Subscribe now

💭 Thought for the week: Was your record year mostly inflation?

Four weeks ago we ran the numbers on a shrinking donor base and said a rising topline was hiding a structural break. A few of you replied to say your board had seen the same chart and concluded you were having a good year.

Data published last Wednesday confirms otherwise.

The base is still going. RKD Group’s mid-year benchmark covers 219 nonprofits across the US and Canada through 30 June 2026. Active donors are down to 2,585,127 - 2% off last year, and 22.2% off the 2020 peak. More than a fifth of the base has gone in six years, and it is still going.

The dollars have stopped covering for it. Revenue per active donor reached $229, a ten-year high, up 27.2% since 2021. That is the number that lands in a board pack. However, adjusted for inflation, it is up 3.6%. Almost the entire record is the currency, not the relationship. We charge roughly what we always did to a fifth fewer people, and the nominal figures have been kind enough to hide it.

The concentration got worse. Gifts of $10,000 and above now make up 68.3% of all revenue, while everything below $10,000 fell 16.9%. General gifts are at their lowest revenue share in a decade. In August we called this concentration risk and said a CFO would refuse it in any other line.

If your income looks flat and your major gifts look healthy, this is the shape you are in. It is not a management failure. It is what a decaying acquisition engine looks like.

The part that should make you optimistic

Two things in the data grew, and they are the same thing in different clothes. Monthly giving revenue hit a ten-year high of $114.9M, up 29.9% since 2021. Mid-level giving grew 7%. Everything transactional shrank; everything built on a continuing relationship expanded, across the same six years that took a fifth of the base away.

RKD’s own read is that monthly giving should be the default acquisition offer rather than an upgrade pitched later. Put plainly: the only parts of the file still growing are the parts where the supporter has a reason to be there next month.

That is the same conclusion as four weeks ago, reached from the other end. Then it was the second-gift problem: fewer than one in five first-timers return, against 65–69% for repeat donors. Two datasets, two countries, one answer. The relationship compounds and the transaction does not.

Running that relevance across a whole file used to be a big-org luxury. It isn’t any more, and that’s the shift we said we’d walk through touchpoint by touchpoint.

Three moves this month

Re-run your $10k line. What share of revenue came from gifts of $10,000 and above last year, and the year before? Anywhere near 68% and it’s something to look into.

Deflate your own headline. Adjust three years of revenue-per-donor for inflation. If the growth survives, well done!

Make the monthly ask the first ask, once. Lead one acquisition campaign this quarter with the recurring option instead of holding it back as an upgrade. Reversible, and it’s the one behavior the data says still compounds.

🧪 From the cohort

Our six cohort organizations - Income Movement, Cameron House, CAUSE, Generation180, Livestrong and Net Impact - are a few months into co-designing agentic giving with us. Two of the six questions on their list are first-gift conversion and dormant-list reactivation, both versions of this week’s story: whether the relationship can be rebuilt where the transaction usually ends.

If you want in on cohort two, the waitlist is open: Sign up here to be the first to know.

Reply with your top-ten gift concentration - what share of last year’s revenue came from your ten biggest gifts - and I’ll send you the donors vs dollars board template plus a one-page version of this inflation adjustment.

Hello@aifornonprofitsnetwork.org

📡 FUNDING RADAR

OpenAI Foundation, $100M for state AI implementation. Committed last month to the Breakthroughs to Follow-Through initiative with the Common Health Coalition, starting with hepatitis C treatment in Alabama, Illinois, Louisiana and Massachusetts. Local nonprofits act as the distributing intermediaries, the grants are deliberately not tied to any AI platform, and technical assistance from pilot to deployment is included.

Health Foundation for Western & Central New York, SEA Change Fund. $1.6M in grants plus $400,000 of capacity-building support, $100,000 each to 16 nonprofits, built as a rapid response to federal funding cuts.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

It is the right question. Most personalization in the sector is a merge field and a segment name. What follows is the actual mechanism we are running, and the challenges we’ll have to overcome.

On August 4 we said we’d go through this touchpoint by touchpoint. This is touchpoint one.

💭 What the engine runs on

Evidence, not house style. We keep a library of more than 200 studies on donor behavior, each one peer-reviewed or checked against a live test. A personalized email campaign will draw on about ten of them. That library does the deciding: when the system picks an ask amount, a send time, or whether to open with impact or with thanks, it’s applying a specific finding rather than a copywriter’s instinct. Findings that don’t replicate in a cohort test come back out of the library.

Enrichment, starting with your own systems. The agents pull a supporter’s CRM history, giving record, event attendance and email engagement (most fundraising platforms, CRMs and mail tools connect without much work), then add outside context you always had a right to and never had time to find: where someone studied, where they work now, what they’ve said publicly about the things they care about. Very little of this is new data. Most of it was already sitting in three systems that never spoke to each other.

Worth saying plainly: enrichment carries obligations. UK and EU GDPR for European supporters, state privacy law in the US, and a supporter’s reaction to being profiled matters as much as the legal position does. We’re using data with explicit consent.

Nonprofits are in an unusually strong position here. A retailer enriching a stranger’s profile is doing surveillance. You’re adding context about someone who already put their hand up for your cause, sometimes years ago, and who has spent every year since being sent mail that reads as though you’ve never met. We almost never hear a supporter complain that a charity knows too much about them. The complaint is the reverse. That’s a better starting point than any commercial sector has, which means the standard for doing this properly can be set here rather than imported later from retail. Every cohort organization is making the call deliberately, and where one of them decides the answer is no, we’ll publish that too.

🧹 What happens when the data is poor

Most organizations don’t have a clean file. They have a decade of one-time donors, no interests captured anywhere, and an unsubscribe rate nobody wants to bring up.

Between them, the six cohort organizations hold over 200,000 email contacts. A large share of those people have heard nothing from the organization holding their address since the day they last gave.

Enrichment is what splits that pile in two. On one cohort file we scrubbed and enriched the whole list down to roughly 10,000 contacts with real engagement in the past year, and rebuilt each of those briefs with employment, education and interest context. That’s the group worth running a personalized program against. The rest aren’t dead. They’re unasked.

And for the unasked, the move isn’t smarter inference. You just have to ask.

One cohort approach: a short, plainly written email to historical donors saying we’d like to know you better, with a two-minute quiz attached. Not a satisfaction survey. Net Impact’s version is themed around impact across the mission, and uses interests as the way in rather than a form to fill.

Questions about what they care about and how they’d like to hear from us. Nobody expects most people to reply. Everyone who does becomes a supporter brief that didn’t exist before, and the raw material for the next twelve months of contact.

Share

🤖 How a journey actually gets sent

The decision engine reads the supporter brief, applies the evidence library, generates the sequence, then behaves like an inbox: it either sends, or it parks the message for a human to approve. Timing is set per supporter rather than by the campaign calendar.

Nobody in the cohort is running this unattended, and we’d be suspicious of anyone who claimed they were.

👋 Come and see it running

We’re running a workshop on 23rd September, 11am (ET) on personalized campaigns and what it takes to unlock them inside your own CRM.

Reply to this email or send a note to hello@aifornonprofitsnetwork.org to sign up for free.

🎯 The personal edge

Each organization is testing this from a different starting position.

Income Movement has built one of the largest economic-security movements in the country. We’re working out what a first ask looks like when it has to be worthy of that trust: personal, respectful, and in the movement’s own voice.

Cameron House, a 150-year-old San Francisco community institution, is testing how warm supporters become first-time and then monthly givers.

CAUSE is building a scalable recurring giving program out of years of scattered supporter data, to fund the civic infrastructure that helps communities participate and lead.

Generation180 is turning one of the sector’s most engaged clean-energy newsletters into a digital giving program.

Livestrong has spent years giving people affected by cancer resources and guidebook downloads, and those downloads captured emails. The question is whether that audience can be re-engaged for support without trading on the moment they first arrived. Livestrong is also building on Ellis, its AI support companion.

Net Impact needs no introduction here: readers met CEO Karen Johns in our February “90 days” case study. Net Impact holds a large alumni file and is testing re-engagement now that those alumni are established in their careers and in a position to give back.

The first Cohort Diary, one of the six writing in their own words, runs next month.

📊 Does any of this work

Too early for the full answer, but the first signal is encouraging. One organization that started with us in May saw donations captured around one of its events rise 12% after adding personalized outreach either side of the event.

One number from one event isn’t a result, and running this in the open means we say so rather than putting it on a slide.

Here’s what we’re measuring across the 200,000 contacts and the campaigns and events behind them, and what we’ll publish over the next three months:

• One-time capture rates from events and digital touchpoints, before and after personalization

• Conversion from one-time gift to monthly, which is the number that decides whether any of this pays for itself

• Reactivation of contacts with no engagement since their last donation

• Where the agent’s output had to be overridden by a human, and why

Baselines are set. We’ll report the KPIs that moved and the ones that didn’t.

✈️ Come and find us in…

Chicago

We’ll be in Chicago from 1st through 5th September, presenting with our partners at Project Equity.

If your organization is in or near Chicago and you want to talk properly about any of this, the cohort, agentic giving, or the state of your own supporter data, reply to this email and we’ll find a time.

NYC Climate Week

Climate Week NYC runs September 20 to 27. We’re supporting Net Impact’s session, Impact Meets Intelligence: Careers in AI, on Monday September 21, 4 to 6pm ET in midtown Manhattan. Come and say hello.

Drop a note to hello@aifornonprofitsnetwork.org and we’ll grab a coffee/beer.

🚪 Join the second cohort

The waitlist for Cohort 2 is open: Join the waitlist now.

Same shape as the first. Real deployments, findings published here as they happen. If you’re sitting on a dormant list, a large pledge file, or an audience you’ve never asked, you’re exactly who we want to hear from.

Or just reply to this email with the state of your own file, and we’ll tell you honestly whether personalization is your next move or whether something else should come first.

Share

That question stops one step short; once the agent has the answer, it has to pay.

Subscribe now

💭 Thought for the week

On August 14, Aashis Luitel of the University of the Cumberlands published a piece in The Conversation about that step. His example is a small one: an agent told just to search for a shirt, buys one instead.

What happens next: The retailer logs a sale, the processor logs a charge, and the AI provider logs a session. Every one of those records is accurate, and as Luitel puts it, “nothing in those records links the charge to the task.”

There’s no evidence chain running from what the person asked for to what the agent did, and because the systems belong to different companies, nobody holds enough of the picture to settle the argument.

Washington has noticed and is putting in place things to be aware of. On July 21, Senator Mark Warner introduced the AI AGENT Act (S. 5051). It defines a “custodial user agent” as one authorized to act for a user in a transparent, documented, limited and revocable way, and it hands NIST the job of identifying protocols or writing standards for verifying that a user delegated authority in the first place. Google has published its Agent Payments Protocol, AP2, which covers some of the same ground by creating records of the limits a user approved and the information each party saw when a transaction is disputed.

Neither is finished. AP2 produces the evidence but doesn’t say who takes the loss, or how long anyone has to keep the records. And NIST’s current work is scoped to agents operating inside a single organization. The consumer agent that crosses company boundaries, which is the one your donor will actually use, has been pushed to later.

A gift is the easiest version of this problem

When a retailer can’t prove authorization, it has a dispute with a stranger over a shirt. When you can’t, you have a conversation with someone who already chose you.

That’s a far better place to be standing. The amounts are smaller, the relationship is longer, and unlike almost every other transaction on the internet, both sides want the record to exist. Your supporter wants their gift recognized properly and you want to know what they actually asked for. Nobody is trying to hide anything from anybody.

So the sector has a reason to reach this before commerce does rather than after. The standards are being written right now, in a bill and a protocol that are both still open, and nobody has asked what a donation looks like inside them. Nonprofit payment processors will move slowly, which means the useful move isn’t waiting for them. It’s keeping your own record of what a supporter’s agent asked for, in your own CRM, starting with the first agentic gift you receive. Something we’ll be testing in the cohort.

The part worth being cheerful about

Everything on Luitel’s list, a task identifier that travels between systems, stated limits, tamper-evident logs, adds up to one thing; a record of what a supporter actually asked for.

That’s the most useful consent and preference data our sector has ever had access to. Fortune ran a piece on Sunday on the $124 trillion wealth transfer running through 2048, quoting Bloomerang’s Steve Isom on what younger donors want. 94% say they’re motivated by an organization being specific about where the money goes, and specific impact language beat generic messaging by 88 percentage points. An agent can only pass on specificity you’ve already published. Write it down, build to receive a structured record back, and your next conversation starts with a segment of one instead of a guess.

That is more or less what the six organizations in our cohort spend the next five months on. When the first of these records starts arriving, we’ll publish what it actually looks like.

Three things you can do this week

1. Send your payment processor one email. What’s your roadmap for agent-initiated payments, and are you looking at AP2? If nothing, join our cohort two waitlist so you have something practical by the end of the year.

2. Decide how you’d handle a supporter who queries a gift their agent made. For most of you the answer is refund it and ask what happened, which is also the answer that keeps the supporter.

3. Look at what your site tells a machine. Program names, ask amounts, what a given amount does. That’s the material an agent quotes when someone asks it where to give.

🚀 Join the second cohort

The waitlist for Cohort 2 is here: Join the Cohort 2 waitlist

Same shape as the first. Five months, real deployments, findings published here as they happen, failures reported to learn from as well as the wins. If you’re sitting on a dormant list, a large pledge file, or an audience you’ve never asked, that’s the profile we’d like to hear from.

✈️ We’re in Chicago

We’ll be in Chicago from 1st through 5th September, presenting with our partners at Project Equity.

If your organization is in or near Chicago and you want to talk properly about any of this, the cohort, agentic giving, or the state of your own supporter data, reply to this email and we’ll find a time. We’d rather spend the week in conversations than in sessions.

🤝 Funding radar

Miller Center for Global Impact accelerator (Santa Clara University). A free accelerator for social enterprises working on women’s economic power or climate resilience. Early deadline August 31, final deadline October 16, for the June 2027 cohort. Details

JD Finish Line Foundation, Louder Than Words Cycle 3: Safe Communities. Up to $10,000 for nonprofits serving underserved communities nationwide. The portal opens September 1 and closes October 15. Details

We Raise Foundation Emerging Leader Grants. $15,000 over two years, split between $10,000 of program funding and $5,000 of professional development, for early-career leaders at Christian 501(c)(3) organizations. Deadline September 30. Details

Worth a look

The $124 trillion handover (Fortune, Aug 16). 75% of millennials say they plan to increase their giving year over year, and about 2% of households hold more than half the wealth in transit. The line worth stealing is Bloomerang’s: donors are ready to trust nonprofits, they just want to see the receipts.

Who’s building AI for good (EY). Women lead half of impact enterprises but only a quarter of AI-driven ones, and women-led startups take 2% of global venture funding. Read it next to your own vendor list.

An AI agent spent your money (The Conversation, Aug 14). The full piece behind this week’s lead, including the five components Luitel argues any accountable agent payment needs.

Reply to this email if you want to meet in Chicago, or if you have a view on any of the above. We read every one.

Subscribe now

That gap between attention and action is how non-profits are seeing their fundraising worlds change. And it’s happening across the sector; a pledge list of more than 100,000 names that hasn’t been contacted in years, a cause with more than a million signatories that have never been asked for a financial contribution or organizations with thousands of warm alumni, that haven’t been reengaged with the right ask at the right time.

That’s the challenge six organizations are setting out to understand, with our partners at Whitelabel. Using best practice behavioral science and AI for good, over the next 5 months, we’ll be publishing the findings here from key questions that include:

• How can we capture more potential individual giving leads from our digital and physical touchpoints?

• How can we personalize every communication to every potential supporter, with the right ask at the right time?

• How do we make sure our cause is the first answer when someone asks AI where to give?

• How do we grow giving on autopilot — through employer matching and government incentives?

• How can we embed these tools into Claude and ChatGPT where our sector is already working so they don’t have to learn a whole new software package?

• How do we learn from every supporter interaction, so each one unlocks new value and new revenue?

Who’s in the cohort?

Organizations spanning community services, clean energy, cancer support, and professional impact:

The first ask: Income Movement

Income Movement has built one of the largest economic-security movements in the country. Together we’ll be working out what a first ask looks like when it has to be worthy of that trust - personal, respectful, and in the movement’s own voice.

Also in the Cohort:

Cameron House - a 150-year-old San Francisco community institution, testing how warm supporters become first-time and monthly givers.

Center for Asian Americans United for Self-Empowerment (CAUSE) - partnering to build a scalable recurring giving program from years of scattered supporter data to fuel the civic infrastructure helping communities to belong, participate, lead, and shape public decisions.

Generation180 - turning one of the sector’s most engaged clean-energy newsletters into a digital giving program.

Livestrong - building on Ellis, its AI support companion for people affected by cancer, and shaping what comes next for supporter engagement.

Net Impact - readers met CEO Karen Johns in our February “90 days” case study; Net Impact now joins the cohort to test AI-supported engagement across its professional network.

What the cohort will test

Each organization is testing a different slice of the same agentic toolkit:

• Agentic donor journeys - welcome, onboarding, and reactivation sequences that reason about what each supporter should see next.

• Hyper-personalized asks - built on a supporter’s interests, not their identity.

• First-gift conversion and monthly upgrades - turning warm audiences into first-time and recurring givers.

• Dormant-list reactivation - waking the supporter data nonprofits already own.

• Match funding, matched by AI - surfacing matching-gift opportunities donors would otherwise never see.

Fuller profiles - We’ll publish more from each organization in its own words, in a new Cohort Diaries column each month.

What we’ll measure, and learn together

Every organization in the cohort is being baselined now on the measures that matter: engagement, conversion, recurring giving. In December we’ll publish the results against those baselines - what worked, what didn’t, and what we’d do differently. No numbers appear without the organization’s consent, and some will be anonymized at their request. The failures will be reported with the same care as the wins; a playbook built only from successes isn’t one.

Be part of the second cohort

Several organizations have already asked about a second wave. If yours should be in that conversation, send and email to hello@aifornonprofitsnetwork.org

🗞️ In Other News

Cohort models are spreading. UMass Amherst’s Ignition Initiative is running its own AI problem-solving cohort with Way Finders, the Food Bank of Western Massachusetts, and HILL for Literacy — more evidence that the sector learns AI better in groups than alone.

The EU AI Act’s transparency rules landed. A useful plain-English read on what the new disclosure obligations do and don’t solve - relevant to anyone with European supporters.

Worth a click

• Nonprofit Tech for Good’s running AI marketing & fundraising statistics page — handy for board decks.

• A CBC piece on Nova Scotia heritage orgs split over AI-made promotion — a small story that captures the whole sector debate in miniature.

Funding roundup

• Anthropic × Gates Foundation — $200M AI-for-good partnership. Announced in May; program details are still emerging, but it signals where big philanthropy is pointing.

• F5 STEM & AI education grants — 2026 guidelines are live for US education-focused orgs.

• Broader lists worth bookmarking: Top AI grants for nonprofits 2026 and this funding guide.

The AI for Non-Profits Network is supported by Whitelabel.ai, an AI native fundraising platform. Clients include Livestrong, the Canadian Red Cross, Net Impact, and Hunger Free Colorado. To read about how we’re helping a local newspaper archive content and reach more people, scroll to the end of this issue.

Donors down, dollars up — why healthy toplines are masking a structural break, and where the way out actually is.

We’ve all done it. A soft quarter, an amber dashboard, and the instinct fires before the meeting ends: add an appeal, send the extra email, squeeze one more ask into November. Volume is the lever that’s always within reach — and the sector data now says volume is precisely what’s failing. Three trends are compounding under all of us. If your income is flat, your file is ageing, and more email is going out for less coming back, you’re not mismanaging anything; you’re watching a structural break from the inside.

The base is shrinking. UK charitable participation is down to around 50% from 58% in 2019 — roughly four million fewer donors, the lowest since CAF’s tracker began — and public donations fell nearly 10% in a single year, 2024 to 2025 (CAF UK Giving; UK Fundraising). The US mirror: donor numbers down 4.5% in 2024 (FEP). Fewer people are giving — not less enthusiastically, fewer people.

The pipeline is thinner. US retention sits at 42.9%, a fifth straight decline; fewer than one in five first-time donors ever gives again, against 65–69% for repeat donors (FEP 2024). And the board-agenda number: small donors (under $100 — more than half of all donors) fell 8.8% while total dollars rose 3.5% (FEP Q4 2024). Big gifts are masking a collapsing base — a CFO would call that concentration risk and refuse it in any other revenue line. If your board hasn’t seen this picture, that’s not negligence; a rising topline is exactly the mechanism that hides it.

The channel is decaying. US nonprofits raised $58 per 1,000 fundraising emails in 2024, down 10%; email revenue for non-political organizations fell 13% (M+R Benchmarks). More email sent, less earned — and the marginal blast spends the one asset that compounds: your supporters’ willingness to open the next one.

The part that should make you optimistic. Look at the retention pairing again: under one in five first-timers return, but donors who make a second gift stay at 65–69%. The sector doesn’t have a generosity problem — it has a second-gift problem. And second gifts aren’t won by volume; they’re won by relevance — what supporters hear connecting to why they gave in the first place. That lever sits entirely within your control, with the supporters you already have. Until recently that kind of relevance was a big-org luxury; AI now makes segment-of-one communication feasible at small-org scale and cost. We think it’s the most consequential shift in fundraising practice in a decade, and we’ll spend coming issues showing how it works, touchpoint by touchpoint.

Four moves to start this month:

• Pull four numbers: three-year donor count trend (not revenue), overall retention, first-time retention, income share of your top ten gifts. Can’t get them in an afternoon? That’s finding number one.

• Add one slide to the board pack — donors and dollars, side by side. Name the masking effect before someone else discovers it.

• Freeze the volume ratchet for a quarter. Cap sends at last year’s level and watch what happens.

• Run the 100-supporter test: does what any 100 supporters hear from you bear any relationship to what you know about them? Where the answer is no, you’ve found your second-gift problem — and your biggest untapped asset.

Next week, we’ll be sharing our grant cohort, and how we’re working with them on this exact challenge above. Watch this space for live learnings and case studies.

Reply with your retention rate and I'll send you the one-slide 'donors vs dollars' board template so you can show your trustees the masking effect this month. Hello@aifornonprofitsnetwork.org.

Share

📡 FUNDING RADAR

Humanity AI — $10M open call: still not open. The 10-foundation coalition promised a summer launch; as of this week it hasn’t dropped. Prep against the four focus areas — democratic institutions, workers’ rights, journalism, education (prep guidance). We’ll flag it the week it opens.

NextLadder Ventures — $1B for AI serving frontline workers. Gates Foundation, Ballmer Group, Stand Together, Valhalla, and John Overdeck, with Anthropic as AI partner. No formal open call yet — but if your org supports case managers, benefits navigators, or social workers, get on their radar now: nextladder.com.

WORTH A LOOK

“Agentic giving” is now officially a thing. Fundraise Up — the donation platform behind The Salvation Army UK and the USO — coined the category last week: donors telling an AI assistant what they care about, and the assistant finding, comparing, and completing the gift. Zero usage data in the announcement — which tells its own story. We’re unpacking what’s real vs. category-making in a coming issue.

Jeff Brooks on the AI slop wave. Three minutes well spent from one of fundraising’s sharpest copywriters: “If you are bad at fundraising, AI won’t help you be better. It’ll probably make you worse. But faster.” A useful companion to this week’s story — volume was failing before AI made it free.

Funders are moving on AI-written grant applications. The NIH now caps investigators at six applications a year and won’t treat proposals “substantially developed by AI” as original ideas. Foundations are quieter but moving the same direction. What that means for your grant strategy — full story soon.

The AI for Non-Profits Network is supported by Whitelabel.ai — whose whole thesis is this week’s story: relevance over volume, personalised supporter journeys, AI-powered giving at small-org scale. If the 100-supporter test stings, that’s the conversation to have. Just reply to this email.

Subscribe now

The $500 million tell

The biggest coordinated bet philanthropy has ever made on AI is quietly showing its hand - and if you read the cards right, there’s money in it for you.

Quick recap if you missed it: last October, ten major foundations - MacArthur, Ford, Mellon, Packard, Omidyar Network, Mozilla, Doris Duke, Lumina, Kapor, and Siegel - launched Humanity AI, a pooled $500 million commitment over five years to keep AI “people-centered.” In May, they made their first moves: $18 million+, including twelve grants of $500K each.

Here’s why this matters to you even if you’ll never apply: first grants are a funder’s revealed preferences. Forget the press-release language - look at who got the checks. The pattern in the twelve:

• Community-led beats credentialed. The portfolio skews toward organizations of the communities affected by AI, not think tanks studying them from a distance. Their own framing: “the communities closest to AI’s impact often hold the clearest vision.”

• Specific beats general. Nobody got funded for “AI ethics.” They got funded for AI + journalism (Pulitzer Center), AI + civic education (Data & Society’s AI Civics initiative, $3M), AI + worker accountability (TechEquity).

• The five key pillars: democracy, education, humanities & culture, labor & economy, and security. If your mission touches one, you’re in the frame.

The actionable part: a $10 million open call is expected to launch this summer - a single pooled application, not ten separate ones, with 501(c)(3) status as the baseline. Full criteria haven’t dropped yet, which means right now is the positioning window. Three things worth doing this week, per the analysis at Granted: map your work to ONE of the focus areas with concrete examples (not broad ethics framing), document how your community actually shapes your decisions, and draft a tight 250-word theory of change so you’re not writing it the week the deadline lands.

And this isn’t a one-off. Coordinated pooled funds are becoming the model for AI philanthropy - OpenAI’s $50M People-First AI Fund ran the same playbook this spring, with community journalism as one of its three focus areas. The organizations that learn to read these signals early will be applying while everyone else is still forwarding the announcement.

What does it actually look like when one of these grants lands at a small nonprofit? Funny you should ask - keep reading.

Read our paper: Building in the Open

📡 Funding Radar

Deadlines and doors worth knowing about this week. All links go direct to the funder.

🟢 Opens Wednesday - Fast Forward Accelerator (2027 cohort) The best-known accelerator for tech nonprofits opens applications July 30 and closes September 8. $25K unrestricted grant, plus mentorship and a curriculum built specifically for nonprofit founders building tech products. You need a working MVP — early-stage is fine. If you’ve built an AI tool for your cause, this is your window. Details & application → · FAQ → We can support - get in touch - hello@aifornonprofitsnetwork.org

🟡 Closing soon — Third Wave Fund: Disability Frontlines Fund Deadline August 7. For organizations led by and serving disabled communities, particularly at the intersection of gender and racial justice. Via Purpose Possible’s funding calendar → We can support - get in touch - hello@aifornonprofitsnetwork.org

🔵 Rolling - Google Ad Grants Still the most under-used free money in the sector: $10,000/month in in-kind Google Search advertising for eligible 501(c)(3)s, rolling applications. If you’re not using it, that’s $120K/year of donor and program visibility on the table. Overview → We can support - get in touch - hello@aifornonprofitsnetwork.org

Case study: 35 years of local news, unlocked

This is the part of the newsletter where I talk about Whitelabel’s own work - this one also links to the main article.

Since 1990, the Two River Times has covered Red Bank and the towns along the Navesink and Shrewsbury Rivers in Monmouth County, New Jersey - local news, school sports, community events, the stories no one else covers. A few years ago the paper went nonprofit, and a growing reader membership is what makes the model work.

We’re now building with them, supported by OpenAI’s People-First AI Fund. Longtime readers will remember our special issue on this fund back in December, when the first $40.5M went to 208 organizations and I argued the money was “an accelerant, not a solution” - that the real barrier isn’t access to technology but the organizational capacity to absorb it. Seven months later we’re on the inside of one of these projects, and I can report: that’s exactly what the work looks like. Two things are being built:

Behind the newsroom: the paper’s 35-year archive, opened up so a reporter can surface the history behind a story in seconds instead of an afternoon of searching. The reporting is theirs and stays theirs - AI is the index, not the author.

In front of readers: three and a half decades of local reporting is a record of the community itself. Members will be able to pull on that thread on demand - the coverage of their street, their school, their business, the events they were part of - personalized to their own history in the town and delivered alongside the print edition they already value. Mission and membership revenue, from an asset the paper already owned.

Here’s the question to steal from this: what’s your organization’s archive? Decades of case files, program records, photos, annual reports, event histories - most nonprofits are sitting on a version of this. It’s usually treated as storage. Funders are increasingly treating it as raw material for exactly the kind of community-rooted AI project the pooled funds want to back.

If that sparked something, hit reply with the word “GRANT” and one line about what you’re sitting on and what you’d build - whether it’s for the Humanity AI call or anything in the Radar above. I’ll tell you honestly whether it’s a fit for our Grant Build Partner engagement - including when the answer is “you don’t need us for this one.”

One more stat before you go

From the Blackbaud Institute’s new Bridging the AI Effectiveness Gap report: 76% of donors want to know when and how your organization uses AI - but only 26% of organizations tell them. A three-paragraph AI transparency statement on your website is the cheapest trust win available right now. (More on this in a future issue.)

And whether or not you ever hit reply: if the Humanity AI call dropped tomorrow, would you know what you’d pitch? That’s the question worth taking into the week. I read every response to this newsletter - the strongest ideas (anonymized) will show up in a future issue.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

• 📝 Special edition: Inside the cohort - what AI-grant winners talk about when the funder isn’t in the room

• 📄 Building in the Open is now public - read the whole paper.

• ⭐ Funding Radar - a free forum today at 2 PM ET, and where the closed rounds go from here

• 🌐 From Across the Network - 200 new parents wanted

Subscribe now

This newsletter is supported by Whitelabel.ai, who also convene the grantee cohort behind this week’s dispatch — and now host the paper it produced.

📝 Inside the cohort: what a year of AI money actually looks like

Last Wednesday, a dozen-plus of the OpenAI Foundation’s first-round grantees met for their third peer-learning session - community centers, youth organizations, a philanthropic intermediary, a workforce agency, a women’s building. No funder in the room, no vendor pitch to sit through. Just organizations a year into spending AI money, comparing notes on what actually happened.

Read the Paper Here

We’ve been convening these sessions since the grants landed, and the paper drawn from them - Building in the Open: The hard part was never the technology - is now public in full. Sixteen organizations, three recorded sessions, six findings, published as a living document that the next cohort will be invited to test and revise. Read it; it’s honest in ways sector reports rarely are. Here is what those findings look like in the room.

The first build is usually a misunderstanding. The session’s most repeated theme wasn’t technical. City Heights CDC in San Diego is building a community AI lab in a neighborhood at high risk of being left out entirely - and started not with tools but with the question “what even is AI?”, aligned to the Department of Labor’s AI-literacy framework. They’re right to start there: Pew’s new national survey finds that even now, fewer than half of US adults have used ChatGPT. The people in this work are years ahead of the people it serves. Comprehension, not capability, is still the gap.

Every project hits the same wall, and the wall is trust. Springboard Opportunities works alongside mothers in affordable-housing communities in Jackson, Mississippi. Their focus groups surfaced both real use and real dependency on AI tools - so before building anything, they’re training families on what these systems are, what they collect, and how to avoid being exploited by them. In Boston’s Dorchester neighborhood, the same instinct: seniors are being taught AI scams first, tools second. Different missions, identical question - what does this technology do to the people we exist to protect?

The tools are dissolving. The Women’s Building in San Francisco had graduate board fellows build them a bespoke fundraising tool that maps each board member’s network against funding categories - the kind of thing that was an enterprise software purchase two years ago, now assembled in a semester and presented at the next board meeting. The Forbes Funds in Pittsburgh has gone further: several hundred AI guides for local nonprofits, open source, claimable by each organization. Nobody in the session talked about buying platforms. Everybody talked about people close to the mission building close to the work.

“Time saved” is still the wrong scoreboard. The most quietly instructive story of the session: one organization wrote its original grant application around a client-navigation tool, then spent the first six months on none of that - staff training, board education, trust-building, partnerships. Their round-two application says so, plainly: here’s what we proposed, here’s how we pivoted, here’s why. That honesty-as-strategy was echoed around the room, including by a funder: the strongest reapplications name the gaps and failures, not just the wins.

Co-design isn’t a nicety - it’s the mechanism. Public Health Advocates has 28 young people in California’s Central Valley working as researchers on one question: how AI can strengthen rather than undermine youth mental health - with the youth leading focus groups and drafting school-board policy recommendations themselves, and a symposium in December. It Gets Better is rebuilding its mental-health navigation tool with its youth ambassadors at the table (and with Whitelabel as build partner - our disclosure as always). Across the session, the organizations making real progress were the ones whose communities are designing, not receiving.

And the hardest one: being the “good actor” is a claim to be tested. One participant described the cost of having pushed AI trainings early in a community that trusted them - visible damage to the brand, followed, slowly, by people circling back. The paper’s version of this finding is blunt: an ethical posture cannot be inherited from your vendor. It has to be earned in public, which is the entire argument for building in the open.

Lessons for nonprofit leaders:

1. Start with the comprehension gap, not the tool. More than half the people you serve have never touched this technology. Your board may not be far ahead of them.

2. Put protection before productivity for vulnerable communities. Scam literacy, data-privacy plain talk, dependency awareness — the cohort treats these as prerequisites, not add-ons.

3. If you pivoted, say so - especially to funders. The reapplications the room rated strongest were the honest ones.

4. Read the paper and steal from it. It’s public and written by peers — six findings, the shared problems behind them, and what the network is building next.

Share

⭐ Funding Radar

• OpenAI People-First AI Fund — Round 2 closed July 15. If you applied: based on last year’s cycle, expect decisions around September. When the 2026 grantee list lands, the peer-learning cohort behind this week’s dispatch will open to new winners — if you’re selected, reply to this email and we’ll bring you in.

• AWS Imagine Grant — Round Two opens August 10, invitation-only. Everyone else: the alert list is the way to not miss the 2027 window.

🌐 From Across the Network

An ask from inside the cohort: 200 new parents wanted. Bonded by Baby — the postpartum mental-health companion app being built with Social Creatures under the People-First Fund (Whitelabel is the build partner) — is beta-ready, in the App Store, and already running with 40 parents through a Mount Sinai program. The team is looking to bring roughly 200 more new parents into testing over the next two to three months. If your organization works with new or expecting parents and you’d like your community involved, reply to this email and we’ll connect you.

We’re still collecting answers to the question from two weeks ago — what’s the one process you’d hand to an AI tomorrow? The patterns are starting to form; keep them coming.

Have an event, case study, gathering, or insight to share with the network? Reply to this email (hello@aifornonprofitsnetwork.org) and we’ll work it into the next issue.

We’re a network of nonprofit leaders interested in how AI is impacting our sector and our mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

• 👀 What caught our eye: A new, free cybersecurity clinic for under-resourced nonprofits — and why it exists now

• 💭 Thought for the week: Uber put its AI people next to the work. The lesson isn’t about Uber.

• ⭐ Funding Radar — one deadline tomorrow, one on Friday, and a correction

• 🌐 From Across the Network

Subscribe now

This newsletter is supported by Whitelabel.ai — helping nonprofits cut through AI noise with practical tools built for mission-driven teams.

1) 👀 What Caught Our Eye: free cybersecurity help for nonprofits, because the threat just got automated

NYU has opened a cybersecurity clinic that will work with nonprofits for free — and the reason it exists is the same technology this newsletter covers every week.

Announced July 7 and housed at the NYU Center for Cybersecurity with support from Craig Newmark Philanthropies, the clinic pairs law and engineering students with organizations that can’t afford serious IT security — training staff, mapping systems for vulnerabilities, and helping draft the governance protocols most small organizations never get around to. Its co-director, Judith Germano, puts the problem plainly: organizations are “still needing to get their basic cybersecurity right. And now the stakes have gone up.”

The stakes have gone up because AI has collapsed the cost of the attack. Deepfake video and voice are now routine instruments of financial fraud, and the same models that can find vulnerabilities deep inside systems for defenders can do it for attackers. The economics that once protected small organizations — too minor to be worth a criminal’s time — no longer hold when the targeting is automated.

Nonprofits are an unusually soft target: trusted brand names, donor payment relationships, sensitive beneficiary data, and almost no security budget. A convincing voice clone of your executive director asking finance to change a vendor’s bank details is no longer science fiction; it is this year’s fraud pattern.

What to do this week. Two concrete things. First, set a verification rule with your finance team today: any request to change payment details or move money — however senior the voice on the phone — gets confirmed through a second, known channel before it happens. Write it down; a rule that lives in one person’s head isn’t a rule. Second, if you have no IT security capacity at all, contact the NYU clinic — this kind of free, hands-on help is rare, and the queue will form quickly.

Share

2) 💭 Thought for the Week: Uber put its AI people next to the work. The lesson isn’t about Uber.

Uber’s CTO did something unfashionable this spring. Instead of buying another platform or announcing another copilot, he took roughly thirty of the company’s most AI-fluent engineers and embedded them, two weeks at a time, inside HR, finance, and legal — next to the people who actually do the work (Fortune).

The sprint design is almost comically humble for a company of Uber’s size. Days one and two: shadow the domain expert. Day three: pick what’s worth automating. Days four and five: build alongside the worker. Then peer validation, then ship on day ten. Sixteen of these “pods” ran in the first two months. The results were not humble: financial pacing reports that took two days now take ten minutes; capital-allocation planning across 150 cities went from fifteen hours to thirty minutes.

Two things about this deserve a nonprofit leader’s attention, and neither is the technology.

The first is where the gains came from. Uber’s CTO is blunt that you cannot automate from process documents: “You have to understand how the work actually gets done.” The pods found their wins in the handoffs, the approval steps, the workarounds nobody had written down — things visible only when you sit beside the person doing the job. The unit of automation was never the task. It was the workflow, and workflows only reveal themselves in person.

The second is what Uber did with the win. It didn’t cut the teams that got faster. It is building a dedicated team to scale the pod model — the productivity gain created a new investment in people. That echoes the pattern we flagged from Gartner’s spend data two weeks ago: the organizations furthest along with AI are raising their spend on people at the same time. The proof point keeps arriving from different directions.

Now the honest counter-evidence, because there’s plenty. Researchers have started calling the opposite pattern “AI brain fry”: a BCG study of nearly 1,500 US workers found self-reported productivity falling once people juggled four or more AI tools, with heavy AI oversight adding 14% more mental effort and 19% more information overload — and workers experiencing it markedly more likely to want to quit (Fortune). David Brooks argued in The Atlantic that when intelligence becomes plentiful, volition becomes the scarce asset — and that many workers will use AI simply to think less. Even Uber’s own leadership admits the AI budget is under real scrutiny. The tools do not produce the Uber result by default. Deployed as a raw efficiency squeeze on unexamined workflows, they produce exhaustion with a dashboard.

Which brings this back to the efficiency plateau we covered in March: 92% of nonprofits using AI, 7% seeing mission-level change. The missing ingredient the pod model supplies is not compute — it’s proximity. Someone fluent with the machine sitting next to someone fluent with the mission, long enough to see how the work actually happens. The sector-wide numbers cited in the same discussion are stark: most organizations claim to be “acting on” AI agents while only a sliver of their staff use one, and fewer than a third of workers at agent-using organizations have had any training at all. The binding constraint on AI value is now human capability. That is a budget line, not a download.

You cannot spare thirty engineers. You don’t need to. The nonprofit version of a pod is one AI-fluent person — a staffer, a volunteer, a borrowed board member’s analyst, a funded fellow — spending two honest weeks beside your caseworker or your grants manager before anyone builds anything. Anthropic’s Claude Corps, which closes its first fellow application window this Friday, is essentially this model with someone else paying the salary (details below). The episode of The AI Daily Brief that pulled these threads together is worth 22 minutes of your commute.

Lessons for nonprofit leaders:

1. Buy proximity before you buy tools. Two weeks of shadowing beats two months of vendor demos. The wins live in workflows nobody has documented.

2. Decide in advance what the dividend funds. If AI frees 200 hours, name where they go — training, relationships, a new program — or efficiency will quietly become the mission.

3. Budget for capability, not just licenses. If your AI line has no training spend in it, you’ve bought the plateau.

4. Watch for brain fry. If AI adoption is inflating email volume and shrinking focused work, you’re running the squeeze, not the pod.

Share

3) ⭐ Funding Radar

• OpenAI Foundation — 2026 People-First AI Fund closes TOMORROW, July 15 at 11:59 PM PT. $50M in unrestricted grants for US 501(c)(3) standalones with $500K–$10M budgets, priority on the $1–8M band. Five questions, 250–300 words each — a final day is enough if you know your community cold. Fund announcement · application portal. If you’re in the last-day scramble and want a second pair of eyes on your answers, reply to this email today — we’ll turn it around fast.

• Anthropic — Claude Corps fellow applications close Friday, July 17 for the October cohort. A year of a trained, paid, AI-fluent person embedded in your organization — the pod model from this week’s essay, funded by the lab. Host organizations also receive $10,000 grants and Claude credits, and host applications remain open across all three cohorts (Chronicle of Philanthropy) · host portal.

• Free online forum — “Work Smarter: How Nonprofits Are Innovating with AI,” Tuesday, July 21, 2 PM ET, hosted by the Chronicle of Philanthropy. Practitioners on how they vet tools, weigh environmental costs, and keep the work human-centered. All registrants get the recording. Register here.

• AWS Imagine Grant — a correction to our earlier bullets: Round One of the 2026 cycle closed June 5, and Round Two (opening August 10) is invitation-only. If you missed it, sign up for the alert list now so the 2027 window doesn’t pass you by — the Pathfinder award runs to $200K plus $100K in credits. Program page.

• Also worth knowing: Illinois became the third state to pass frontier-AI safety legislation on July 6 — mandatory third-party audits for the largest model developers, effective 2028 (WTTW). With California and New York, states covering an estimated 40% of the US AI market are now setting the rules your vendors will build to, whatever happens in Washington.

4) 🌐 From Across the Network

Last week we asked what one process you’d hand to an AI tomorrow. The question clearly hit a nerve — several of you told us in person it’s harder than it looks, which is rather the point. It stands for another week: reply with yours, and we’ll pull the patterns together once we have a critical mass worth sharing.

Our peer-learning paper with the Phase 1 cohort, Building in the Open: The hard part was never the technology, continues to circulate. If you’d like a copy, reply and we’ll send it over.

Have an event, case study, gathering, or insight to share with the network? Reply to this email (hello@aifornonprofitsnetwork.org) and we’ll work it into the next issue.

We’re a network of nonprofit leaders interested in how AI is impacting our sector and our mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

• 📝 Special edition: The Segment of One - a dispatch from a room full of charity finance leaders

• ⭐ Interesting News and Funding Calls - two deadlines inside two weeks

• 🌐 From Across the Network

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through AI noise with practical tools built for mission-driven teams.

Subscribe now

📝 The Segment of One

A dispatch from this morning’s Charity Finance Group roundtable.

I know the look a room gives when the “AI person” is introduced. Shoulders tense. Here we go - more hype about a technology that will fix everything and cost nothing.

So let me tell you the most useful thing said all day. It wasn’t mine, and it wasn’t about technology at all.

It was this: AI is a strategic question, not a technology one. The organizations getting it wrong bought a clever tool and went looking for a problem to point it at. The ones getting it right started with a problem worth solving and a number they wanted to move.

Which is a relief, honestly, because it means you don’t need to understand the technology to make a good decision about it. You need to understand your own organization.

Here’s how I’ve come to think about where AI actually earns its place in a charity. Three buckets.

1. Give your people their time back

The least glamorous bucket, and probably the most valuable. One of the charities we work with runs a grant for people going through medical treatment. Every application went out to a third-party eligibility checker, at real cost, hundreds of times a month.

We replaced it with a smart form that works out eligibility as the applicant goes, and nudges people who stall - you might still qualify; go check your insurance and come back. It paid for itself within weeks of going live. More to the point, it gave staff their time back for the work only humans should be doing.

There’s a hopeful footnote here, and it’s the one we led with in last week’s issue: Gartner’s 2026 CMO Spend Survey found that the organizations furthest along with AI are raising their spend on people at the same time — labor’s share of budget rose from 21.9% to 24.5% year over year, and a shortage of skilled people, not technology, was the most-cited barrier. (The usual caveat: that’s corporate marketing data, not nonprofit data — but the pattern is what travels.) This isn’t the redundancy story we were all braced for. It’s a capacity story.

2. Stop segmenting. Start talking to one person.

Here’s the idea I can’t stop thinking about. For decades we’ve segmented donors - bucketed them by age, gift size, zip code - because that was the best we could do. We couldn’t talk to everyone individually, so we talked to groups.

That constraint is gone. AI lets you have a genuinely relevant, one-to-one conversation at the scale of two thousand donors, or two hundred thousand. I’ve started calling it the segment of one.

Concretely: instead of blasting the same monthly email to the whole list, organizations can surface wildfire updates to the supporter who gave to wildfires, and something else to the supporter who cares about something else. Same newsletter. Different, relevant content. One social-impact charity we work with saw a meaningful double-digit uplift in post-event giving simply by following up with a message tuned to what each attendee actually engaged with - instead of a generic “thanks for coming.”

But let me be honest: this is not a silver bullet. We are all drowning in email, and the answer is emphatically not to send more. The whole point of the segment of one is relevance, not frequency. Done badly, personalization is just noise with your name on it. We’re keen to explore this concept more with non-profit orgs, so drop us an email on hello@aifornonprofitsnetwork.org to partner up.

3. Turn what you already do into something new

The bucket that surprises people. When one org made 30 years of audience specific resources genuinely accessible through AI, they didn’t just help their audience - they ended up holding one of the richest pictures of what that audience actually need. That opened partnership and revenue conversations with other partners nationwide. A value exchange that didn’t exist the year before.

What I’d actually tell you to do

Not this: write a sweeping AI strategy the size of a phone book.

This: pick one painful, measurable process, something with a clear cost and a clear number. Run a small, low-risk experiment. See if the metric moves. If it does, do it again somewhere else.

Two things to hold onto while you do. First, it’s almost always a data problem before it’s an AI problem - most organizations don’t lack tools, they lack usable data, and that’s the real work. Second, do it with consent. The segment of one runs entirely on data people knowingly gave you. Creepy is using data someone didn’t know you had. Relevant is using what they told you. Stay firmly on the right side of that line and you keep the one thing a charity can’t afford to lose: trust.

The charities pulling ahead right now aren’t the ones spending the most, or the ones with the biggest tech teams. They’re the ones who picked one thing and started.

So - what’s the one process you’d hand to an AI tomorrow if you could? Hit reply and tell me. I read them all.

— Jak

Share

3) ⭐ Interesting News and Funding Calls

• OpenAI Foundation — 2026 People-First AI Fund closes July 15 at 11:59 PM PT — now under two weeks out. $50M in unrestricted grants for US 501(c)(3) standalones with $500K–$10M budgets, priority on the $1–8M band. If you’re a fit and haven’t started, this is the week - let us know, we can help!

• Free forum — “Work Smarter: How Nonprofits Are Innovating with AI,” Tuesday, July 21, 2 p.m. EDT — two weeks out. Nonprofit and foundation staff on how they vet tools, weigh environmental costs, and keep the work human-centered. All registrants get the recording and transcript. Register free.

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

• 👀 What caught our eye: Gartner’s new spend data undercuts the “AI means fewer staff” story

• 💭 Thought for the week: AI can buy a small team time - but only people can do what the sector is actually for

• ⭐ Interesting News and Funding Calls

• 🌐 From Across the Network

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through AI noise with practical tools built for mission-driven teams.

From the network: Hunger Free Colorado is going live with Whitelabel, using its operational AI layer to manage distribution and SNAP access across an ecosystem of 1500 smaller food banks. Case study coming soon to network readers or find out more at whitelabel.ai

Subscribe now

1) 👀 What Caught Our Eye: the more AI-mature an organization is, the less likely it is to be cutting people

The story we’re always told is that AI is a substitution play: same output, fewer staff. New spend data points the other way - the organizations furthest along with AI are spending more on it and pointing it at growth, not headcount.

Gartner’s 2026 CMO Spend Survey, fielded January to March among 401 marketing leaders across North America, the UK and Europe, found that organizations with mature AI capabilities allocate 21.3% of their budget to AI, against a survey average of 15.3% - and they direct a far larger share of spend toward innovation rather than efficiency or cost control: 34.2% versus 27.2% for the average org. In other words, the further along you are, the more you’re using AI to do new things, not to do the same things cheaper.

The same survey found labor’s share of budget rising - from 21.9% in 2025 to 24.5% in 2026 - and named a shortage of skilled people, not a shortage of technology, as the single most-cited barrier to getting value from AI. Maturity and investing in people are moving together, not in opposite directions.

One caveat: this is a survey of corporate marketing departments, most at companies above $1bn in revenue. It is not a study of nonprofits, and the budgets bear no resemblance to ours. But the pattern is revealing - the orgs that get good at this reach for more ambition, not fewer staff.

Why it matters for nonprofits. If there is ever a sector with more to do than hands to do it, it’s ours. A maturity curve that leads to “what could we finally take on?” rather than “who can we let go?” is the version of AI adoption worth fighting for - and now there’s data underneath it.

What to do this week. Pull up your own AI spend, however small, and ask which column it’s in: efficiency or innovation. If everything you’re doing with AI is about trimming cost, you’re running the immature playbook. Name one thing your mission can’t currently reach, and point a small experiment at that instead.

Share

2) 💭 Thought for the Week: AI can buy the time - but only people can do what the sector is actually for

This week, three foundation presidents - of McKnight, Freedom Together, and MacArthur - wrote about what a year of funding “everyday acts of courage” taught them. The scenes are worth thinking about: a Minneapolis pastor organizing thousands of volunteers to bring groceries to immigrants too afraid to leave home; an Oregon wildfire survivor running weekly Peace Meals; Buddhist monks finishing a 2,300-mile walk for peace. Their Courage Project made small, flexible grants — $10,000 to $50,000, $2.5 million across all 50 states - not to institutions but to neighbors. Their conclusion: the most effective civic leadership is coming from ordinary people doing slow, relational, deeply human work.

It is worth reading that in the same week this newsletter spends a section urging you to get good at AI, because the two stories pull in opposite directions. AI is sold as scale: more reach, more output, more people processed per hour. The Courage Project is about the opposite - proximity, presence, showing up. And the thing it names as the engine of civic repair, trust, is precisely what does not scale and cannot be automated.

With institutional trust at historic lows - just 22% of Americans say they trust the federal government to do the right thing - the rarest asset in this sector is the relationship between a person in need and a person they believe. That asset is built in hours that look, on an efficiency dashboard, like waste: the home visit that didn’t strictly need to happen, the volunteer who stayed late to listen. AI is very good at finding and eliminating exactly that kind of “waste.” Point it at the wrong target and you optimize the soul out of the work - an organization that processes more people and reaches fewer.

This is the risk of an AI-mature sector I don’t hear discussed enough: not that the tools fail, but that they succeed at the wrong things. Funders reward what’s measurable and scalable; AI makes the measurable cheaper to chase. The relational, hard-to-count work is the first thing a pure scale logic defunds, because it photographs badly in an impact report. The Courage Project stands out precisely because it funds the invisible.

The counterargument isn’t empty: relational reach can be extended by good technology - a well-built tool helps a tiny team answer more people, faster, after hours, in their own language.

The Thing that caught our eye this week was data that said mature organizations use AI to do more, not to cut; this reading tells us what the “more” should be. Let AI take the grant report, the donor spreadsheet, the duplicate data entry - drudgery that was never anyone’s reason for joining this sector - so the humans are freed to do the proximate, un-scalable work that rebuilds trust. Use the machine to protect the relationship, not replace it.

Lessons for nonprofit leaders:

1. Sort every AI use case into two piles: “frees a human for relationship” or “replaces a human relationship.” Lean into the first; move slowly on the second.

2. Protect the unmeasurable. Before AI optimizes a workflow, ask what relational value the “inefficient” version was quietly producing.

3. Watch the funder incentive. If your reporting rewards only what scales, you’ll quietly defund your own proximity. Name that risk to your board.

4. Spend the reclaimed hour on people, not throughput. The time AI buys back is the point — what you do with it is the mission.

Share

3) ⭐ Interesting News and Funding Calls

• OpenAI Foundation — 2026 People-First AI Fund closes July 15 at 11:59 PM PT — now under two weeks out. $50M in unrestricted grants for US 501(c)(3) standalones with $500K–$10M budgets, priority on the $1–8M band. If you’re a fit and haven’t started, this is the week - let us know, we can help!

• Free forum — “Work Smarter: How Nonprofits Are Innovating with AI,” Tuesday, July 21, 2 p.m. EDT. A one-hour online session with nonprofit and foundation staff on the front lines of AI adoption — how they vet tools, weigh environmental costs, and keep the work human-centered. Covers grant-winning, donor-data management, workload, and even AI avatars as fundraisers. Useful whether you’re just exploring or already experimenting; all registrants get the recording and transcript. Register free.

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

• 👀 What caught our eye: A new report on the workplace strain AI is adding — and the judgment loop that nonprofits can least afford

• 💭 Thought for the week: Big Law just automated the bottom of its pyramid. Nonprofits are about to do the same thing, by accident

• ⭐ Interesting News and Funding Calls

• 🌐 From Across the Network

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai

Subscribe now

1) 👀 What Caught Our Eye: the AI rollout has a bill, and it’s landing on your team’s state of mind

A new report names a cost of the AI rollout that almost nobody puts in the business case - not budget, but strain.

Headspace’s 8th annual Workforce State of Mind report, based on a YouGov survey of more than 700 US workers and managers fielded in April, finds that AI is now the most common workplace change - 70% say their organization adopted new AI tools in the past year - and a leading driver of what it calls “chronic strain,” the low-grade overload that never fully switches off. 44% say that strain has hurt their ability to apply good judgment when using AI and new technology. Over a third say the strain has risen in the past year, half describe themselves as emotionally checked out, and 58% have had no resilience or change-management training to cope with any of it.

Why it matters for nonprofits. We are the hardest case in this data - thinner-staffed, deficit-running, mission-driven teams already prone to burnout, now adopting AI on the same curve as everyone else with less slack to absorb it. The loop is the part to sit with: the tool meant to lighten the load becomes another source of it, and the resulting strain erodes the judgment that using AI well actually requires. Capability was never the constraint. The people are.

What to do this week. Ask the team not only “will this new tool save time?” but “what will you stop doing to make room for it?” Adopting AI without subtracting work is how you manufacture strain.

Share

2) 💭 Thought for the Week: who trains the next generation when AI eats the entry-level work?

Big Law firms spent the last year automating the bottom of its own pyramid - and in doing so ran an experiment the nonprofit sector is about to repeat without meaning to.

In a single year, generative-AI use among legal professionals more than doubled, from 31% to 69%. It absorbed the base of the pyramid first: first-pass document review, legal research, contract markup - the high-volume, low-judgment work that, for a century, was how a first-year associate learned to be a lawyer. Yet 54% of firms still offer no AI training at all.

The uncomfortable part is not that machines do junior work. It is that the junior work was never only about output - it was the apprenticeship. Strip it out and you save a salary line this year and lose your training pipeline over a decade. Firms noticed late, and are now improvising replacements, rotating new lawyers through “AI compliance” roles, because they automated the classroom before building a new one.

Nonprofits run the same hidden pyramid. The development assistant learns your donors by drafting hundreds of thank-yous; the coordinator learns the community by cleaning the intake data; the comms associate learns the mission by writing the newsletter before anyone reads it. Those are precisely the tasks today’s models are cheapest at. With many organizations in deficit, the rational move when the assistant leaves is not to rehire but to hand the work to the model. Each decision makes sense alone. The aggregate is a sector that, in ten years, has no senior people who came up through the work - because no one did.

This is not an argument against the tools. The organizations moving fastest aren’t avoiding AI; they’re pairing it with structure - governance-first deployments pushed roughly 12 times more projects into production than the move-fast crowd (Stanford Digital Economy Lab). And Claude Corps, which we covered last week, is a $150M bet on putting early-career people and AI together inside nonprofits - a new apprenticeship for a world automating the old one. Whether it works is open. That it’s the right question is not.

So the choice every leader is making this year - which work to hand the model - is also a choice about who learns your organization from the inside, and who runs it in 2036. Make it on purpose, not one unfilled vacancy at a time.

Lessons for nonprofit leaders:

1. Make the two lists. The entry-level tasks you’d most like to automate, and the tasks people learn your organization through. The overlap is where to slow down.

2. Redesign the junior role; don’t delete it. Supervising and editing AI output still trains someone - but only if the job still exists.

3. If you cut entry-level hiring, fund the replacement. Mentoring, rotations, time to learn the mission directly. The old apprenticeship was free and invisible; its replacement won’t be.

4. Ask your board one question: in five years, where do our senior people come from?

Share

3) ⭐ Interesting News and Funding Calls

• ChatGPT slipped below 50% of the AI assistant market for the first time - 46.4%, with Gemini at 27.7% and Claude at 10.3%, per Sensor Tower's 2026 State of AI report. Worth a note for any org standardizing its workflows on a single provider: the leader you onboarded your team onto in 2024 isn't guaranteed to be the leader in 2027. Portability is a strategy, not a nicety. (roundup)

• Anthropic - Claude Corps host applications. Open across all three cohort start dates (Oct 2026, Jan 2027, Aug 2027) via the host portal. Fellow applications close July 17 for the October cohort.

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

In The Briefing this week:

• 👀 Claude Corps: Anthropic's $150M, 1,000-fellow bet that an AI lab can be a labor program

• 💭 Special Issue: The 2026 OpenAI Fund is open. Here is what the Phase 1 cohort has learned - and how the network can help with your bid.

• ⭐ Interesting News and Funding Calls

• 🌐 From Across the Network

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai

Subscribe now

1) 👀 What Caught Our Eye: Anthropic launches Claude Corps

On June 11, Anthropic launched Claude Corps: a year-long, in-person fellowship that will place 1,000 early-career hires inside at least 400 US nonprofits over the next two years. The headline number is $150 million.

The structure is a three-way partnership. Anthropic funds the program and provides Claude expertise. CodePath, Anthropic’s existing nonprofit partner and the largest US provider of collegiate computer-science education, acts as the fellows’ employer of record. Social Finance, a nonprofit registered investment advisor, handles measurement and evaluation and is building a longer-term financial vehicle to scale the program. Fellows earn $85,000 plus benefits, get five hours of weekly training on Claude, and spend the rest of the year embedded with a host. Anyone over 18 with under two years of full-time work experience is eligible; no degree required. The first cohort of 100 starts in October; the next two begin January 2027 and August 2027. Fellow applications close July 17 for the October start.

The host list is broad and named. Goodwill Industries International, Code for America, the IRC, RAINN, the YMCA of Greater Charlotte and Greater Cincinnati, the Montgomery County Food Bank, StriveTogether, Year Up United, Braven, Code the Dream, Team Red White & Blue, Heartland Forward, REEF, SoundOff, MyFriendBen, and Pacific Community Ventures are all in the first wave. Anthropic announced Claude Corps alongside its new Policy on the AI Exponential framework - the program is its own labor-disruption response, in public.

Why it matters for nonprofits. The pitch to host organizations is unusually plain: a fully funded, fully trained year of technical capacity, embedded in your org, paid for by the lab making the tools. The cost is what you would expect: a fellow who needs management, a workflow you did not have before, and a year of co-dependence on one vendor’s stack. Both sides of that ledger are real.

What to do this week. If you would host a fellow, the host portal is open now and rolling across all three cohorts - apply early, not late. If you would like a sober second opinion on what a fellow could realistically own inside your organization before you write the host application, reply to this email - we are happy to think it through with you so you can make the strongest application possible.

Share

2) 💭 Special Issue: The 2026 OpenAI round is open, and here is where the network can help

On June 15, the OpenAI Foundation opened the 2026 People-First AI Fund. The commitment is $50 million in unrestricted grants. The window is one month: applications close at 11:59 PM PT on July 15. For most readers of this newsletter, this is the round to apply to. We want to put the round in front of you, put six months of cohort evidence next to it, and be plain about what we can do to help.

The 2026 Fund is sharper than the 2025 one. Three focus areas now named: community service-navigation (legal aid, public benefits, disability and accessibility); community arts and cultural organizations; community journalism and media. Eligibility tightened to US 501(c)(3) standalones with $500K–$10M operating budgets, priority on the $1–8M band. Select community foundations and regranting orgs are newly eligible. Grants up to 10% of annual budget, unrestricted, no overhead cap.

The application is clear: five questions, 250–300 words each. The Fund is not asking applicants to have an AI product. It is asking them to describe their community, the trust they have built with it, and how they plan to think about AI over the next twelve months. The pitch is institutional, not technical.

Where this network sits, and what we are offering this issue. Two things worth saying plainly. We have been the sole convener of the informal Phase 1 cohort - twelve nonprofits, regular sessions since the grants landed, shared workflows and recorded transcripts. We are sharing the peer-learning paper drawn from those sessions, Building in the Open: The hard part was never the technology to early interested partners (reply to this email or send a note to hello@aifornonprofitsnetwork.org.

And Whitelabel - the sponsor of this newsletter and the convener of the cohort - is also a build partner to several Phase 1 winners; the clearest example is Bonded by Baby, the postpartum AI companion app we are designing, developing and delivering with Social Creatures under the Fund.

If you are drafting a 2026 application, reply to this email. We will share Building in the Open ahead of public release, walk you through what the question set rewards, and where useful, introduce you to grantees who have already done the work. If you are thinking past the application - what you would actually build, or whether the build you have in mind is the right one - we are happy to talk. We are doing it now with several Phase 1 winners.

Six findings run through the paper, none of them about model capability.

1. The first build is usually a misunderstanding. Adoption stalls on what AI even is, not on whether the tools work. A board expects a chatbot, the team built a predictive model, someone asks “where’s the actual AI?” The friction is comprehension before it is anything else.

2. Every project hits the same wall, and the wall is trust. Two shapes of trust, every time: where sensitive data lives, and what the system says to a person who is not okay at 2 a.m. Different missions, identical question.

3. The tools are dissolving. Several cohort organizations are drifting off packaged software toward bespoke, internally built tools. The new hire that matters is not “an AI person” - it is somebody close to the mission, fluent with the machine.

4. “Time saved” is the wrong scoreboard. AI does not give you your time back. It forces a choice about what your time is for, and that choice exposes what a team actually values.

5. You build trust by being honest, not by selling. The organizations getting buy-in admit the trade-offs out loud and design with users, not for them.

6. The cohort treats “good actor” as a claim to be tested, not a slogan. Open sharing of workflows is what stops the same nonprofit problem from being solved twelve separate times across the sector.

Four lessons worth taking back to your team:

1. Apply if you are eligible. Unrestricted, no overhead cap, no formal reporting — these are unusually clean grant terms in the 2026 climate. The $1–8M operating-budget band is the Fund’s stated priority window.

2. Answer the five application questions in the order they are asked. The first four are about your community and the trust you have built with it. The AI question is the fifth. The structure is the brief.

3. Ask us for an early copy of the whitepaper before you draft. Six findings from twelve Phase 1 grantees save you from reinventing them in your first six months of funding.

4. If you sit outside the eligibility window, read the round as a map. The Fund’s tightening to small and mid-sized standalone community organizations is itself a signal of where institutional AI philanthropy is heading next.

Send us an email now if you are thinking of applying and need support.

Share

3) ⭐ Interesting News and Funding Calls

• OpenAI Foundation - 2026 People-First AI Fund. $50M in unrestricted grants. Applications open now via the Apply portal; close July 15 at 11:59 PM PT.

• Anthropic - Claude Corps host applications. Open across all three cohort start dates (Oct 2026, Jan 2027, Aug 2027) via the host portal. Fellow applications close July 17 for the October cohort.

• CEP’s State of Nonprofits 2026 — still the most authoritative read of the operating conditions an AI investment conversation has to land into. Headline: 46% of CEOs say their own burnout is “very much” a concern (up from ~30% in 2025); 39% are running a deficit, up from 22% in 2022. Report.

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

• 👀 What Caught Our Eye: Inside the OpenAI grantee cohort - what’s actually being built, and what the room kept circling back to

• 💭 Thought for the Week: Making Value from Efficiency

• ⭐ Interesting News and Funding Calls

• 🌐 From Across the Network

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai

Subscribe now

1) 👀 What Caught Our Eye: Inside the OpenAI grantee cohort

On May 21, seven nonprofits funded by the OpenAI People-First AI Fund joined the second of our peer-learning sessions.

What is being built is more varied than the public framing of the fund suggests. Project Equity is building AI to for employee ownership. It Gets Better is mid-rollout on a more engaging public-facing mental-health guide for LGBTQ+ youth, after a structured consultation with its youth ambassadors that Brian Wenke is documenting on Substack. The Women’s Building is treating internal AI literacy as the first deliverable and the tools as the second. The Forbes Funds is building cross-platform evaluation. Seal Future Foundation has internalized grant writing, and has a community app entering beta. Social Creatures is building an AI companion to its postpartum-loneliness program for the moments a clinical group cannot reach.

The bottleneck is no longer technical literacy; it is the choice of what to build.

Why it matters for nonprofits. This is, in scope and seriousness, close to what most foundation-funded AI work will look like in eighteen months.

What to do this week. Pick one workflow your team has automated. Ask whose hours were freed and where those hours went. If the answer is “back into the same job, faster,” that is a leadership decision, not a technology one.

2) 💭 Thought for the Week: The honest question isn’t what AI saves you. It’s what you’ll do with the time

Halfway through that same session I found the discussion turned interesting. The sector should get away from the verbiage of “time saved” and skip past “time optimized” too. We started to discuss feeling busier than ever, with all these tools around us to help us be more efficient.

The freed hours from efficiency gains will be absorbed back into the same job, faster - more grant applications, more dashboards, more all-staff Slack channels: we fill the time we have with work, so when AI hands us back time, we fill that with work too. Whether this happens is a leadership choice.

That choice lands into one of the worst-conditioned sectors in a generation. The Center for Effective Philanthropy’s State of Nonprofits 2026 report, published two weeks before the session, found 46% of nonprofit CEOs now report burnout as “very much” a concern - up from just under 30% in 2025 - and 39% of organizations are running an operating deficit. The structural pressure is to absorb productivity gains as more throughput. It is the board asking next quarter why the development team has not doubled its output now that “AI is doing half the work.”

A second thread ran underneath: the only real career risk now is resistance to learning. Describing a sector that has handed its workforce a new burden of continuous adaptation and is calling it opportunity.

There is a version where the time is used differently; a CEO being more forward-facing for the mission rather than absorbed by operations or more reach in inaccessible times. It was all about more presence.

Five lessons worth taking back to your team:

1. Stop reporting AI in hours saved. Report it in hours redeployed. If you cannot say where the hours went, you have a productivity tool, not an AI strategy.

2. Decide the redeployment in advance. Name what the freed capacity is for before the workflow ships.

3. Treat AI adoption as culture work first, tools work second. Education before deployment is more replicable than most pilot projects.

4. Audit the burden you are pushing onto staff. Continuous learning is now part of the job. Make space for it in workload, evaluation and pay.

5. Ask your board the harder question, in writing. Is the intent to free hours for the mission, or to do the same work with fewer people?

Leave a comment

Share

3) ⭐ Interesting News and Funding Calls

• CEP's State of Nonprofits 2026 — released May 14, and the sector backdrop against which any AI investment conversation now needs to land. Headline numbers: 46% of CEOs say their own burnout is "very much" a concern (up from just under 30% in 2025); 57% say foundation grants have been harder to secure since January 2025; 39% are running a deficit, up from 22% in 2022. Report.

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

👀 What's Caught our Eye: Anthropic and the Gates Foundation put $200M on the table

💭 Thought for the Week: Humanity AI's first cohort went to its friends. The open call is where the coalition either widens out or doesn't.

⭐ Interesting News and Funding Calls

🌐 From Across the Network

Subscribe now

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through the AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai - They’re giving $50k in technical services to run fundraising pilots - sign up!

1) 👀 What Caught Our Eye: Anthropic and the Gates Foundation put $200M on the table

On May 14, Anthropic and the Gates Foundation announced a four-year, $200 million partnership across global health, life sciences, education and economic mobility - grants, Claude credits, and Anthropic engineering time pooled into a single commitment. Anthropic calls it the largest deal of its kind between an AI company and a global philanthropy.

The structure is interesting: The largest single slice goes to health work in low- and middle-income countries, where roughly 4.6 billion people lack access to essential health services. Named programs: AI-assisted vaccine and therapy screening for overlooked diseases, starting with polio, HPV and eclampsia/preeclampsia. Named partners: the Institute for Disease Modeling, which sits inside the Gates Foundation, on malaria and tuberculosis forecasting. The education strand puts Claude behind K-12 tutoring tools in the US and foundational literacy and numeracy apps in sub-Saharan Africa and India. The mobility strand funds portable skills records and career guidance for people moving into or retraining for the US workforce. Datasets, benchmarks and African-language data are due to be released as public goods later this year.

The partnership itself is also newsworthy - Anthropic is putting in compute, credits and engineers. Gates is putting in domain access, implementing partners, and political cover. Both sides come out with something they could not buy on the open market.

Why it matters for nonprofits. Some version of this partnership is about to be pitched to large nonprofits and mid-sized funders. It will not be at this scale; the structure might be similar - one question to ask is what each side is putting in and what each side is taking out.

Subscribe now

Share AI for Non-Profits Network

2) 💭 Thought for the Week: A coalition’s first grants always go to its friends

The Humanity AI coalition - ten of the largest US foundations, including MacArthur, Ford, Mellon, Mozilla, Omidyar and Packard - announced its first round of grantmaking on May 12. The headline number was $18 million. Eight million went to nine organizations at half a million each, including the AI Now Institute, the Center for Democracy and Technology, the Partnership on AI, the Pulitzer Center and Student Defense’s SHAPE AI. Three million went to a new “AI Civics” program led by Data & Society with the Digital Public Library of America. The remaining ten million will go out in an open call this summer.

Four days later, Inside Philanthropy made the obvious point in print: every inaugural grantee had already been funded by one of the ten founders. John Palfrey, MacArthur’s president, told the magazine those organizations were the ones with “a long track record of making a difference on AI” who were “simply short of capital.” Fair enough as a starting move. Chantal Forster, formerly of the Technology Association of Grantmakers, put the sharper version on the record: “The existing set of grantees are all East-Coast- and West-Coast-based - what’s missing are groups in the heartland, groups in the South… it’s ignoring the rural question.”

The grantees announced this month are excellent organizations. They are also the ones anyone who has been to a tech-policy convening in the last decade could have listed on the back of a napkin without consulting the press release. The first cohort is not a coalition betting on new voices. It is a coalition consolidating around its existing ones.

The legible grantee - track record, staff, indirect-cost rate, relationships - is the safe grantee. Funders feel pressure to move quickly, as Palfrey said, and quickness costs you reach. The organizations Forster is describing - rural, Southern, building from inside affected communities rather than commenting from outside them - do not have the same intake apparatus. They will only appear in the open call, and only if the open call is structured to find them.

So the test is the open call. Ten million dollars, opening this summer. The criteria, the outreach, the geography of the review panel, and the indirect-cost rate of the eventual grants are the four things that will tell you whether Humanity AI is doing AI philanthropy differently, or whether it is the same circle with a better logo on the cover sheet.

A couple of quick thoughts:

1. Watch the open call, not the inaugural list. The first cohort tells you who the coalition already trusts. The open call tells you whether it is trying to enlarge that circle.

2. Ask your own board the same question. When your organization last hired, funded or partnered - was the first move to your friends?

Leave a comment

Share

3) ⭐ Interesting News and Funding Calls

• Humanity AI’s $10M open call - opening this summer. Criteria, application window and panel composition not yet announced. Watch the Humanity AI site and join the mailing list there.

• CEP’s State of Nonprofits 2026 - released May 14. The most authoritative recent read of the operating conditions the sector is in. Headline numbers: 46% of CEOs say their own burnout is “very much” a concern (up from just under 30% in 2025); 57% say foundation grants have been harder to secure since January 2025; 39% are running a deficit, up from 22% in 2022. Read this before any board conversation about AI investment. Report.

• Code for America + Anthropic - the SNAP Policy Navigator - announced at the Code for America Summit on May 8. A Claude-based tool that gives caseworkers a plain-language, sourced answer to specific SNAP policy questions, with eligibility decisions deliberately left with the human. The most concrete public-interest AI deployment in our window, and a useful model for human-in-the-loop scoping.

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

👀 What's Caught our Eye: AI search is eating nonprofit organic traffic - and that may not be all bad

💭 Thought for the Week: The boring half of the donor file is where the AI math actually compounds

⭐ Interesting News and Funding Calls

🌐 From Across the Network

Subscribe now

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through the AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai - They’re giving $50k in technical services to run fundraising pilots - sign up!

1) 👀 What Caught Our Eye: AI search is eating nonprofit organic traffic - and that may not be all bad

The 2026 M+R Benchmarks Study, published last month, shows that organic search traffic to US nonprofit websites fell steadily through 2025. M+R points to AI Overviews on Google and the migration of search queries into Claude, ChatGPT, and Gemini.

Organic search still accounts for 39% of all visits, so the channel is some distance from crisis. The line has bent, though, and every monthly slice in the data tells the same story. Read alongside the Blackbaud Institute’s finding that digital fundraising revenue grew about 11% across the year, the picture becomes harder to read than the headline allows: traffic falling, money holding.

At AWS Imagine in London last week, more than one communications lead told us that the visits they were losing had been the wrong visits all along - informational lookups, school assignments, casual readers who were never going to give. AI tools, on this reading, are stripping out an audience that always cost more in server time than it yielded. The interpretation is generous, and lines up uncomfortably with a second strand in the Blackbaud data: gifts under $1,000 fell by 1.1% in 2025, while gifts of $1,000 or more rose by 4.7%. The visitors a generic AI chat answer would always have satisfied are the ones disappearing first.

Why it matters for nonprofits. The acquisition funnel is shifting before the gift conversation is. If fewer prospects reach the page at all, the trust signals on the pages they do reach carry more weight, and the share of new donor relationships beginning in organic search becomes a quarterly board question.

What to do this week. Pull eighteen months of organic-search traffic to your donation pages and read the trend rather than the total. Then run three honest questions about your organization through an AI Overview and a chatbot. Few teams have looked.

Subscribe now

Share AI for Non-Profits Network

2) 💭 Thought for the Week: The boring half of the donor file is where the AI math actually compounds

Monthly Giving Awareness Week ran from May 11 through 15, coordinated by RKD Group, GivingTuesday, and Positive Equation. Its second outing drew a fraction of the press that the agentic-AI platform launches at AFP ICON pulled the week before. Given the numbers underneath, the proportions are upside down.

The 2026 M+R Benchmarks shows that monthly giving accounted for 27% of all online revenue across the US sector in 2025. Many organizations now take a third or more of their revenue from sustainers. Recurring giving is the largest and most reliable growth lever in the fundraising stack, and it has very little to do with AI.

Most of the AI-in-fundraising conversation right now sits at the top of the funnel - prospecting, list-building, outreach to people who have never heard of the organization. Very little of it sits with the sustainer who is about to lapse. The vendor opportunity is at the top of the funnel, and the discourse follows.

Whilst there has been a few big announcements of AI donor management tools - they usually point at acquisition and major-gift cultivation. The math lives further down the file. For a mid-sized US nonprofit, a five-point lift in monthly-giving retention compounds across years at flat cost-to-serve, and tends to be worth more than a fifty-point lift in acquisition that has to be re-earned every quarter. Retention is still handled in most teams as a hygiene function, passed to whoever is junior enough that no one fights over the brief.

What AI contributes here is real, and quieter than the vendor brochures suggest. A model can spot a sustainer whose engagement pattern has shifted six weeks before they cancel. It can draft a reactivation email that references the actual program a lapsed donor used to support, rather than an all-purpose ‘we miss you’. It can sit on top of a CRM export and tell a fundraiser which 12% of sustainers will probably upgrade if asked, and which 4% will resent the ask. Several of those things move the numbers more reliably than anything in the agentic catalog.

We are writing with skin in the game. The five $50,000 technical services pilot grants we announced last month are aimed precisely here. Retention, reactivation and the middle of the donor file are what the cohort will be testing AI against, with real numbers and external evaluation. Acquisition is a genuine problem; we think the leverage on the other side of the file is larger.

Five lessons worth taking back to your team:

1. Look at sustainer retention before acquisition tools. A point of retention compounds across years; a point of acquisition has to be re-won every quarter.

2. Treat reactivation as a strategic project rather than a year-end campaign. AI moves the math here more than anywhere else.

3. Audit where your AI vendor budget actually goes. If it all sits at the top of the funnel, you are paying to chase strangers while the people who already love you quietly drift.

4. Ask your CRM provider what they can show you about sustainer churn signals. A vague answer is a strategic gap with a software face.

5. Be honest about whose problem you are solving. Sustainer retention is invisible from a board slide; that is why the AI conversation skips it.

Leave a comment

Share

3) ⭐ Interesting News and Funding Calls

• OpenAI Foundation: Wave 2 of the People-First AI Fund - $9.5M in board-directed grants for nonprofits in health, education, and community resilience, expected to be announced shortly. Wave 1 sent $40.5M to 208 US nonprofits. Worth reading the Wave 1 grantee list before pitching.

• Blackbaud Institute 2025 Trends in Giving — total giving up 4.3% in 2025, but gifts under $1,000 fell while gifts above $1,000 grew. The midlevel donor squeeze is real. Resource page.

• 2026 M+R Benchmarks Study — the annual reference data on digital fundraising for the US sector. Read it.

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

👀 What's Caught our Eye: Donors say AI summaries aren’t moving them

💭 Thought for the Week: Live from AWS Imagine London - what the strongest sessions had in common

⭐ Interesting News and Funding Calls

🌐 From Across the Network

Subscribe now

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through the AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai - They’re giving $50k in technical services to run fundraising pilots - sign up!

1) 👀 What Caught Our Eye: Donors say AI summaries aren’t moving them

The Give.org Donor Trust Report 2026 was published last week: only 7% of donors say AI-generated summaries influenced their giving decisions in 2025.

The report draws on a December 2025 survey of more than 1,500 US adults and nine years of trend data. Openness to being approached by charities reached 42.1%, the second-highest level since 2017, and 63% of Gen Z respondents say they might give more if asked.

The trust gap is the problem. Nearly 68% of respondents say trusting a charity is essential before they will give; only 18% report actually having high trust in the charities they encounter. And what closes that gap is not, on this evidence, AI. Among donors who research before giving, the dominant influences are charities’ own websites (54%) and third-party monitors like Charity Navigator and BBB’s Give.org (39%). AI summaries trail well behind.

This matters for two reasons. Every fundraising vendor right now is promising to help you “show up better” in AI-generated answers - the data suggests donors are not making decisions there. And Bennett Weiner, who runs Give.org, said disengagement is increasingly about “uncertainty about whom and what to trust” rather than indifference. Donors are hunting for credibility, not eloquence.

What to do this week. Open your About, Impact, and Annual Report pages and read them as a donor who has never heard of you. Is the leadership team named? Is the financial breakdown visible without a download? Do you link out to your third-party ratings? Fix those before you spend another hour on AI search optimization. The trust signals donors actually use are still the ones you fully control.

2) 💭 Thought for the Week: Live from AWS Imagine London - what the strongest sessions had in common

I spent Tuesday morning at AWS Imagine Non-Profit in London. The key message “Build responsibly with AWS - Governance, Ethics, Accountability.” Two years ago, that would have been the warm-up before the real pitch about agents and models. Today it was the key pitch from AWS with interesting case studies from across their network.

After three years in which every cloud-vendor event for our sector has felt like a relay race of new acronyms, Imagine 2026 was almost stubbornly unglamorous about its tooling. The strongest sessions were about applied AI in multiple non profit contexts; how to choose a pilot, how to talk to a donor in their own language, and how to give up on things that aren’t working. The headline isn’t the technology. It’s the operational discipline a sector squeezed by government-funded scarcity rarely allows itself.

Three moments from the morning I thought relevant to share with the network:

The first was The HALO Trust, the mine-clearance non-profit, as well as the huge data they are ploughing through to find mines through satellite and drone imagery, they were also describing how they claw back roughly two months of staff time a year on donor reporting. Saying something in the language each funder wants to read, with the specific framing they care about. The win is translation, not automation. HALO has not replaced a fundraiser with an AI. It has stopped asking its program staff to write the same impact in seven different dialects of donor.

The second is on how to choose a pilot, delivered with an honesty that is rare in this circuit. The framework was simple. Pick a project where the problem is clear and the positive outcome is one that can be moved. Pick a partner you would happily share bad news with - credibility matters more than capability. Start small and build a positive feedback loop across your organization before going wide. Keep things in pilot for longer than feels comfortable. Find the problem before you reach for the solution. And “get comfortable with failing.” Most non-profit cultures cannot afford to fail publicly. The pilot is the one place where they can fail privately, and the leaders who lose that permission tend to lose the upside too.

The third was where conferences and sessions like this help you think completely outside what you normally think of. Like turning to ballet dancing to understand why grave diggers are being injured. The Royal Academy of Dance, just named a Pathfinder Award winner in the generative AI category, is modernizing Benesh Movement Notation - a 1950s system originally designed to write ballet on paper - as the structured backbone of an AI-powered encyclopedia of dance knowledge. The speaker mentioned, almost in passing, that a team in Paris had used the same notation system to analyze why gravediggers were getting injured. The notation is that versatile because it describes the human body moving in space, which turns out to be useful well beyond ballet. The takeaway from the talk was not “AI saves dance.” It was the inverse: a small specialist non-profit was sitting on a piece of niche structured knowledge that nobody else in the world had, and AI is what finally made it operationally useful.

The temptation for non-profits is to copy what enterprise is doing with AI. The actual opportunity is to surface the specific, hard-won, often analogue expertise that makes your organization indispensable in the first place - and to be willing to fail in private while you figure out how to put it to work.

Five lessons I found useful:

1. Pick the pilot before you pick the platform. Vendors will help you with the second; only you can do the first.

2. Donor reporting is a translation problem, not an automation one. Solve for “in their language” before you solve for “at speed.”

3. Choose partners you would tell the truth to. The most sophisticated tool is worth less than a credible, patient collaborator.

4. Keep projects in pilot for longer than feels comfortable. Promotion to “live” creates pressure to claim success before you have earned it.

5. Get comfortable with failing - and make sure your governance has a way to retire a project. If it doesn’t, you don’t have a governance policy. You have a one-way door.

AWS are releasing a number of Imagine grants for non-profits - and we’re keen to partner with you if you are thinking of applying - just reply to this email or drop us a note on hello@aifornonprofitsnetwork.org.

Leave a comment

Share

3) ⭐ Interesting News and Funding Calls

Data Shows Gen Z Is In The Beginning Stages Of Philanthropy (Non Profit Times)

Von der Leyen, Hillary Clinton back new push to childproof AI (Politico)

Enterprise AI Is Stalling: 46% of Initiatives Fall Short Despite Rising Investment (Yahoo)

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

Every nonprofit in the network has tried AI. Most have a paid ChatGPT seat or two. A few have piloted Copilot. Almost none have moved a single donation metric because of it.

That’s the gap we want to close - not with another webinar, not with a vendor pitch deck, but with five real, funded, end-to-end deployments. So this week’s issue is a special one.

Or go straight tot he grant readiness check

Share

👀 01 · The problem we’re funding to fix

Everyone’s using AI, but who’s seeing the benefit?

We’ve spent the last six months talking to nonprofit ops, fundraising and digital leads across the network. The pattern is consistent: tools are everywhere, outcomes aren’t. Teams report time saved drafting emails but almost none report a measurable lift in donations, click-through, or donor satisfaction.

The orgs that are seeing returns share three things - they treat AI as a teammate not a tool, they have one internal champion empowered to remove friction, and they ship inside enterprise-grade safeguarding from day one.

So we partnered with Whitelabel to fund five nonprofits to do exactly that - properly, end-to-end, in the open.

★ THE DROP

5 grants · $50K each · 6-month pilot

AI for Nonprofits Network (the convening side) × Whitelabel (the build side)

Five nonprofits. Six months. Full pilot deployment of an AI fundraising stack - funded end-to-end, then pro-bono extended through 2027 and 2028 for teams that adopt and want to keep going.

Take the 90-second pre-app

Subscribe now

🎯 02 · Who this is for - score yourself

What’s the catch, and is your team a fit?

The catch is honest. This isn’t a tool you bolt on. It’s a six-month operational change to how your team raises money, and we’re only funding five teams who can actually carry that.

Interested? Complete a 90-second readiness checker. It isn’t designed to gatekeep funding - it’s designed to make sure the grant lands somewhere it’ll move. Tick what’s true for your team - you’re a strong fit if you can honestly check at least four of six.

★ Self-score in 30 seconds

★ 01 · Champion — Named internal lead with authority. One person empowered to remove blockers across fundraising, ops and IT.

★ 02 · Mandate — Board & exec are bought-in. Senior team is genuinely open to experimentation inside live fundraising - not “let’s see the slide deck first”.

★ 03 · Compliance — Enterprise safeguarding. You’re ready to deploy AI inside SOC 2 and donor-data hygiene from day one.

★ 04 · Stack ready — Donor data is exportable, somewhere. CRM, email tool, payment processor - even messy. We can integrate. We can’t conjure data that isn’t written down.

★ 05 · Capacity — 1 hr / week from the champion. Six months of pilot needs steady contact, not heroics. If your champion is already at 110%, this isn’t the round.

★ 06 · Open — Comfortable being a public case study. Open data, open results, shared with the network. The grant trades funding for visibility.

Score 4 / 6 or higher? → strong fit. 90 second pre-app here

What we’ll fund

• Full onboarding & deployment time from the Whitelabel team for the six-month pilot window.

• All AI token usage across agents, models, and integrations for the duration of the pilot.

• Two further years of pro-bono software licensing through 2027 and 2028 for teams that adopt the stack.

What we won’t

• Cash grants - this is a deployed pilot, not unrestricted funding. The value lands as software, time and reduced overhead.

• Replacing your fundraising team - the AI sits alongside humans, takes a clearly scoped role, and reports up.

• Side-of-desk pilots - if there’s no champion and no mandate, we’d rather hold the spot for a team that’s ready.

• Black-box deployments - every tool and integration is visible to your team and the network.

The grant is structured so that even after the pilot, your donor processing overheads go down, not up. The grant doesn’t end with a renewal cliff, it ends with a cheaper, smarter fundraising operation than you started with.

🔧 03 · What you’ll actually deploy

The pilot, the metrics, the timeline

Here’s what the pilot will actually test, and the three uplift metrics every funded team will be measured against.

What the pilot tests

01 · Autonomous donation agents - State-of-the-art tests on hyper-personalized email campaigns, AI landing pages, and a new digital front door designed around the donor - not the CMS template.

02 · Hyper-personalized newsletter & touchpoints - Each supporter gets content shaped to their giving history, interests and engagement pattern - at the scale and frequency a comms team of three could never reach manually.

03 · An AI team member on fundraising - A dedicated agent that owns monthly-donor and individual-donor giving strategy alongside your humans - segmenting, nurturing, and following up like a fundraising hire who never sleeps.

★ The three uplift metrics · measured monthly

1. Increase in donation uptake across individual and / or monthly donors — target +10%

2. Click-rate to donation interest across all measured outbound touchpoints — target ↑

3. Satisfaction lift on existing-donor engagement measured at every touchpoint we can attribute — target +%

Interesting in the uplift? - take the 90-second pre-app here

Timeline

• Phase 01 · May–Jun 2026 — Apply & qualify. Express interest, complete the readiness check, conversations with the partnership team.

• Phase 02 · Jul–Dec 2026 — Pilot · 6 months. Full deployment, weekly check-ins, monthly metric reviews, open case-study writeups.

• Phase 03 · 2027–28 — Sustain. Pro-bono licensing continues, 50% commission reduction, continued integration support.

🌐 04 · Before you apply - a few honest notes

A couple of pieces of context before you decide whether to apply - what others in the network are doing, and where this grant sits relative to where the field is heading.

Operator note · “The model isn’t AI vs. humans”

Most of the orgs already shipping say the same thing, the lift comes from giving the AI a clearly scoped role, not from replacing anyone. The pilot is built around that lesson.

Funder note · Why a grant, not a discount

A discount asks small teams to find budget they don’t have. A funded pilot lets you prove the lift first, then absorb the (now smaller) cost into next year’s plan.

Compliance note · Enterprise-grade, by default

Every deployment ships with the certifications a corporate IT team would expect - because nonprofit donor data deserves that bar, not a lower one.

The honest bit · Five spots. Self-select carefully.

If you’re stretched, mid-rebrand, or your board hasn’t bought into experimentation - apply for the next round. This one is for teams that can move in July.

★ Ready to be one of the five?

Drop your details, take the readiness quiz, and we’ll be in touch within five working days.

→ Express interest in 90 seconds here

5 grants · $50K each · pilot starts July 2026

That’s it for this week. Forward to one nonprofit lead who needs to see this or, if that nonprofit is yours, hit the link above.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

👀 What's Caught our Eye: Everyone’s using AI - but who’s seeing the benefit?

💭 Thought for the Week: Five useful prompts, and the work they’re doing

⭐ Interesting News and Funding Calls

🌐 From Across the Network

Subscribe now

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through the AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai

1) 👀 What Caught Our Eye: Everyone’s using AI - but who’s seeing the benefit?

A new benchmark study from Virtuous and Fundraising.AI, based on a survey of 346 nonprofits, suggests Ninety-two percent of nonprofits are now using AI in some capacity, but just seven percent say it has produced major improvements in what their organization can actually accomplish. The authors call the gap the “efficiency plateau” - and it is, on the evidence, where most of the sector is stuck.

The underlying numbers are specific. Eighty-one percent are using AI individually, without shared workflows. Forty-seven percent have no AI governance policy at all. The pattern is familiar: someone in fundraising uses ChatGPT to draft appeals, a program manager runs reports through Claude, comms has a Gemini subscription nobody authorized. First drafts get faster. Routine tasks get automated. Mission impact barely moves.

What separates the seven percent who are seeing the needle move?

• Clear governance

• Documented workflows

• Cross-functional ownership

• Consistent measurement.

The organizations getting transformation out of AI are not the ones with the most sophisticated tools. They are the ones treating AI as an organizational capability to be built, rather than a personal productivity hack to be discovered.

Three practical implications. First, audit honestly: how much of your organization’s AI use is individual experimentation versus shared practice? If it’s mostly the former, you are probably on the plateau. Second, a governance policy is no longer optional - and it does not need to be complex - we outlined one in our recent workshop that you can download for free (just drop me a message). A one-page document covering acceptable use, data handling, and disclosure is enough to start. Third, pick one shared workflow and redesign it end-to-end around AI, with defined roles and a way to measure whether it is working - again - we can help here, just reach out.

2) 💭 Thought for the Week: Five useful prompts, and the work they’re doing

The dominant public narrative about AI in nonprofits is still largely programmatic - chatbots for beneficiaries, efficiency gains in service delivery, AI-assisted grant writing. These are real and worth attention. But from my recent conversations with executive directors there are a few more private and less discussed patterns: AI is becoming a thinking companion for the person at the top, doing work that used to be done alone, or not at all.

What unites the most useful executive uses of AI is that none of them are really about output. Nobody getting real value is using AI to write the strategic plan or the statement to staff. They are using it to think more clearly before doing those things themselves. Below are five prompts worth saving somewhere accessible. Each one is written to be pasted in and adapted.

The board pack. When a hundred pages land in your inbox the night before the meeting, the temptation is to skim. The better use of AI is to orient before the reading, not around it.

“I’ll paste in the board pack below. Give me (1) the three decisions the board is actually being asked to make, (2) the two items most likely to generate pushback, and (3) anything that looks inconsistent with previous papers.”

The meeting prep. Executive directors walk into meetings cold more often than they would admit. Ten minutes with an AI tool and the right prompt is not the same as an hour of genuine preparation, but it is meaningfully better than nothing, which is what the alternative usually is.

“I have a 30-minute meeting tomorrow with [person]. Here’s what I know: [paste]. What are the three things they most likely want from this conversation, what’s the question I should be ready for but probably won’t be, and what’s a useful opening I could offer?”

The difficult email. The draft sitting in your drafts folder is either too blunt or too mushy. AI is useful here not because it writes better than you do, but because it catches what you cannot see in your own prose.

“Here’s a draft email to [person]. Context: [situation]. My goal is to [outcome]. Rewrite it so it’s direct but warm, and flag anything in my original that could be read the wrong way.”

The steelman. Executive leadership is lonely. Most EDs have nobody in their organization who can push back on a half-formed decision without it becoming political. AI, despite its well-documented tendency toward sycophancy, can be instructed to argue back.

“I’m planning to [decision]. Here’s my reasoning: [paste]. Argue the strongest possible case against. Don’t strawman - give me the version a thoughtful, sympathetic critic would actually make. What am I not seeing?”

The narrative finder. The numbers are in front of you; the story is not. This is the use that most clearly illustrates AI’s value as scaffolding rather than substitute — it finds the pattern, you decide what to do with it.

“Here’s our impact data from the last quarter: [paste]. What’s the actual story here? What’s surprising, what’s worth celebrating, and what should we be honest about? Don’t spin - I want to understand what’s really happening.”

A final thought: The EDs getting the most out of AI right now are not the ones with the most sophisticated setups. They are the ones who have figured out that the technology is most useful when it is making them slower, not faster - slower to walk into meetings unprepared, slower to send emails they will regret, slower to commit to decisions they have not really examined. The value is almost never in the first response the AI gives. It’s in the second and third, after you’ve pushed back, corrected its assumptions, and made it work harder. Treat it as a staffer worth training, not an oracle worth trusting, and it earns its place in your week.

Leave a comment

Share

3) ⭐ Interesting News and Funding Calls

How to influence the future of AI (Charity Digital)

Renaissance Philanthropy Launches Forecasting Hub To Predict AI’s Impact On The Workforce (Business Insider)

Mozilla’s open source AI alliance is taking on Big Tech (The Logic)

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

👀 What's Caught our Eye: San José City Council voted unanimously to spin the GovAI Coalition out into an independent nonprofit

💭 Thought for the Week: The Volunteer Problem for Nonprofits Implementing AI

⭐ Interesting News and Funding Calls

🌐 From Across the Network

Subscribe now

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through the AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai

1) 👀 What Caught Our Eye: The Safety Net for AI in the Public Sector Just Got a Permanent Home

For the past two years, one of the most important AI governance initiatives in the United States has been run almost entirely by volunteers. That changed this week.

On April 7, the San José City Council voted unanimously to spin the GovAI Coalition out into an independent nonprofit - a milestone for an organization that has grown from a single video call with 50 agencies in November 2023 to a network of more than 3,000 members across 900 public agencies.

The coalition’s origin story is instructive. It began when San José’s AI and digital privacy team, frustrated that vendors were reluctant to share information about their products’ data practices, reached out to counterparts in other agencies and found the same concern everywhere. The solution was collective action. The result was a free, openly available library of policy templates, procurement guidance, and AI governance frameworks that any government agency - or nonprofit - can download and adapt.

The coalition’s primary output has been a growing library of practical resources: fact sheets on AI assistants, a public-sector AI skills training playbook, and model documents for procurement and data-sharing agreements. These are exactly the tools that we hear as a network many small and mid-sized nonprofits lack the budget to develop themselves.

The transition to nonprofit status is being funded by the Packard Foundation and will take approximately nine months, with an interim fiscal sponsor in place in the meantime.

As San Diego CIO Jonathan Behnke, a coalition board member, put it: “The current model employs very dedicated volunteers and the transition to a nonprofit will remove some of those constraints and allow dedicated staff to devote more time to delivering value for the coalition membership.”

What this means for your organization: The GovAI Coalition’s resources are free, publicly available, and designed for organizations without dedicated AI staff or legal teams. If your nonprofit has been putting off developing an AI governance policy because it felt too complex or too expensive, this is your shortcut. And watch this space: as the coalition transitions to nonprofit status, opportunities for the social sector to engage directly - and shape the standards being developed - are likely to grow.

2) 💭 Thought for the Week: The Volunteer Problem for Nonprofits Implementing AI

Here is a question that keeps coming up in my conversations with nonprofit leaders across this network: what happens to your volunteers when the admin work disappears?

I’ve been speaking with executive directors and operations leads over the past few weeks - organizations of every size and type, from food banks to arts charities to housing advocacy groups - and a pattern is emerging. Leaders are excited about AI - they’re automating data entry, streamlining intake processes, offloading the scheduling and follow-up emails that used to eat whole afternoons. For teams that have been running on empty for years, the relief is palpable.

But almost nobody is asking the harder question underneath it.

Ask any leader what eats most of their volunteers’ time, and the answer is usually some version of the same list:

• Data entry.

• Sorting donations.

• Fielding intake calls.

• Scheduling.

These are the tasks that consume hundreds of thousands of volunteer hours every year - and they are precisely the tasks AI is now capable of automating, at scale, for a fraction of the cost.

Volunteers are not primarily motivated by a paycheck. They show up because they feel useful, because they want to make a tangible difference, because working alongside a stranger toward a shared goal creates a kind of connection that no chatbot can replicate. Strip away the administrative scaffolding that gives many volunteers their foothold in an organization, and you don’t just lose labor hours. You risk losing people.

This isn’t abstract. The US nonprofit sector relies on roughly 60 million volunteers annually. But that army of people is not simply a source of free labor - and the leaders I speak to know this better than anyone. Volunteers are the civic infrastructure that sustains community connection, public trust, and donor pipelines. Research consistently shows they are among the most likely people to become major donors. The volunteer relationship is, for many organizations, the on-ramp to everything else.

What I’m hearing across the network is that the sector hasn’t yet developed language for this tradeoff, let alone a strategy for managing it. The dominant narrative around AI in nonprofits is almost entirely operational: save time, reduce errors, do more with less. That framing treats volunteers as a means to an end rather than as stakeholders with their own relationship to your mission and their own reasons for showing up.

The opportunity, and I genuinely believe there is one, lies in organizations that redesign their volunteer programmes around relational depth rather than task completion. Mentoring. Advocacy. Storytelling. Community listening. The kind of work that can’t be automated, and shouldn’t be. But that future requires deliberate choices made now, before the efficiency wave reshapes volunteer programmes by default rather than design.

The conversations I’m having across this network give me real cause for optimism. What’s missing is a shared framework for acting on those instincts. If you’re already wrestling with this - or if you’ve found approaches that are working - I want to hear from you. Hit reply.

Leave a comment

Share

3) ⭐ Interesting News and Funding Calls

AI companies know they have an image problem. Will funding policy papers and thinktanks dig them out? (The Guardian)

We gave an AI a 3 year retail lease in SF and asked it to make a profit (Andon Labs)

Nonprofit AI Artists Alliance Seeking Community Input (Santa Barbara Independent)

Evaluating Scalability, Reproducibility, and Impact of GenAI and Agentic AI in the Water and Wastewater Sector (The Water Research Foundation)

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.

👀 What's Caught our Eye: The Letter That Changed Everything - How AI Is Freeing Nonprofits From Their Own Inboxes

💭 Thought for the Week: Big Tech Comes Bearing Gifts

⭐ Interesting News and Funding Calls

🌐 From Across the Network

Subscribe now

This newsletter is supported by Whitelabel.ai - helping nonprofits cut through the AI noise with practical tools built for mission-driven teams. Find out more at whitelabel.ai

1) 👀 What Caught Our Eye: The Letter That Changed Everything - How AI Is Freeing Nonprofits From Their Own Inboxes

For a child living in poverty in Colombia or Zambia, a letter from their sponsor is more than correspondence, it is proof that someone, somewhere, cares. For Children International, the Kansas City-based organization supporting impoverished youth across the globe, making sure those letters actually arrived, translated, processed, and delivered, was consuming an enormous amount of staff time. Time that was supposed to be spent on the mission.

The organization’s solution was elegantly simple: AI agents now handle bulk translation of the letters that flow between donors and the children they support. What once required hours of painstaking human effort is now largely automated. As Tim Batcha, Children International’s vice president of Global Information Technology, told the Microsoft Global Nonprofit Leaders Summit last week, “too much effort was going toward day-to-day operations instead of advancing the nonprofit’s core mission.”

Children International’s story is a useful corrective to the way AI tends to get discussed in the sector, as either an existential threat or a magical fix. This is neither. It is an organization that identified a specific, high-volume, low-discretion task that was quietly draining its capacity, and deployed AI to take it off the plate. The mission work - the relationships, the programs, the human judgment - stays human.

Also at the Summit, Denver-based Project C.U.R.E., which ships medical supplies to clinics around the world, described using AI to search nearly four decades of equipment repair manuals and predict future inventory needs. Its archives had grown too large for any one person to navigate. Again: not a moonshot, a filing cabinet problem, solved.

The practical lesson for nonprofit leaders is to resist the temptation to think about AI at the organizational level until you’ve thought about it at the task level. Walk through a week in the life of your busiest team member. How many hours are spent on translation, data entry, scheduling, report formatting, or triaging routine inquiries? That is your starting point. The mission case for AI is rarely abstract - it is the hours won back, and what your people do with them.

2) 💭 Thought for the Week: Big Tech Comes Bearing Gifts

There is something seductive about the current moment in nonprofit AI funding. Microsoft pledges $5 billion in support for the sector. OpenAI distributes $40 million in unrestricted grants to hundreds of nonprofits. Google, Amazon, and LinkedIn follow with their own programs. After decades of watching technology pass the sector by, nonprofit leaders could be forgiven for feeling that Silicon Valley has finally noticed them.

But should we at least interrogate why?

Full disclosure: we recently celebrated members of this network who won OpenAI funding, and we are actively preparing applications for the latest Gates funding round. We are not writing from the sidelines. Which is precisely why this conversation matters.

None of this is to say that tech companies are acting in bad faith. But good faith and aligned interests are not the same thing, and the difference matters enormously when you are deciding which tools to build your organization around, which platforms to trust with your beneficiaries’ data, and which vendor’s vision of the future you are quietly endorsing with every procurement decision you make.

Big Tech grant programs are primarily motivated by expanding adoption of their own platforms, demonstrating social impact, and building reputational value in the nonprofit sector. Applications that succeed typically emphasize specific, measurable use cases for the funder’s own technology. In essence; the grants are structured to reward organizations that use the funder’s tools. That is not philanthropy in the traditional sense, it is customer acquisition with a halo.

Many nonprofits work with communities already being harmed by algorithmic systems - benefits software that wrongly denies claims, predictive tools that entrench bias in child welfare, automated hiring systems that screen out underrepresented candidates. Organizations that have spent years documenting these harms are now being courted by the same industry to adopt its next generation of tools, grant funding attached.

None of this means refusing to engage. When nonprofits remain absent from the debates shaping how AI is regulated and deployed, efficiency and control tend to define the rules - while values of equity, community accountability, and human judgment get incorporated, if at all, after decisions are already made. Engaging on our own terms is far better than not engaging at all.

But the posture matters - every Big Tech partnership deserves the same scrutiny as any major donor relationship: What does this funder want from us? What are we implicitly endorsing by taking this money? Who owns the data we feed into these systems, and what happens to it if the partnership ends?

We are applying for the Big Tech grants because we believe we can use it to genuinely advance our mission and be part of the conversation. How do you see it? Take the money where it serves your communities? Know exactly what you are trading for it?Never let the funder’s roadmap become your strategy?

Leave a comment

Share

3) ⭐ Interesting News and Funding Calls

Nonprofit Research Groups Disturbed to Learn That OpenAI Has Secretly Been Funding Their Work (Futurism)

Tech Nonprofits to Feds: Don’t Weaponize Procurement to Undermine AI Trust and Safety (EFF)

Navigating the Growing Challenge of Shadow AI in the Enterprise (The Fast Mode)

4) 🌐 From Across the Network

Have an event, case study, gathering or interesting insight you would like to share with the network? Drop us a note by replying to this email.

We’re a network of non-profit leaders interested in how AI is impacting our sector and driving mission. You’re receiving this email because you expressed an interest in the network. If you think this was a mistake, you can unsubscribe using the link below.