global rates, with 40% subsidy by the government.

What’s actually running on it? Sarvam AI’s sovereign LLM development, BharatGen’s multilingual foundation models, and over 30 AI application projects across healthcare, agriculture, and governance.

The friction is real though. Only ~17,000 GPUs are fully operational. Queue times persist. The companies that will win aren’t waiting for government allocation—they’re raising capital and buying their own GPUs.

But the symbolic value matters. Public compute signals state commitment, which de-risks private investment. That might be its most important function.

2. India will have it’s own LLM (soon)

Eight companies are now building indigenous Large Language Models (LLMs) under the IndiaAI Mission’s Foundation Models pillar, with approximately ₹2,000 crore in grants.

The lineup includes:

1. Sarvam AI (sovereign LLM for governance, Hindi-first, 120B parameter model),

2. Soket AI Labs (open-source multilingual for defense/healthcare, 120B parameter model),

3. Gnani AI (voice AI, real-time multilingual speech, 14B parameter model),

4. Gan AI (text-to-speech, 70B parameter model),

5. BharatGen from IIT Bombay (bilingual multimodal, 2B–1T parameter range model),

6. Fractal Analytics (large reasoning model for STEM/medical, 70B parameter model),

7. Tech Mahindra (enterprise-focused, TBD on the size), and

8. Avataar AI (domain-specific industrial models, TBD on the size).

Sarvam AI, backed by Lightspeed and Khosla Ventures, expects to release India’s first sovereign LLM by February 2026, timed for the AI Impact Summit. The model will power use cases like “2047: Citizen Connect” and “AI4Pragati.”

The debate nobody seems to be resolving: should publicly-funded models be open-source? Sarvam’s flagship is closed-source today. DeepSeek-R1, by contrast, was released with open weights.

“Sovereignty” is a politically useful framing, but competitiveness is what actually matters for startups and enterprises. If Sarvam’s model can’t match GPT-5 or Claude 4 on English tasks while excelling on Indic languages, adoption will lag regardless of who owns the weights. The real test: will Indian enterprises actually deploy these models, or keep buying from OpenAI and Anthropic?

3. Big Tech commits over $80 billion to India

The hyper scaler land grab is on.

Microsoft committed $17.5B over 2026–2029—their largest investment in Asia. Satya Nadella met PM Modi in December and the package includes a new Hyderabad data center region (live mid-2026), sovereign cloud solutions, skilling for 20M people, AI integration into e-Shram & National Career Service platforms serving 310 million informal workers.

Google is putting $15B into building India’s first “gigawatt-scale” AI hub in Visakhapatnam - three data center campuses, a new international subsea gateway on India’s east coast, partnerships with AdaniConneX and Airtel.

Amazon committed $35B by 2030 for AWS expansion and AI for 15M small businesses.

OpenAI opened a Delhi office for enterprise sales. Anthropic is coming to Bengaluru in early 2026 with Indic language support.

But the most interesting play might be Perplexity’s exclusive deal with Airtel: 360 million customers get free 12-month access to Perplexity Pro (worth ₹17,000). It’s the first time a frontier AI company has bundled with a telecom in India — and potentially a template for distribution at population scale.

The infrastructure investments are table stakes. What’s genuinely novel is using telecom distribution rails to acquire users at near-zero CAC. We expect every frontier AI company to pursue similar partnerships. The winner in consumer AI in a price-sensitive, less AI penetrated market like India may not be whoever has the best model but rather whoever cracks distribution first.

4. India is the world’s 3rd largest AI consumer

The demand signal is no longer speculative.

OpenAI: India is ChatGPT’s second-largest market outside the US. Downloads hit 46.7M in Q2 2025 alone.

ElevenLabs: India is their largest market by signups, second-largest by enterprise revenue. Founder Mati Staniszewski put it well: “India is already voice-first—roughly one-third of internet users use voice search each month, the highest globally.”

Anthropic: India ranks second globally in Claude usage, behind only the US.

Perplexity: India became their largest market by monthly active users in Q2 2025—downloads up 600% YoY.

India isn’t just a cost arbitrage play for AI companies anymore. It’s a demand center. But demand without supply-side capture is just value extraction. Right now, Indian users are training American models. The consumer AI opportunity is enormous, but only if Indian companies can build products that win on experience, not price.

5. India may lead in Consumer AI innovation

2025 saw the emergence of India-first consumer AI categories that don’t have clean Western analogues. This is where things get interesting.

AI Astrology is having a moment. Apps like AstroTalk (5M+ downloads), Shani.ai, Melooha, and KundliGPT combine Vedic astrology with LLMs. Users get personalized Kundli readings, dasha predictions, and voice-based consultations. The astro-spiritual market in India is valued at $40+ billion - and AI is lowering the cost of access dramatically.

This isn’t “woo-woo.” It’s a massive market with clear willingness to pay, cultural acceptance of digital delivery, and a perfect fit for AI’s core capability: personalized advice at scale.

AI Companions are emerging, too. Rumik ($1.5M raised), Mello, and Kavana are building for India’s loneliness economy. The thesis: 700M internet users, vernacular-first, voice-first, mobile-first. The demographic tailwinds are real - urban loneliness, delayed marriage, nuclear families.

AI Learning products are proving out faster than expected. Arivihan ($4.17M from Prosus and Accel) and Supernova (backed by Kae Capital, Lumikai, AdvantEdge) are building AI tutors that work in vernacular, price for Bharat, and solve for specific skill gaps—board exam prep and spoken English respectively. The thesis: 250 million Indian kids can't afford quality tutoring, but they have smartphones. AI tutors can deliver 1-on-1 personalization at ₹300/month instead of ₹5,000+. 80% of Arivihan's users come from Tier 3 cities and rural areas.

India is uniquely positioned to lead in Consumer AI: a billion-plus users, mobile-first behavior, and daily interaction with imperfect systems that demand AI-native solutions. The next wave of global consumer AI will be trained in the chaos, scale, and diversity of India.

6. Voice AI set to transform the Enterprise

Voice AI became the default interface for enterprise India in 2025. This is no longer a nice-to-have - it’s table stakes for reaching the mass market.

India has 22 official languages and hundreds of dialects. Text-based AI struggles with code-mixing (Hindi-English), regional accents, and low-literacy users. Voice solves this.

Gnani AI is building a 14B parameter voice model under the IndiaAI Mission, already deployed for multilingual customer service at banks and telcos. Sarvam AI is powering Hindi in Meta’s Ray-Ban glasses - making it the first Indian AI company to power a global consumer product. ElevenLabs launched their Voice Actor Marketplace in India, projecting $1M+ payouts to Indian voice talent by the end of 2025, with Hindi and Tamil support on v3 models.

From contact centers to government services to consumer apps—if you want to reach the next 500 million users, you build voice-first. We’ve been saying this for two years: voice isn’t a feature, it’s a platform shift. The companies building voice-native from day one have a structural advantage over those bolting it on later. The question is whether Indian voice-AI companies can build defensible moats before frontier labs catch up on Indic languages.

7. The capital landscape is changing

The funding landscape for Indian AI moved beyond traditional VC in 2025. What’s emerging is a blended capital stack - private VCs, corporate CVCs, and government -that’s unique to India’s AI moment.

Government is showing up with real money. The headline is that PM Modi launched the ₹1 lakh crore ($12 billion) Research, Development and Innovation (RDI) Fund in November - the largest public commitment to private-sector R&D in India’s history. The six-year scheme offers low-interest (sub-3%) or even interest-free unsecured loans, equity, and patient capital to private companies and startups for high-risk, high-impact projects in sunrise sectors like AI, biotech, quantum, and semiconductors, aiming to bridge the gap between lab research and commercial success. The fund also includes a Deep-Tech Fund of Funds to back high-risk startup bets that traditional VCs won’t typically invest behind.

AI-native India funds are emerging with distinct theses. Activate, Together Fund, Boundless Ventures, Neon Fund all launched new funds this year as AI specialists, bringing domain depth, early technical conviction, and hands-on support to founders building from day zero.

Global capital is deploying earlier and smaller. The Accel + Google Ventures partnership - a $2 million co-investment initiative specifically for Indian AI startups - signals willingness to get in at the ground floor, not just Series B and beyond.

Meanwhile, Indian VCs are going West. Peak XV, Elevation, and Together Fund have all hired US-based partners and are ramping investments in American AI startups. Lightspeed India has been building in the US since 2021 and operate actively across the India–US corridor, tightly integrated with their global platform. Follow the returns, deploy into the frontier, and bring learnings back.

The question is whether this blended capital stack can sustain the patience required for deep tech bets. Government money is patient but bureaucratic. VC money is fast but impatient. The best outcomes will likely come from founders who can navigate both: taking government grants for R&D runway while raising private capital for go-to-market velocity.

The opportunity for AI-native capital in India has never been better. Most generalist VCs still don’t have the pattern recognition to evaluate pre-product AI companies. That gap, while being real, won’t last forever.

8. AI-Led Services: India’s quiet breakout category

In 2025, one of India’s most under-appreciated AI success stories was AI-led services. Instead of selling software alone, a new generation of companies combined AI systems with service delivery to solve messy, real-world problems—sales, customer support, finance ops, healthcare workflows, compliance, and more.

India’s advantage here is structural. Deep operational talent, cost-efficient execution, and proximity to global customers allow Indian teams to deploy AI directly into production environments. These companies use AI to replace labor, compress timelines, and guarantee outcomes, often charging on performance rather than seats.

What’s notable is that many of these businesses don’t look like traditional SaaS at birth. They start service-heavy, learn fast from real usage, and gradually automate themselves into high-margin, defensible platforms. Squadstack now runs one of the largest AI-enabled call centers in the country, Realfast is automating Salesforce deployment, and Bridgetown Research is combining AI with expert analysts to deliver faster investment research for global funds.

What unites these companies is not software elegance, but ownership of outcomes - revenue generated, systems deployed, or insight delivered.

9. The Layoff Reckoning: 50,000+ jobs at risk

AI didn’t just create jobs in 2025. It eliminated them.

TCS cut 12,000–20,000 employees, roughly 2% of their workforce. The CEO, K. Krithivasan, cited “skill mismatch” and “limited deployment opportunities.” They introduced a 35-day bench policy: employees not assigned to a project within 35 days face termination. “Silent layoffs” are becoming the norm: no public announcements, just discreet internal communications.

Industry-wide, analysts estimate 50,000+ IT jobs are at risk due to AI automation, primarily in backend operations and support functions.

Context matters here. India’s IT sector employs 5.67 million people and contributes 7.5% of GDP. TCS alone employs over 600,000. The sector’s traditional model - large headcounts, labor arbitrage, pyramid hiring - is under structural pressure from AI-driven productivity.

The counterpoint: Infosys hired 17,000 in Q1 FY26 and plans to recruit 20,000 graduates this year. HCLTech added 5,196 freshers in Q2. The story isn’t uniform.

But the direction is clear. The IT services model was built on labor arbitrage. AI challenges that model. The companies that survive will be the ones that transform into AI-services companies rather than the ones that try to protect headcount. This is the part of the AI story nobody in India wants to talk about publicly. The second-order effects on the middle class are significant and under-stated.

10. India hosts the Global AI Impact Summit 2026

For the first time, the Global AI Summit series comes to the Global South.

The India-AI Impact Summit will be held February 19–20th, 2026, at Bharat Mandapam in New Delhi, hosted by Honorable Prime Minister Narendra Modi under the IndiaAI Mission.

What’s expected: heads of state, global leaders, researchers, industry executives, the release of India’s sovereign LLM, launch of new foundational model initiatives, AI Governance Guidelines, 300+ exhibitors from 30+ countries.

The framing is distinctly Indian - three “Sutras” (People, Planet, Progress) and seven “Chakras” (Human Capital, Inclusion, Safe & Trusted AI, Resilience, Innovation, Democratizing AI Resources, Social Good).

The summit follows the UK AI Safety Summit (2023), AI Seoul Summit (2024), and Paris AI Action Summit (2025). The symbolism is deliberate: India wants a seat at the table where AI’s future is decided.

What makes this interesting isn’t the summit itself - it’s what it represents about India’s evolving position in AI’s global conversation. Whether this translates into influence over actual AI development trajectories or remains a parallel conversation is the thing to watch.

Questions to ask in 2026

1. Can India’s sovereign LLM compete with the frontier globally?

2. How soon can an Indian native-AI startup cross $100M ARR? How many learnings from their playbook could be applied to others?

3. Can India’s ChatGPT moment be a mass AI Tutor or AI Doctor for the next billion users?

4. Will the IndiaAI Mission’s compute actually reach startups at scale enabling them to do rapid innovation?

5. Can India retain the AI talent it produces or will the US continue to vacuum it up?

6. Will AI unlock a new wave of Indian services companies, or trigger widespread job losses?

7. Can vehicles like the RDI Fund bridge the gap between lab-stage research and venture-scale companies?

8. How strategically important will India become to global AI labs beyond usage, talent, and data?

9. Can Indian conglomerates like Reliance, Adani, and Tata become long-term sovereign AI builders?

10. Will the AI Impact Summit produce binding outcomes for startups and AI practitioners or just communiqués?

The infrastructure is being laid. The capital is flowing. The talent exists. The demand is real.

What’s missing is proof: companies at scale, AI research, and homegrown products that show India can lead in AI, not just participate.

That’s the test. The summit in February is the starting gun.

This is the final edition of Activate Signal for 2025. See you in the new year—and at the Summit.

We spend most of our time talking to builders across India’s AI ecosystem. Over the past few months, as we move towards the next phase of the AI S-Curve, the conversation is quickly moving towards value. The question is no longer “can AI do this?” but “who will build the AI-native version that’s 10x the status quo experience?”

This edition is a call to builders across 3 core themes where we see the highest potential for India to win. This is nowhere near a comprehensive list, but here are three vectors where we think the conditions are right to create category-defining companies from India in 2026: AI-led Services, Consumer AI, and Sovereign AI.

AI-led Services

AI-led services replace labor arbitrage businesses with AI-native operations. Instead of scaling headcount linearly, these companies use AI to deliver professional services - accounting, legal, insurance, healthcare administration - at 10x lower cost with comparable or better quality and adding additional strategic value. The business model isn’t SaaS, it’s about delivering outcomes and charging for results.

India has decades of domain expertise powering global services businesses. The talent exists - chartered accountants, lawyers, insurance underwriters - at $80K/year versus $400K in the US. India already runs more than a third of global healthcare BPO operations. The infrastructure is proven. The arbitrage now is replacing that labor with AI-led workflows while keeping the domain knowledge in-house.

Month-end Close Automation

Mid-market companies ($50M revenue) pay $60K-$120K/year for accounting services that take 10 days to close each month. We think building an AI-native accounting firm that does real-time reconciliation from bank feeds, automated journal entries, and same-day month-end close is a big opportunity. The wedge is speed (same-day vs 10 days) before expanding into a full financial operating system for every SMB.

Startup Legal Automation

Startups spend $15K-$25K on incorporation, contracts, and employment agreements in year one. We think an AI-native law firm for tech/modern companies that automates incorporation, standard contracts, and regulatory filings, starting with commoditised services where buyers tolerate comparable quality for 10x lower cost, before expanding into AI-assisted due diligence, is a massive whitespace to build in.

The moat here comes from a proprietary contract database and a regulatory corpus. Whoever builds the best retrieval architecture for legal documents likely wins.

Healthcare Claims Processing

US healthcare administrative costs are over $1 trillion annually. Claims take 30-60 days to process. Denial rates hit 15-20%. An AI-native revenue cycle management (RCM) company that does automated medical coding, real-time eligibility checks, same-day claims submission, and predictive denial management could capture a lot of the value there.

This is a must-have for providers - you improve their cash flow immediately. And given India already powers major healthcare BPO operations, the next generation great Indian companies in this domain will be AI-led services companies that own the entire workflow, not just offshore labor.

Commercial Insurance Brokerage

Commercial insurance brokerage generates $80-140B in annual revenue in the US alone. Corporate clients still get quotes manually, wait days for policy documents, and deal with brokers capturing 10-20% margins on relationships rather than service quality.

The AI-native brokerage that gets instant quotes across multiple carriers, has dynamic pricing models, and has 24/7 AI agents handling renewals and endorsements. The dominant business model here is transaction fees at scale and not SaaS. You’re not selling software but are delivering insurance outcomes and capturing the spread.

Corporate Travel Optimisation

Corporate travel is a $1.4 trillion global market. Companies still book through legacy TMCs charging service fees while offering limited optimisation. Building the AI-native travel platform that learns company travel policies, predicts trip patterns, negotiates dynamic rates with suppliers, and auto-books optimal itineraries is a massive opportunity.

The insight here is that travel data creates compounding advantages. Every booking improves pricing models and supplier negotiations. The right wedge could be to start with mid-market companies, prove 15-20% cost savings, then expand to enterprise.

Tax-filing Automation

SMBs pay accountants $5K-15K annually just for tax filing—work that’s largely mechanical data entry and form completion. Build AI that connects to accounting systems, auto-generates tax returns (federal, state, local), files electronically, and handles amendments. Charge $500-$1,000 per return—10x cheaper than human accountants, 90% gross margins.

The moat here is building trust with SMB owners who fear tax audits and not the AI. Offering audit insurance and white-glove support for the 5% of returns that get flagged is an additional strategic value that upstarts could bundle into their own offering.

Freight Forwarding

Global freight forwarding is a $200B+ market. And yet, freight forwarders still coordinate shipments manually - calling carriers for quotes, negotiating rates, tracking containers via email chains, filing customs paperwork. The workflow hasn’t fundamentally changed in 30 years despite being entirely information-based.

An AI-native freight forwarder could be someone that does automated carrier selection based on route/cost/reliability, instant customs documentation, predictive delay alerts, real-time shipment visibility.

We think a good wedge here could be to start with India-US trade lanes (leverage regulatory knowledge), then expand to Asia-Europe corridors. And the right business model for this is to charge transaction fees per shipment. The moat comes from developing dee carrier relationships + proprietary rate/reliability data.

IT Services: Enterprise Software Implementation

Enterprise software implementation is a $50B+ annual market. Companies spend $300K-$2M implementing Salesforce, SAP, or ServiceNow, with projects taking 6-18 months. Systems integrators bill $150-$300/hour for work that’s 60% configuration, 20% data migration, 20% custom development.

AI Agents that automate the repetitive 80%: read existing business processes from documentation/interviews, generate configuration recommendations, auto-migrate data with validation, create custom workflows via natural language would result in implementation time dropping from 12 months to 3 months at 1/3rd the cost.

Indian IT services firms (TCS, Infosys, Wipro) do $100B+ annually in traditional implementation. The AI-native player that proves faster/cheaper implementations will take market share rapidly. Charge fixed-price per module implemented—customers prefer predictability over hourly billing.

AI Services for AI Implementation

Every enterprise deploying AI faces the same problems: unstructured data handling (PDFs, images, audio), vector database management, prompt engineering pipelines, model evaluation frameworks, compliance/governance tracking. They’re building these capabilities in-house because no standard “AI implementation services” category exists yet.

We think there’s a large opportunity to build the AI services firm for AI implementation: data pipeline setup for unstructured data (OCR, audio transcription, video analysis), vector database architecture and optimisation, RAG system development, evaluation harness creation, compliance documentation for regulated industries.

This is a greenfield category. Fortune 500 companies will pay $500K-$2M for AI implementation services, similar to what they pay for cloud migration or ERP implementation. India’s IT services DNA married with deep AI expertise should result in a winning combination. The right wedge could be to start with a vertical that has higher willingness to pay & has strict compliance needs.

Data Training & Labelling

The market for data labelling is now moving from high-volume, low-value, general-purpose tasks to specialised, high-value tasks spanning complex modalities like: voice transcription with dialect labeling, video action segmentation, 3D point cloud annotation for robotics, multimodal data pairing.

India’s advantage in its operational excellence in managing distributed labelling teams and quality control processes lend itself naturally to the domains of specialised labeling for voice (Indic languages, code-switching patterns), video (action recognition, scene understanding), and physical AI (LIDAR, sensor fusion).

Consumer AI

Consumer AI in India will reshape how people learn, earn, heal, and express themselves. The real opportunity lies in reimagining everyday systems like education, healthcare, finance, and commerce through deeply personal, culturally aware AI experiences.

For India’s 700 million internet users, the most impactful AI products won’t be tools but companions, tutors, advisors, marketplaces built on trust and local context. AI must understand exam culture and family dynamics, adapt to home remedies and local diets, navigate UPI-led micro-transactions, and speak in regional languages rich with idioms and emotion.

With its digital public rails, linguistic diversity, and mobile-first behavior, India has the chance to leapfrog the world in consumer AI, creating products that don’t just localize global ideas but set new global standards.

AI Primary Care Doctors

Indian households can’t afford ₹500-1,000 per doctor visit for common ailments. Clinics are overcrowded - 1.3-1.4M doctors serve 1.4 billion people (1:1000 ratio). Telemedicine startups showed demand exists but hit quality ceilings with human doctors.

There’s an opportunity to build AI doctors for primary care: symptom assessment via conversational interface, preliminary diagnosis with confidence scoring, treatment recommendations (OTC medications, lifestyle changes, when to see human doctors), prescription delivery integration. Critical product features become the ability to work in Hindi and 10+ regional languages, understand local disease patterns (dengue, malaria seasonality), and recommend local medications available in India.

AI Tutor

Indian families spend $10-11B annually on private tutoring and coaching, largely for competitive exam prep (JEE, NEET, UPSC, CAT). Existing edtech provides recorded lectures but lacks personalisation - a student struggling with calculus gets the same video as everyone else.

With dropping costs, there’s an opportunity to build AI tutors that provide personalised learning: adaptive question practice (easier/harder based on performance), instant solution explanations in student’s preferred language, topic mastery tracking aligned to exam syllabi, doubt resolution via conversational AI, peer comparison to identify improvement areas.

A good place to start could be JEE/NEET math and physics as the content there is highly structured with repeated problem patterns post which one can scale to UPSC, CAT, and state board exams once product-market fit is proven.

The moat is likely to come from proprietary question banks, performance data showing learning velocity improvements, integration with test series platforms.

AI Farmers

India has 145 million agricultural households, 86% owning <2 hectares (smallholder farmers). They face information asymmetry on: when to plant/harvest based on weather, pest/disease identification, fertilizer recommendations, market prices for selling produce. Government extension services reach <20% of farmers.

AI farming advisors that have a voice-first interface in regional languages (Hindi, Punjabi, Marathi, Telugu, Tamil), provide hyperlocal weather forecasts with planting recommendations, image-based pest/disease diagnosis via phone camera, soil testing interpretation, market price alerts for selling at optimal times, government scheme eligibility checking is a huge opportunity.

The right place to start could be in Punjab/Haryana (wheat/rice, high smartphone penetration, commercial farming mindset) and prove out 10-15% yield improvements or cost savings before potentially scaling to states like Maharashtra, Karnataka, Andhra Pradesh.

The hard problems are making AI work offline or in low-bandwidth conditions (2G/3G). And the smart distribution hack here could be to use WhatsApp as distribution channel, and not build a separate app.

AI Astrologers

Indians spend approximately $500M-$1B annually on astrology consultations, with estimates of the online market at $163M in 2024 and growing to $1.8B by 2030. Think of AI-native astrologers that deliver personalised predictions based on birth charts, planetary positions, and life events.

The key is to deliver outcome-based trust where users pay when predictions prove accurate/helpful. One could start with highest intent wedges like Kundli matching for marriages before expanding to career guidance and auspicious timing (mahurat). Voice form factor matters more here than text. And the sticky user behaviour is going to come from someone who makes it sound like the family astrologer and not an AI chatbot.

AI Nutritionists

Indian dietary needs are distinct - vegetarian/vegan preferences on different days of the week, regional cuisines, religious restrictions, and Ayurvedic principles. Existing nutrition apps are built for Western diets. We seen an opportunity to build AI nutritionists that understand dal-chawal-sabzi combinations, suggest swaps within regional cuisines, and create meal plans around Indian grocery availability.

The right products could be monetized by charging for personalised meal plans (₹999/month) and even potentially partnering with grocery delivery for affiliate revenue as well as upselling to fitness coaching bundles.

AI Wealth Advisors

Indians pay $1.5-2B annually for wealth advisory services, but most middle-class families can’t afford ₹15K-40K yearly fees. Thus, AI wealth advisors that recommend mutual funds, optimise tax (80C, ELSS), plan for goals (child education, home purchase), and rebalance portfolios is a big opportunity in an existing whitespace.

The best entry point is to start with simple tax optimisation - immediate, measurable value before expanding to goal-based investing. Trust is likely to come from tangible outcomes: “We saved you ₹45K in taxes this year.”

AI Companions for Young Adults

Character.AI has shown that AI companions are a multi-billion dollar category globally, with 20M+ monthly active users. But the India-first AI companion hasn’t been built yet. It needs to understand the Indian cultural context: navigating family expectations about marriage and career, code-switching between English and Hindi/regional languages, and discussing cricket and Bollywood without Western references.

Voice-first experience is critical, along with innovations on the memory front to make it sound like a friend, not an assistant. We believe that the best products will be monetized via a paid subscription from the get-go price-gated on long-term memory/conversation access, priority response, voice calls, etc.

Micro-Dramas

India’s 600-750M short-form video consumers spend 45-60 minutes daily on YouTube Shorts, Instagram Reels, and Moj. But content creation is still expensive. An AI-generated micro-drama that has 60-90 second episodic content with serialised narratives, hyper-personalised to user preferences, leveraging the drop in cost of content creation with Generative AI, enabling infinite niche genres - fantasy content, saas-bahu drama shorts, horror episodics, sports commentary reels, among others.

AI-remixed Casual Games

India’s mobile gaming market has 500M+ users, but most games are Western ports. Building casual games where levels, difficulty, and narratives adapt in real-time based on player behaviour is a big opportunity using Generative AI to create user-generated content that gets AI-remixed, resulting in infinite content, minimal development cost.

A nice place to start could be simple puzzle games (Candy Crush-style) and then adding regional language support, integrating UPI for micro-transactions (₹10-50 per power-up). Infinite content creates infinite engagement as well as monetisation potential for the power users - and retention is everything in gaming.

AI Shopping

E-commerce today is built for search, not discovery. But in India, shopping is emotional, social, and deeply contextual, shaped by festivals, trends, influencers, & family recommendations. AI can move commerce from transactional to intuitive, from browsing to understanding.

An AI-native shopping layer could learn your taste the way Spotify learns music: reading aesthetic cues, budget habits, cultural moments, and peer signals. It could generate personalised storefronts, bundle outfits across retailers, or simulate products in your life before purchase.

On the backend, AI can rewire the entire stack: dynamic merchandising that adjusts to real-time demand, supply chains that predict rather than react, marketing that replaces ads with conversation, and retention engines that personalize every touchpoint. The result: lower CAC, faster turns, and higher repeat rates.

Sovereign AI

Sovereign AI is about building India’s AI infrastructure stack - from silicon to models to inference - optimised for Indian use cases, languages, and regulatory requirements. Foundation models are commoditising for general tasks, but they’re not trained on Indian cultural context, they incur latency from US/EU data centers, and they don’t meet data residency requirements for regulated industries. The opportunity is building AI infrastructure that Indian startups and enterprises must use due to economic, regulatory, or quality advantages. Most critically, Indian builders need to deeply understand the local context that global model providers ignore: regional languages, cultural nuances, regulatory constraints, and cost sensitivities.

Purpose-Built Voice & Multimodal Models for India

The speech models from OpenAI, Google & ElevenLabs are trained predominantly on Western English. They fail on Indian accents, code-switching (mixing Hindi-English mid-sentence), and regional languages. The Vision models don’t recognise Indian cultural context either - they can’t identify a rangoli, explain the significance of Diwali decorations, or understand visual references from Bollywood.

Thus, this presents an opportunity to build purpose-built foundation models for India: voice models trained on Indian accents across 22+ languages with code-switching support, vision models that understand Indian cultural artefacts (clothing, food, festivals, regional architecture), multimodal models that combine voice, vision, and text for Indian use cases.

The moat lies in proprietary training data: 100,000+ hours of conversational Indian speech, 10M+ labelled images of Indian cultural contexts, curated text from Indian literature, news, and social media.

The GTM here is to license models to Indian startups building voice assistants, customer support bots, educational apps and monetize that via APIs & Agents.

Inference-as-a-Service with India Advantages

Every Indian AI startup faces the same problem: running inference on Anthropic/OpenAI models incurs 200-400ms latency from US/EU data centers. For voice applications (customer support, tutoring), this latency is unacceptable. Healthcare and fintech applications require data residency in India for regulatory compliance. And international API pricing (OpenAI charges $10-$60 per million tokens) eats into margins for high-volume use cases.

An inference-as-a-service infrastructure optimised for India that hosts popular open-source models (Llama, Mistral, Qwen) and India-specific models in Indian data centers (Mumbai, Bangalore), delivers <50ms latency, guarantees data residency compliance for DPDP Act and RBI regulations, prices at 30-50% below international APIs due to lower Indian infrastructure costs is a big opportunity.

This could become the default inference provider for Indian AI startups. The flywheel could look something like this: aggregate demand across startups to negotiate better GPU pricing, optimise inference kernels for cost efficiency, add value-added services (prompt caching, batching, fine-tuning).

Custom Silicon for AI Workloads

The AI compute stack is disaggregating. NVIDIA GPUs are generalist - great for training and real-time inference but overkill for batch inference, inefficient for specific modalities (voice, vision), and expensive ($30K-$40K per H100). The next wave is custom silicon optimised for specific AI workloads: batch inference (processing millions of documents overnight), real-time inference (voice assistants requiring <100ms latency), modality-specific inference (dedicated vision chips, dedicated voice chips).

We’re bullish on the custom silicon opportunity for AI: inference-optimised ASICs that deliver 3-5x better performance/watt than GPUs for specific workloads, targeting the $10K-$15K price point (much lower than H100s). The initial product wedge could be on batch inference workloads (data labeling, document processing, offline model evaluation) where latency flexibility enables better hardware optimisation.

Unique Indian Datasets as Moat-Building Assets

The most defensible AI companies have proprietary datasets that competitors can’t easily replicate. India offers unique datasets that global players haven’t captured yet: regional language text (literature, news, social media in 22+ languages), vernacular voice data (accents, dialects, code-switching patterns), Indian medical records (disease patterns, treatment outcomes for tropical diseases), agricultural data (soil types, crop yields, weather patterns by micro-region), financial behaviour (UPI transaction patterns, credit behaviour for thin-file customers).

Building dataset companies that collect, clean, and monetize Indian-specific data is a big opportunity: by partnering with state governments for public datasets (land records, agricultural data), aggregating data from healthcare providers (with consent and de-identification), collecting voice data through consumer apps, curate text from regional publishers and content platforms.

AI for life-sciences

Modern protein design, drug discovery, and agricultural biotech are empirical discovery problems in combinatorially large search spaces. There are 10^180 possible protein sequences for a 150-amino acid protein - more than atoms in the observable universe. Traditional wet lab approaches test 10^3-10^4 candidates per year. AI can navigate this search space computationally, predicting protein structures, drug-target interactions, and crop trait optimisation orders of magnitude faster than lab experiments.

India is the world’s pharmacy - producing 60% of global vaccines and 20% of generic drugs by volume. The country has deep biotech talent (IISc, CSIR labs, pharma R&D centers) at Indian costs. Clinical trial costs are 40-60% lower than US/EU. Agricultural biotech expertise spans ICAR institutions and agritech startups. And regulatory pathways through CDSCO and GEAC move faster than FDA/EMA for certain applications.

India AI News Roundup

The most impactful AI developments & announcements shaping India in recent weeks.

Vinod Khosla: AI, an enabler of universal access, increased equity, and global influence for India (ET Guest Column)

MeitY releases India AI governance guidelines to build safe innovation ecosystem

Accel partners with Google to back Indian AI startups with $2M co-investment

Sarvam.ai to launch India’s first LLM by early next year

India is set for its first AI IPO as Fractal Analytics gets SEBI nod

Startup Signals

Spotlighting brand new emerging AI startups from India every month, early and undiscovered.

Bolna.ai — Voice AI for Indic Languages

Bolna powers conversational voice agents across Hindi, Tamil, Telugu, and 10+ Indic languages. Built specifically for Indian accents, code-switching, and regional dialects with <300ms latency. High-quality voice for customer support, sales outreach, and user engagement. Current traction: 40+ enterprise customers, processing 2M+ voice minutes monthly.

https://www.bolna.ai/

MedSee.ai — AI-Powered Radiology Workflow Platform

MedSee is building the universal AI-powered radiology workflow platform that streamlines imaging, reporting, and collaboration. For radiologists, it accelerates diagnosis with AI-assisted reporting, intelligent highlighting of abnormalities, and auto-fill measurements directly into smart reports. For clinicians, it provides instant access to complete imaging insights with seamless collaboration. For patients, it delivers clear, comprehensive digital reports with embedded reference images. Currently in private beta with select radiology centers across India.

https://medsee.ai/

August AI — 24/7 AI Health Assistant on WhatsApp

August is India’s AI health assistant delivering personalised medical insights via WhatsApp. Multi-modal platform accepts voice notes, text, and lab reports (PDFs) to provide expert-level health guidance. Scored 94.8% on USMLE—outperforming GPT-4 (87.8%) and Google MedPaLM 2 (86.5%). Built by Beyond, a Bengaluru based health-tech startup founded in 2022. Currently serving 2.5M+ users and 100K+ doctors across 148 countries. The platform helps users understand health concerns, analyze reports, manage medications, and schedule appointments—available free on WhatsApp.

https://www.meetaugust.ai/

This edition of Signal maps India’s top Generative AI companies across 4 layers - Infrastructure & Core AI, Horizontal Agents & Automation, Vertical Applications, and Consumer AI - to showcase depth, not just breadth.

The methodology is simple: we focused on Gen AI natives (founded in the last 2-3 years) and trajectory shifters (companies whose growth fundamentally accelerated because of generative AI), with the end result being a list of 75 breakouts. We tracked funding velocity, technical ambition, and adoption signals. What emerged is a full-stack ecosystem and signals of where India’s AI economy is being built right now.

Why India’s AI activity looks different?

India’s AI companies are building for fundamentally different realities:

• Multilingual by default: English is one of many. Sarvam, Krutrim, and Gnani.ai exist because Hindi, Tamil, Bengali, and 18 other languages must be first-class citizens.

• Voice > Text: India is a country where a lot of consumer services are delivered voice-first, leading to the early success of companies like Nurix and GreyLabs.

• Vertical and services-led: Decades of IT, BPO, and SaaS expertise mean India’s builders understand how real workflows run, from customer support to logistics to finance. Many of India’s AI startups are emerging not from hot trends, but from reimagining complex boring problems with automation, context, and scale.

• Mobile-first, Android-heavy: 95%+ smartphone penetration is Android. Apps must be lightweight, work on 4G (not 5G), and assume intermittent connectivity.

• Young, exploratory user base: 65% of India is under 35. Consumers are willing to experiment with AI companions (Rumik, Mello), AI entertainment (Dashverse), and AI-first workflows (Kruti agents) in novel ways than what’s available in Western markets.

Just as India leapfrogged landlines to mobile, and physical banking to UPI, the country is now leapfrogging GUI-based software to agentic, voice-first AI. This is a distinct paradigm that may preview global behavior.

Layer 1: Infrastructure & Core AI

Let’s start at the bottom of the stack - the foundational layer that determines what’s possible for everyone building above it.

For years, India was an integrator of AI, not a producer of it. Companies licensed models from OpenAI, Anthropic, or Google, then built applications on top. That’s slowly changing. A cohort of ambitious teams is now building the rails themselves: sovereign language models, GPU compute infrastructure, hardware alternatives, and developer tooling.

This isn’t just about national pride or data sovereignty (though both matter). It’s about economic necessity. As data residency regulations tighten globally, as inference costs become a meaningful line item, as customisation needs exceed what general-purpose models can provide, owning the infrastructure becomes strategic.

Frontier Models

· Sarvam AI: Full-stack generative AI platform for India—both models and application layer tools, optimized for Indic languages

· Krutrim: India’s first large-scale multilingual foundational LLM, positioning as sovereign AI infrastructure

· Two AI: Glocalized AI models designed for the “next billion” users across emerging markets

· Bharatgen: Frontier research and model development

Compute & Hardware

· Neysa: GPU cloud for enterprises needing India-resident compute—data sovereignty, latency, cost efficiency

· Agrani Labs: Building Nvidia alternatives—hardware infrastructure for AI compute designed for cost-efficiency and accessibility

Developer Tools

· Composio: Integration platform connecting AI agents to 150+ enterprise tools—the connective tissue between deterministic software and non-deterministic AI agents

· Portkey: LLMOps stack for monitoring, managing, and routing across multiple models

· True Foundry: End-to-end ML deployment and scaling toolkit for production environments

· Maxim AI: GenAI evaluation and observability—critical for moving from pilots to production

· Simplismart: Build, deploy, and manage custom deep learning models

Layer 2: Horizontal Agents & Automation

If Layer 1 is the operating system, Layer 2 is the application framework. This is where the action is right now - companies building AI that doesn’t just answer questions, but executes workflows autonomously.

The promise of agentic AI is simple: instead of “copilots” that suggest next steps, you get systems that complete entire workflows: qualifying leads, handling support tickets, managing procurement, drafting contracts.

What’s interesting about India’s agentic companies is that they’re not just building tools for developers; they’re also building platforms that let any enterprise create custom agents without writing code.

Agentic AI - Horizontal

· UnifyApps: No-code AI agent builder for enterprises—drag-and-drop interface to create agents that orchestrate across SaaS tools

· Ema: Autonomous enterprise automation—agents that handle end-to-end processes like procurement, compliance, and workflow orchestration

· Atomicwork: Modern ITSM/ESM platform powered by universal agents (voice, vision, action AI) for HR and IT workflows

· Sifthub: AI sales engineer that qualifies leads, answers RFPs, handles technical objections autonomously

· Metaforms: Automates customer support and engagement via agentic workflows

· Thesys: AI-powered design execution engine for infrastructure software

· Emergent Labs: Agentic vibe-coding platform—writes, tests, and deploys code from natural language prompts

· DhiWise / Rocket.new: Programming platform converting designs into developer-friendly code for mobile and web

· Presentations.AI: AI-powered platform automating presentation creation

Voice & Conversational AI

· Nurix AI: Voice & chat assistants for enterprises—boosting conversions and cutting support costs

· GreyLabs AI: Agentic voice AI for India’s BFSI sector—conversational analytics platform

· Smallest: Voice AI solutions for contact center use-cases (lightweight, mobile-optimized)

· Gnani.ai: Speech recognition, text-to-speech, and voice agents focused on Indic languages

Layer 3: Vertical Applications

This is where India’s ecosystem is deepest - domain-specific, data-rich, and deeply integrated vertical apps that are imagining how industries actually work. Doctors diagnosing faster, lawyers drafting smarter, teachers educating better, and creators building sharper.

Instead of building generic copilots, founders are embedding intelligence into the workflows and datasets of a single domain, where context is king. This is where real defensibility, customer love, and monetization start to appear.

And nowhere is this shift more natural than in India - a country built on services, outsourcing, and process expertise. The same DNA that powered India’s IT & BPO revolution is now fuelling a new wave of vertical AI companies, all of who require human-in-the-loop expertise. From finance to logistics to education, the country’s next breakout AI companies are likely to be vertical, not viral.

Healthcare

· Qure.AI: Radiology AI for TB, stroke, chest scans—deployed across India and globally

· Niramai: Non-invasive, AI-driven early breast cancer detection

· Eka Care: Personal health records + AI-based diagnostics

· RapidClaims: Autonomous medical coding and documentation platform

· Spry: Cloud-based practice management + RCM for physical therapy clinics

· Confido: AI-powered healthcare solutions

· August AI: True health companion—personalized wellness and care guidance

· Wysa: AI mental health companion with global adoption (3M+ users)

· Zoca: AI-native voice driven marketing solutions for salons and spas

Fintech

· OnFinance AI: Full-stack compliance OS for financial services

· Moneyview: AI-driven digital lending and financial management

· Alltius: AI assistants for enterprises and BFSI domain

· Wokelo AI: Human-driven, AI-assisted’ platform for knowledge workers in business research

· Soma: Commercial insurance for complex risks with sub-10-minute response times

Education

· Stimuler: Audio-AI to help the global population speak English better

· Professor Curious (fka ZuAI): Autonomous test prep that mimics expert tutors

· Arivihan: India’s first fully automated vernacular learning platform

Media & Creative AI

· Dashverse AI: Full-stack AI entertainment company—creating AI-native content as production costs collapse

· Gameramp: AI Agents and APIs to automate LiveOps, better forecast LTV and unlock scale

· Gan.ai: Personalized video generation at scale for marketing and sales

· Pixis: Generative AI automating digital marketing campaigns

· Beatoven.ai: AI-generated, royalty-free music for creators and brands

· Dashtoon: Platform to create, publish, and monetize AI-generated comics

· Invideo AI: Video creation with text-to-video features

· Murf: Lifelike synthetic voice generation

· Writesonic: SEO-optimized, plagiarism-free content creation

· AfterShoot: AI-based image culling and editing for photographers

· Runable: General-purpose AI Agent for creative content workflows

Industrial & Manufacturing

· Pre6: Vertical AI agent for manufacturing use-cases—automating production workflows, quality control, and operations

· Systemantics: Industrial robots for factory automation

· Ati Motors: Autonomous industrial vehicles for warehouses and logistics

· Planys: Underwater robotics powered by AI for inspection and maintenance

Agriculture

· Kissan AI: AI agent platform for agriculture—advisory and market linkage

· Krishi Sathi: Conversational AI for farmers (crop advisory, weather, pricing)

Legal Tech

· Lexlegis AI: Using AI to power justice at scale

· SpotDraft: Simplify contracting with LLMs—templates, chat querying, smart insights

· Lucio: AI powered document review, legal research and drafting

AI-Enabled Services

· Aeos Labs: AI services for videos and creative tools

· Pepper Content: AI tools for content creation, distribution, and measurement

· Realfast: AI-first Salesforce implementation and consulting company

· Kim.cc: Agentic BPO for Customer Support

· Squadstack: AI-enabled outsourcing/operations for call centers

· Gushwork: AI-assisted SEO for revenue generating traffic

· Bridgetown Research: Gather and analyze primary & secondary data at unprecedented scale, speed, and cost efficiency

Layer 4: Consumer AI

This is where India diverges most sharply from the West. AI companions, mental health chatbots, astrology bots - things that are unique to Indian consumers.

India, with its massive mobile-first population and comfort with digital experimentation, is uniquely positioned to lead here. Just as India skipped desktops to go mobile, it may skip “tools” to go straight to AI-native experiences.

Founders are blending design, emotion, and intelligence, building products that feel less like software and more like superpowers. In a world full of AI infrastructure, the next big consumer companies may be built on delight, not datasets.

· Kavana AI: Playful, personal, and immersive AI-first apps

· Rumik AI: Personal AI built for humans—conversational companion

· Mello AI: AI friends designed for Indian users

· HiAstro: AI-powered astrology and personalized insights

India AI News Roundup

The most impactful AI developments & announcements shaping India in recent weeks.

Google stakes $15B to build a gigawatt-scale AI hub in Visakhapatnam

Anthropic to open India office next year, CEO Dario Amodei announced on his recent India visit

Sarvam AI lands Meta partnership to power Ray-Ban Glasses with full Hindi support

Reuters: AI chatbots are displacing India’s call-center work with 70-90% automation

IndiaAI Impact Summit to be held on February 19-20, applications for pre-events now open

Startup Signals

Spotlighting brand new emerging AI startups from India every month - early and undiscovered.

Specvia — AI for Manufacturing Specifications

Specvia is re-imagining the engineering workflow by making specifications themselves first-class, structured, and AI-augmented. Instead of treating specs as documents, they treat them as living, traceable artifacts — a shift that resonates well within complex manufacturing-led engineering orgs.

https://www.specvia.com/

Sing One Song — AI Coach for Singing

Sing One Song is bringing AI into the world of vocal coaching, lowering the bar for anyone to discover their “signature song” and improve through voice analysis. It’s a reminder that consumer-AI isn’t just about chatbots or assistants; it can also be about self-expression, enabling human skills with AI.

https://singonesong.com/

Kello AI — AI for Recruiting in India

Kello AI is reimagining hiring — using AI to source, shortlist, and rediscover hidden talent across millions of profiles. Built for how teams in India actually hire, it’s turning recruitment into an intelligent, continuous workflow.

https://kello.ai/

GPT-5 was supposed to be the AGI moment. Sam Altman posted a death star 1 day before launch, beta users were posting how software is dead & jobs are gone, etc. full on Terminator hype.

What we finally got was a great model, perhaps better than anything else on the market, but was it really that exponential? The unanimous answer to this is not quite. It’s definitely a big jump for free users moving from 4o, but to those who were used to o3, it was another incremental improvement.

2. MIT Report: 95% of GenAI pilots at companies are failing

A week later came this report from MIT’s NANDA lab claiming that only 5% of GenAI enterprise pilots convert into full deployments. Challenges included integration with complex data systems, broad scopes for agentic AI, and models not fully replacing humans. It said enterprises are extremely eager to adopt AI, but underwhelmed by the practical ROI impact these tools are having 2+ years into the cycle.

This report is full of amazing insights, including what the most successful enterprise buyers are doing. If you’re interested in getting access, please respond to this email and we’ll send it to you 1:1.

3. Meta hiring freeze

Finally, along the same time, The Wall Street Journal reported that Meta has frozen AI hiring after going on a crazy spree last few months. Some outlets also reported that they will consider restructuring including layoffs in the AI teams.

Here’s our take: this isn’t “peak GenAI,” it’s hype getting repriced to operations. GPT-5’s launch reminded everyone that progress is real but non-linear. Stronger reasoning and coding, yes; an AGI moment, no. That gap between sky-high expectations and day-one experience, plus early rollout hiccups, simply accelerated an overdue reset. MIT’s report pins the bottleneck where it belongs: data access/quality, integrations with systems of record, workflow redesign, governance, and change management. Meta’s hiring pause is just incidental at the same time as everything else adding to the noise, but this was bound to happen after giving out $100M+ offers. We welcome all of it. The next phase is about execution quality, not model access.

Enterprise

Practically, in the enterprise, value will accrue when you can turn models into closed-loop, agentic workflows with humans in the loop where it matters: tickets resolved, claims processed, code merged, orders fulfilled. The game is not about tokens or latency any more - those are input metrics. It’s instead about cost per completed task, cycle-time compression, error rates, and P&L impact. Builders should anchor on three things: (1) proprietary or privileged data advantages in narrow use cases, (2) rock-solid integrations and reliability in messy enterprise environments, and (3) outcome-based pricing with clear SLAs. Buyers should insist on reference architectures, evals that measure real-world performance (not demo scripts), and governance guardrails from day one.

For India, this shift is both a risk and an opening. The risk: large IT services firms—built on headcount-heavy delivery, face pressure as global buyers start to treat AI vendors less like SaaS providers but more like consulting or BPO partners, demanding deep customization and operational outcomes. That means Indian outsourcing could feel the squeeze if it doesn’t retool for AI-first delivery. But the opportunity is equally large: India has the talent density, service DNA, and cost base to become the world’s integration and orchestration hub: building, customizing, and operating agentic workflows at scale. Startups that embrace the “business service provider” mindset, and marry domain expertise with proprietary data, can carve out defensible niches. For the country, the play is moving up the stack: from low-cost labor arbitrage to outcome-priced AI operations. If India executes, it doesn’t just weather the shift, it becomes the back office of the AI decade.

Consumer

The hype is cooling into durable utility: the near-term winners won’t be sci-fi “AGI companions,” but everyday helpers that compress chores into taps: drafting and editing, shopping and travel planning, photo/video cleanup, tutoring, and lightweight personal finance. Distribution favors incumbents (OS, keyboards, messaging, browsers) as assistants become an ambient feature rather than a standalone destination. That raises the bar for independents: pick a narrow, high-frequency job-to-be-done, build a tight feedback loop with user data (with explicit consent), and outperform the platform default on speed, reliability, and familiarity. Personalization will matter more than parameter count: on-device + hybrid models unlock low-latency, privacy-preserving experiences that remember preferences, style, and context without feeling creepy.

For India, this transition has two sharp edges. On the one hand, distribution gravity tilts toward U.S. and Chinese platform giants whose apps dominate Indian consumer attention, like WhatsApp, Android keyboards, YouTube, Chrome. Startups that try to go head-to-head as “general AI assistants” will likely get squeezed. On the other hand, India’s unique consumer internet landscape such as vernacular content, UPI-led payments, aspirational creators, and price-sensitive mass adoption, creates distinct local use cases that global incumbents won’t prioritize. Think: AI tutors tuned for CBSE/State syllabi, shopping concierges that reconcile UPI offers across e-commerce sites, regional-language content cleanup/editing, or micro-SaaS tools for the 100M+ Indian freelancers & SMBs. Startups that combine cultural context with domain specificity can thrive even within platform ecosystems.

At a country level, the opportunity is profound: if India builds trusted, low-cost, and localized consumer AI utilities, it can leapfrog digital adoption curves in education, healthcare, and commerce, creating millions of “AI-enabled users” who would otherwise be left behind. The risk is ceding consumer surfaces entirely to global platforms and watching value capture flow abroad. The playbook: embed in the platforms Indians already use, but own the niche where local data, context, and trust matter most.

India AI News Roundup

The most impactful AI developments & announcements shaping India in recent weeks.

Reliance Intelligence is the new Jio for Mukesh Ambani: $100M JV with Meta for Enterprise AI in India

OpenAI Plans India Data Center in Major Stargate Expansion

Indian AI startup Emergent hits $10M ARR in two months from launch

E2E Networks shares rise 10% after winning ₹177 crore order from MeitY’s IndiaAI mission

AI Impact Summit 2026: India sets stage for global Artificial Intelligence leadership

Abhishek Singh, Mitesh Khapra of India featured on the 2025 Time 100 AI List

Startup Signals

Spotlighting breakout AI startups from India every month—fresh ideas, bold founders, and the signals worth tracking.

NyayaNidhi – India's first AI Law Firm

India’s legal system generates millions of documents annually, but most remain trapped in PDFs and archives. NyayaNidhi brings structure to this data by automating case analysis, precedent research, court-ready documentation, and digital case tracking across languages and jurisdictions. Nyayanidhi is one of the world’s first full-stack AI law firms, where the AI-native digitisation redefines how enterprises engage law firms in a transparent and 10x more efficient way.

https://nyayanidhi.com/

Smallest AI – Real-time Voice Agents for Financial Services

Smallest AI powers lightning-fast, human-sounding voice agents that handle complex contact center calls with near-zero human fallback. From auto-dialers and call rescheduling to voice cloning and real-time TTS, it’s building the infra layer for AI-powered enterprise voice at scale.

https://smallest.ai/

MetaForms – AI Infrastructure for Faster, Smarter Market Research

MetaForms is rethinking how surveys are built, deployed, and analyzed. Its platform uses AI to turn raw briefs into ready-to-run surveys, optimize form logic, and track results in real time. It’s giving market research teams an edge on speed, accuracy, and scalability and all without reinventing their workflow.

https://www.metaforms.ai/