As startup employees increasingly use AI to write documents, emails and Slack messages, internal communication problems are fast emerging — with CEOs hurriedly drawing up new policies to combat them.

Last week, San Francisco-based sales intelligence startup Clay published its AI Writing Policy, encouraging employees to think first before using AI to write. In one of its guiding principles, it holds employees responsible for ensuring a document reflects their own thoughts before sharing it.

“If a reviewer asks, “What did you mean by this line?” it’s not acceptable to reply with “Oh sorry, AI wrote that, ignore it,” the policy authored by engineer Sophie Alpert reads.

Dutch startup Polarsteps also published its own AI writing policy this summer. CEO Clare Jones noticed that documents were getting longer and less comprehensible, and messages between team members were often jumbled.

“It was harder to understand what each other was trying to say, because [the writing] had all those words like “genuinely” and “directionally positive” — and oh my gosh, “quietly” appeared in everything,” she told Sifted. “I was having to squint at the document to really understand what’s being said — you know, what’s the thing behind the drama?”

Subscribe now

Sloppifying company comms

In a fast-moving startup environment, communicating clearly and concisely is important to keep teams aligned.

Large language models can’t yet produce accurate, well-sourced text on their own. Even with perfect prompting, LLMs can muddle sentences, exaggerate facts, and, often, make things up, making it all the more important for humans to oversee their output.

Online, startup employees have raged about colleagues who simply ‘copy paste’ output from chatbots without reading through or editing it first.

Victor Riparbelli, cofounder of AI avatars startup Synthesia, sent a Slack message to his employees in June after noticing AI-slop was sneaking into company documents. While he hasn’t banned the use of LLMs in writing, he’s asked employees to use them more intentionally to avoid wasting time.

“When we write too long and verbose documents we place the burden of distilling the content onto the readers, so instead of one person spending ten mins on sharpening a doc, we have ten people spending ten additional minutes reading slop, leading to net-negative productivity,” he wrote.

Beyond comprehension issues, another commonly cited problem of using AI to write is it can erode individual personality and voice, and make everything sound the same.

A recent piece by Harvard Business Review highlighted CEOs risk “flattening” their leadership style by using AI to communicate, losing employee trust in the process.

Jones thinks the risk is broader, potentially impacting a team’s ability to connect with one another.

“One of our values is dedication to craft and that is reflected in our app and in its design, but also in how we behave,” says Jones. “How you communicate is also your craft as a team member, as a leader and a manager. And when you outsource that to an LLM that doesn’t know you, or your team, or the company, you’re losing so much.”

For example, Jones says her Dutch colleague used the idiom “cucumber time” in a recent conversation, referring to the quiet summer months when everyone is on holiday and the news cycle is slower — something Jones hadn’t heard before.

“It [brought] such a lovely moment of connection, and you miss all that when you screen out all the quirks of language that make us different,” she says.

A new industry standard?

There are some benefits of using AI to communicate in ways other than text.

Large language models can present information in varying formats to suit the individual’s needs. Jones recently had to present an annual review to a colleague who isn’t fond of reading; she recorded herself speaking and used an LLM to make it into a podcast.

Voice AI also enables team members to talk through an idea for a product or sales strategy aloud, while the AI takes notes for them.

“That’s the lovely thing. There are loads of ways you can use AI to be more inclusive to help people get context and information in the way that works best for them,” says Jones.

Laura Gonzalez Florez, chief of staff at Synthesia, agrees, suggesting that “a concise summary, a translated explanation, a video walkthrough, or an interactive experience where employees can ask questions or practise a scenario,” can often facilitate better understanding among colleagues and a “measurable improvement in how people work.”

But regulating the use of AI in writing is likely to persist as people tire of slop. The tech industry itself is moving to set new standards.

This month, Anthropic announced it intends to watermark text generated by its AI models in line with the EU AI Act’s Transparency Code, and LinkedIn launched a “Seems like AI slop” button. Substack has also partnered up with Pangram to detect AI-generated content in posts and articles, a move which sparked controversy among some users of the platform.

Sarah Drinkwater, a solo general partner at VC firm Common Magic, says AI writing policies are necessary for startups as they help team members distinguish where generating content is appropriate, and where it “creates slop, wastes time and leads to unoriginal thinking.”

“For startups creating cultures that allow them to ship and learn at speed, while creating distinctive products and edge, this deeply matters,” she says.

“I think we’ll see many organisations, from companies outside tech to schools and governments, adopt AI writing policies to set their parameters and create trust internally and externally.”

This article was also published on Sifted and mentioned in its Startup Life newsletter. Subscribe here.

Today’s platforms feel far from gentle. The endless scroll features and algorithms — designed to maximise user engagement — pushing us to consume ever more content and click on ever more ads is tiring to say the least.

Nowadays, I try (and often fail) to steer clear of social media, or at least moderate my consumption — and I’m not alone. Aware of the effects being “too online” can have on our mental health and self-esteem, some are drifting to slower platforms (Myspace, much to my delight, is making a comeback, according to Vogue) or are finding new ways to stay informed off-screen.

The concern over the harms caused by social media has led governments around the world to take drastic action. In December 2025, Australia became the first country to ban social media. France passed a law banning under 15s from platforms in July, but it has since been overruled, and the UK is also consulting on a ban for under 16s. Some critics argue complete abstinence doesn’t solve the underlying problem.

It strikes me that all media — be it news, Netflix or shopping apps — can be just as addictive and overwhelming for the senses as Instagram or TikTok.

To find out more about how our digital consumption affects our wellbeing, I reached out to Lucia Komljen, a strategist, sociotechnical researcher and founder of Have A Nice Future, an advisory practice focused on responsible media innovation.

We discuss the problems with online platforms, the rise of AI-generated media, and why social media detoxes aren’t always effective. We also get into Lucia’s new project, Media Rehab, a programme helping people understand their digital diets and rebuild them in a more positive way.

The interview has been condensed for length and clarity.

If you could pinpoint one or two things, what developments do you see in the current media landscape that are potentially problematic or not supportive of our wellbeing?

One of the things that has concerned me for quite some time is that media and the platforms on which they live — whether that’s entertainment, information, social, but also what I call functional media like your real estate apps, shopping apps, data apps, as well as media advertising — all of that has been overwhelmingly engineered for profit. And to take it a step further, it’s in part been engineered to divide, to addict and to sow chaos.

The maximum time spent and the quantity of our consumption is very much at odds with what benefits us on a sensory level, on an emotional level, on a mental and cognitive level. Essentially, the mechanisms that underpin these platforms are not very human-friendly.

One of the key mechanisms that is currently very anti-human is the way recommendation algorithms1 work and who they serve. Now, we’re in a new media cycle where we have the combination of recommendation algorithms with synthetic media. It worries me again that media will continue to serve those creating it, rather than those receiving it.

There is a long list of existing harms based on the impacts of the current media cycle, such as shared reality erosion, and the exploitation of our attention and emotions. There’s also the exploitation of financial means: we’re being sold to everywhere, right?

Sadly, it feels like media inequality is also going to deepen. Media that delivers quality, meaning and factually accurate information is increasingly being priced at a premium, only there for those who can pay for it.

What does synthetic media mean — and why might this form of media be particularly nefarious?

Synthetic media is media that has been at least partially or completely generated by AI. Ages ago I identified six categories of synthetic media — there’s probably about twelve now.

The dominant use case for generative AI is synthetic ‘slop’ media — cheaply engineered for irresistibility — and ‘nerve agents’, a kind of evolution of misinformation.

One of the risks of synthetic data is the sheer volume of media being produced. There are questions around the sustainability of the storage of this media and the resource cost of that.

Then there’s cultural concerns. This is not the first time we’ve had tools to help us create media, such as photo, audio, or anything that makes production faster, cheaper and easier, and it’s not necessarily always a bad thing. But there are a lot of skilled, imaginative creators who think the process is what ultimately creates value and meaning.

The risk of more creators — whether that’s bedroom creators or a legacy business in the media industry — using this technology is what Kyle Chayka from the New Yorker calls 'the ‘flattening of culture.’ It’s linked to the debate of whether output is really original — and if you are actually creating new culture — if you work with large language models.

The third risk is: if we think about the way that so-called fake news or disinformation or adversarial news has behaved to date — with AI, you have the opportunity to supercharge that. Things can be tailored to push very specific individuals’ emotional buttons. The intensity is just being dialled up — and the more intense the message, the more intense the emotional response.

I love your metaphor of the problems with the media landscape being akin to dietary challenges — media can either nourish or malnourish us. Can you unpack that for us?

The food analogy likens our consumption of media to how the food industry behaves, to how specific foods behave, and the impacts of that consumption. That’s obviously not new.

If you look at our history of using this analogy… I think starting in the 80s, there was a sociologist called Carl Jensen who was the first to use the term “junk food news.” You’re probably also familiar with the metaphors of coach potatoes and TV junkies which have always kind of been there.

The analogy today is still predominantly negative. We refer to media as junk or slop, or as a guilty pleasure. So I thought about how to balance that by expanding the vocabulary to include some positive metaphors.

I think the risk of using negative labels to describe media consumption is that we end up doing what we did with food: we perceive it as demonic, for example. The prescribed solution is always abstinence — but abstinence (via detoxes or social media bans) can only get you so far.

I’d like to dig into the idea of social media abstinence a little later on. Before I do, can I ask what you consider to be an ‘unhealthy’ media diet?

That’s a very individual thing, and something I was keen from the outset to not stake any claims around. We’re all very complex beings so there is no black and white, ‘this is good’ vs ‘this is bad.’ I suppose what I encourage people to do is to understand their diet across three dimensions:

• How is what you’re consuming making you feel now — and what is the long tail of that feeling?

• What are the functional benefits you are getting from your media? (I have a taxonomy of six categories, which includes pleasure or winding down)

• How is your diet aligned or misaligned with your daily routines?

Then, they can make their own decisions around what to do about it.

Many of us are aware that phones and social media apps can be addicting. But even though we have that awareness, we’re all still kind of hooked. Why is it so difficult for many of us to regulate our consumption?

The interesting thing on this particular topic has been that what people feel about their media consumption doesn’t match what some evidence suggests.

If you ask the people who self-reported, the majority of them feel that they’re addicted either to a device or a particular app or a particular genre of media. But there’s a substantial camp of scientists who will say these people’s behaviour doesn’t actually look like addiction when you compare it to something like substance abuse.

To your question of why can’t we shake it? I think there just haven’t been the right kind of remedies in place — and again, different things will work for different people. All we’ve been given to date have been detoxes, apps that block other apps, and now there’s even a physical device, ‘Brick,’ to block you from apps.

Why do you think abstinence from devices is not always the right remedy?

I don’t think we’re properly tuned into the nature of that addiction and what it is specifically that addicts.

Again, to think of parallels in the food space. If you go to a nutritionist because you have a particular ailment, you go on a journey with your nutrition in order to remedy it. And that journey takes you inward: you track your food for a week and how it makes you feel, and then you start experimenting with eliminating certain foods. And then after that period of diagnosis, you move into therapeutics. We don’t have the equivalent for social media.

When it comes to detox programmes in particular, those tend to be delivered as ideas: ‘here’s the five steps to do it.’ There’s been research showing sleep can improve during a detox, but recent papers have also shown loneliness can creep up, as well as a sense of unhappiness because you’re no longer in touch with your interests. The way that certain aspects of social media can bond us with each other — that falls away. So even detoxes, whose sustained effects are very minimal, have their challenges.

You’ve developed your own programme to help people understand and improve their digital diets called Media Rehab. Can you explain the core idea behind it?

Media Rehab is currently a programme, although I’m exploring different iterations of it at the moment. Over twenty one days, participants unpack their diet, scrutinise what’s in it and then rebuild it in a way that serves them. They explore these three dimensions mentioned earlier: how media makes me feel, how it serves me functionally, and how it aligns with my daily rhythms.

I think there’s some really good stuff on these platforms; we just have to get better at understanding what it does to us, and what benefits us and what doesn’t. Then, you can adjust your diet based on what your mental intolerances are. For example, if you get enraged very quickly by watching posts about terrible things happening on the other side of the world, can you find another way of staying informed?

It’s the same with food allergies: if you’re lactose intolerant, you can take a Lactase and still eat icecream. But if the afflictions are severe, there’s probably a strong case for reducing the amounts of that particular food, or removing it altogether. It’s about using platforms with awareness.

What have you seen or learned from the changes people experienced during the programme?

I think one thing that has been universally appreciated is that it’s not a black and white ‘switch all this stuff off because it’s terrible and here’s the reasons why.’ It allows for you to find joy again in what we’re increasingly told is a polluted ecosystem. There are still some unspoiled fruits.

The other thing people seem to appreciate too is this idea of, once you understand what your diet looks like and how it makes you feel, you make commitments around what to keep, what to reduce and what to move towards.

I had one guy participate who lives in a city in the United States where ICE (Immigration and Customs Enforcement) was very present. He was describing to me that he tended to sit on a bus, scroll through Instagram and just see ICE do terrible things, while also looking out of his window and seeing ICE do terrible things. He said: I just can’t do that anymore, it’s really bad for my mental health.

He said that finding more neutral sources for news outside of social media, and dosing and scheduling his time spent on platforms made a significant impact, even three months after he participated. It’s interesting to see how people are making their way back to public service media in the States like Ground News, and now PBS and NPR.

Some people also said that they’d never considered this category of functional media as part of their media diets. You know, the sitting in bed and looking at houses that you can’t afford, spending multiple hours a day tuning into a dating app to find out you were ghosted, or also shopping media. Every platform behaves like Netflix these days — there’s so much gorgeous video to sell, sell, sell. To realise these have an impact too was something they found quite valuable.

There’s political momentum in the EU to hold platforms accountable for addictive design, and other potentially harmful mechanisms. How far do you think responsibility lies with individuals to manage their media diets and how much lies with platforms and with the designers of technology?

Ideally, we’d have a situation where the onus to the extent that it is today wouldn’t be on individuals. As you mentioned, we are seeing some progress in light of the whole raft of lawsuits against US companies2, and there are a lot of initiatives at EU level where design issues will start to be addressed.

Can we really trust those platforms to address them to the extent that we have perfect non-consumer harm conditions? Probably not. But let’s be optimistic that these initiatives will start to reduce the onus on people to manage it completely on their own.

In the UK, there’s also a bigger drive towards media literacy. It’s now being adopted into the school curriculum in 2028 — but I think adults are still completely underserved.

Overall, the initiatives that are happening both at a company and a policy level are just not enough yet. It’s very piecemeal.

With regards to media literacy for adults, what would you like to see happen?

I’m actually starting a Masters in Media and Communications Governance to dig into this properly! One idea I want to explore is using nutrition as a model, looking at how it’s historically been presented in policy and pop culture and what lessons we can borrow.

The diet / nutrition analogy isn’t new, but people treat it very superficially. My research to date suggests there is much more to unearth and adapt.

What makes nutrition a useful model is that, unlike media literacy (usually a one-off learning experience), nutrition is part of people’s daily decision-making process. It covers both preventing problems and building good habits, adapts to what people can actually afford or access, and comes in different forms — medical, cultural, ideological — that meet people where they already are.

Subscribe now

I don’t blame her for being confused at my outreach. I introduce myself as a technology journalist reporting on AI — what does that have to do with writing?

In my mind, at least, it has everything to do with it. At a time when the world feels like it’s being engulfed by AI and many aspects of society, from education to the workplace, are changing with speed, the practice of reading and writing can keep us grounded and connected to each other.

I reached out to Madhvi following the launch of her new children’s book Move Your Body!, which celebrates movement as a means of creative expression. The timeline from writing to publishing took two years, even though the book is just 300 words. It was a reminder of how crafting something worthwhile takes time.

Madhvi is a fiction writer, essayist, playwright, as well as a performer and creative facilitator. She’s part of a South Asian spoken-word performance group called Sluts in Sarees, runs a Feminist Book Club in Berlin and frequently hosts writing retreats, most recently on a farm in Brandenburg.

Sitting outside the cafe on a swing chair in the pouring rain, mugs of tea in hand, Madhvi and I talked about her early memories of writing, why it’s important to keep children reading in the digital age, and why literature is a powerful tool to help us imagine better alternatives for society.

The interview has been condensed for length and clarity.

How did you get into writing? Was it something that started in childhood, or did you come to it later in life?

I think I always liked writing. I was always a really quiet child. I see that with a lot of artists, like musicians and writers: they were really quiet and somehow writing became an outlet to express themselves in some way. But it’s a very solitary kind of outlet.

I think it also came from the fact that my mother was an immigrant in the UK and English was her second language. She was a single mother, so she went out to work, and part of her job was to take minutes. She’d write them first in Gujarati and would then translate them into English — but she was always really insecure about speaking English.

I think that made me realise that language is so powerful. It could make the difference between what you could communicate or not, what was heard and what wasn’t.

Do you have any early memories of the first things you wrote?

I started with creative writing — I still mostly do creative writing, and I used to disappear off into my own fantasy world a lot. With my mother as a single parent, she didn’t have much time to entertain me, so I had a lot of time to myself to build my own worlds. I started writing a lot of stories about the moon and pirates.

I have an Irish friend who is an actress, and as a child, she used to put on religious masses with her siblings, and that’s how she started acting, devising and directing.

I guess that’s how children learn to understand the world to some extent, through acting out the real world in play. You’ve written various books for children, and recently published Move Your Body!, a story about the joy of moving in your own way. Why did children’s books become an important medium for you?

Oh, this is going to be a long answer! Like I said, I think I still feel connected to what was important in my childhood, and books were an escape. They were also a way to learn. They were also a way to see yourself doing different things.

I think it’s also important to have books with diverse characters that other children can see themselves in. I did a workshop last week and a woman there, her daughter is Black and wears her hair in braids, which she doesn’t like because none of the other kids in her Kita1 have them. She said she read my book and then wanted to have her hair in two braids, like one of the characters in the book. I think that’s really important for confidence.

My book before this one was called Whisper, Shout: Let it Out, which is about expression and having the confidence to get your voice out there and be part of the world. This probably relates to AI somewhat: there’s very confident people, mostly white men, that totally dominate the conversation, and there are some voices that don’t get heard.2 You have to be connected to your own voice and confident enough to know that it matters.

I recently ran a workshop for women who work in business and, statistically, women talk less in meetings or they get interrupted more, they get questioned more and so on. I was doing a few creative techniques to help women use their voices at work. There were women who were very accomplished who had all the privileges in the world and they still had very critical inner voices. I kind of felt like it’s almost too late at that point when we’re adults to build this confidence; we have to start way younger.

How do you weave those messages into the storylines?

Move Your Body! is more like a workshop book, so you get promised to get taken on a journey. The premise is that moving is magic: when we move, we can become anything.

The reader is taken through a series of different elements, like water and fire, where the child is asked to make a decision for themselves. Like, what sea creature do you want to be? Do you want to be a seal? Do you want to be a dolphin? A hummingbird? What’s the movement you want to do to portray that creature? And the reader does that movement in the moment.

It’s the idea that everyone can choose their own movement. You give them a choice, so they have to connect to their own body and their own self to decide.

It’s interesting to talk about this in the context of AI. Speaking to parents and educators, many are concerned about the world that their children are growing up in, and how to prepare them for it. There’s a lot of discussion on how to teach things like resilience, empathy and judgement. Do you think that reading can be a medium through which to develop these skills?

Totally, I think empathy and reading go together. With the book I just mentioned about movement, and also the book before that about voice, these are embodied actions. AI is not embodied. When you use AI, you’re often in your head, typing, looking at a screen. Movement allows you to connect to yourself in different ways.

Another concern is about keeping children reading. Research shows reading literacy in some developed nations, including Germany where we sit, is declining. What are your thoughts on this?

There’s a literacy crisis right now. I think the publishing industry is basically competing against YouTube. There’s really great stuff on YouTube and there’s great apps out there for learning, but that just means that less reading is happening.

I think parents putting books first is really important, because that reading habit has to be built. We’re not naturally readers — if I stop reading for a few months, it’s hard for me to get back into a book because it requires a certain amount of focus and concentration. Reading has to happen early to show how valuable it is in understanding the world around us.

Speaking of AI, you can’t use it to write children’s books because there are certain innate, ancient patterns of rhythm and sound that require a lot of brain work to write. I’m currently writing a book about a girl that lost her balloon. On one side, I’m telling the story, and on the other side, I’m rigorously working with rhythm and sound to make it flow. It’s almost mathematical.

That’s interesting — it’s kind of like the words have their own underlying beat and resonance to them. What function does rhythm have in children’s books?

I don’t know much about the origin of language, but we come from an oral culture where stories were told and passed down to the next generation; a lot of things were recited over and over again. We remember things through rhythm and song; it’s also how we speak to babies.

Our bodies also remember rhythm. There’s something very innate about that: language is intellectual in that you have to know the words in order to understand the concept, but it’s also embodied.3

You recently said on an Instagram post: “I realise more and more how sacred ideas, expression, and words are to humanity.” Why is this particularly important in the world today?

Words have lost their connection to reality somehow. I think we have a situation politically… like, people have worked on the meanings of words or concepts for a long time in academia — terms like woke, or fascism or racism — and a lot of historical research has been done. There’s been a real deepening of human knowledge. And now, all of a sudden, it’s just being emptied out. Certain words and phrases are being used in the wrong context and their meanings are being turned around.

The only thing that gives words meaning is how much respect we give them. I think we have to stay connected with our communities, with our history, with the people who came before us, no matter who they were, and to understand things in a context that kind of makes sense. Because now it’s just turning into nonsense.

I think Naomi Klein elaborates on this topic most articulately in Doppelganger. She says: “What if our books, and our movements as they are currently constructed (often in ways that resemble corporate brands), are only changing words? What if words — written on the page or shouted in protest — change only what people and institutions say, and not what they do?”

I feel that words being taken out of context and blown up is something that can happen a lot on social media. Now, there’s also a lot of AI-generated content and slop, which is often empty of meaning. It feels like a really crucial time to acknowledge that words have weight.

I think so too. I’ve been running a feminist book club in Berlin for three years. Anyone can join as long as they’re a woman. We read a lot of bell hooks and Angela Davis — I think it’s great that people are reading quotes from their books and sharing them on social media, but it’s really important to engage with the books in their entirety.

We also recently read The Sex Lives of African Women by Nana Darkoa Sekyiamah, which features stories of female sexual experiences and desire from across the African continent and its global diaspora.

Sometimes we collaborate with bookshops or other groups, but mostly we’re meeting in people’s private homes and just talking — reading and talking about these books and what we think about them.

What I love about book clubs is that someone inevitably tells a really personal story linking back to what they’re reading, and then the discussion sometimes ends up not being about the book in the end at all. It can be very moving.

Yes, at some point when nobody’s talking about the book anymore, we’re just talking to each other. I was recently reading Susan Sontag’s On Women, in which she asks: where are the women all talking to each other? We have a lot of differences. We have a lot of things to sort out. We have to be able to listen to each other — and a book is a great way to do that.

Last time, we read Girl on Girl4 by Sophie Gilbert and a woman in the bookclub who is from Syria said that none of this applies in Syria. I was like: ‘wow, all of it applies to my childhood.’

So it’s very interesting to see how we can build up solidarity and conversations and understand where different people are coming from so that we can have a feminist movement. What happened to the feminist movement where women were talking to each other and organising? It seems like it’s there, but is it there?

I love the community-making effect that reading can have. Sometimes, the way I connect with friends and family of mine back home in the UK is through discussing interesting books or articles we’ve read. I think there’s ways in which we can cross physical and emotional boundaries by being able to talk about books.

I think for sure, if we’re both discussing the same book and we disagree about it, we’re still on the same page somehow. It’s great because it forms a basis for conversation.

I think it’s also really important in the Feminist Book Club because, I mean, we’re in Berlin and we’re totally in a privileged bubble, but I think a lot of women feel like they’re alone when talking to their families, or whoever. And just to be able to meet over a book and to know that there are other people who share your views — it’s really nice.

For me, it also opens up other worlds, you know? I just read Butter by Asako Yuzuki, which is about feminism and food in Japanese culture. It made me see how we’re all connected and our emotions are the same, it’s just the worlds we inhabit or the possibilities that we have that are different.

Books open up the power of imagination. We need imagination to come up with different ideas, fantasies or possibilities for the future. There are people who are like: if not capitalism, then what else? Or, there’s public discussions about where the money is going to come from to finance our healthcare systems. Are we really saying that we can’t think our way out of it?

This is a really powerful way of explaining why I personally think reading will endure, even as people increasingly consume information digitally, and AI tools are being used to generate text, or condense information into bite-sized pieces. When we read fiction, we don’t just read a story: we understand the history and cultural context in which a book is set, we build a relationship with characters who harbour a range of emotions and experiences, and we observe the successes and failures of the past that serve as lessons for the future. Reading helps us think, to follow an argument over time and form our own opinions. We’ll need various ideas from various brains to help society adjust to the changes brought about by AI.

Thank you Madhvi!

Subscribe now

📚 Madhvi’s book recommendations:

• Doppleganger: A Trip into the Mirror World, by Naomi Klein. The best explanation of our current times.

• The Word for World Is Forest, by Ursula K. Le Guin. A reverie on colonisation and consciousness.

• And if you’re interested in how our breath, rhythm, sound and our bodies are connected to words, line and poetry: Rules for the Dance by Mary Oliver and
  Momo,by Michael Enders. I love how children’s books can tackle massive topics and be emotional and magical at the same time — this one is about capitalism!

On the subject of reading:

A New York Times article published last week shows how books can help us to process painful emotions, and understand societal changes.

In Ukraine, many children’s books have been published since Russia invaded the country aimed at helping children and their parents cope with the trauma of war. The books cover a range of themes, including loss and the terror of bombing. They also explain in child-friendly terms how the war is reshaping everyday life and society in Ukraine. Well worth a read.

I’m happy it led me to Jena Wuu’s Substack, Feed For Thought, which is all about how expertise and trust are evolving in the digital age.

Jena has worked in tech and public policy for 15 years, and formerly served as head of global information ecosystem programs at Meta, a role focused on improving access to trustworthy information during big world events such as COVID-19 and the war in Ukraine.

What initially drew me to Jena is her thinking on the efficacy of content moderation in stemming the tide of misinformation on social media. Platforms have faced backlash at various points over the years for failing to protect users from misleading or harmful content, but Jena argues that pressuring platforms to moderate content better is only part of the solution. The other part — getting enough credible information into online spaces — is often overlooked.

This was a wide-ranging conversation, and I learned a lot during the hour we spent together. We talked about trust in the digital age, why institutions need to be more present online, and how individual actions such as liking and commenting, though seemingly unimportant, are powerful in shaping the online ecosystem.

The interview has been condensed for length and clarity.

Can you kick us off by explaining what’s meant by an information ecosystem and who the various actors are shaping public understanding today?

The information ecosystem is essentially anywhere where somebody consumes media. I don’t see a distinction between entertainment or something like news, because I think that every piece of media that you consume informs your understanding and perception of the world.

So that can be anything from TV, films, radio, newspapers to digital platforms like Facebook, Instagram, TikTok. I also include things like WhatsApp groups, Discord and Slack groups too — anywhere where information flows.

A friend of mine recently asked me what the difference is between media and content, but I actually think they’re the same thing. It’s just the connotation is different. Media sounds like something official, from an institution, whereas content is peer-to-peer, and often associated with influencers or social platforms.

You’ve argued on your Substack that authority is no longer necessarily owned by institutions but is earned in various spaces on the internet. Can you explain more about that shift and how it’s changed how people decide who or what to trust?

Authority is basically a mixture of expertise and trust. Previously, if you had expertise it was connoted by your affiliation with an institution, whether that’s a university or media outlet. Expertise is essentially knowledge learned through education, training and experience. Trust is earned through somebody believing that you have their interests in mind and are acting in goodwill.

These two things used to be combined and that’s what created authority. But now, these things are earned separately — having credentials alone is not enough to earn somebody’s trust, it’s just one factor. Other factors would be: do you seem relatable to me? Does it seem like you’re experiencing something I’m going through? Do I know you or are you familiar to me? Do other people I trust recommend you? My hypothesis is that institutions have so long relied on who they are to have trust that they haven’t really invested in earning it.

Reuters Institute’s 2026 Digital News Report shows that trust towards news is at its lowest point since 2015. Meanwhile, half of survey respondents get their news from creators and, increasingly, people are turning to AI chatbots for news. Some might refer to this as a crisis of trust. What do you think about that?

It could be a crisis of trust or it could be an opportunity for people to really examine who they trust and why. In the past we would just trust institutions because they were the institution, but now there have been challenges with institutions and legacy media for a long time. I think that it actually could be an opportunity for people to interrogate the decisions they make and how they participate on online platforms.

In your role at Facebook, you devised a program to address misinformation and disinformation. What did that involve, and what challenges did you face?

There’s a difference between disinformation and misinformation. Disinformation is false information that is purposely perpetuated for nefarious reasons. Misinformation is just false information that’s shared without necessarily a negative intent.

At Facebook, I created an information ecosystem program out of recognition that there were some gaps in the way the company, but also policymakers, were thinking about the “misinformation problem.” There was so much focus on removing bad content. The whole industry has invested a lot in third party fact-checking, which is an important part of the ecosystem, but the whole thing is a balancing act.

While content moderation is important, it’s not enough by itself. It relies on misinformation to surface and get some kind of circulation to begin with for it to get removed. But it also relies on an assumption that there is sufficient credible information online that’s there to take its place — and that is simply not accurate.

At the scale that digital platforms operate — and that people are using these platforms to look for information across different cultural contexts and languages — we cannot rely on journalists alone to fill this gap.

So the emphasis has been placed on content moderation, when we should also be looking at the other side of the equation: getting more credible information on the internet?

Right, and I think institutions whose objective it is to produce knowledge for the purpose of public benefit need to communicate that knowledge effectively.

What I found at Meta was… When Covid-19 broke out in December 2019, Mark Zuckerberg made this commitment to the World Health Organisation to provide as many free ads as it needed to disseminate accurate information about Covid-19, public health and preventive health measures. We then quickly expanded it to national health ministries because the pandemic was a global issue, but the response was being managed on the national or local levels.

The original assumption we made was that the institutions lacked the budget to buy ads, and so it was supposed to be an ads donation program. But what we quickly learned was that 80% of the public health agencies we were working with had never used social media for the purpose of spreading public information; they’d only really viewed it as a marketing or fundraising tool.

To me, that was really shocking: for as much focus as there is on these platforms as sources of misinformation, there was very little focus on them as sources of information. So we ended up expanding the programme to help institutions create Facebook accounts, and teach them how to create good content to share credible information.

There’s a lot of distrust for platforms which is perfectly understandable. But if we say, to use Facebook as an example, that it’s Mark Zuckerberg’s responsibility to moderate content, then we’re also giving him the power to decide what is credible information or not. We’re basically giving him permission to create Facebook in his visage, like Elon Musk has done with X. So, the question is: do we actually want to give that power to platforms, or not?

You’ve said that part of the problem with getting enough credible information online is that a lot of institutions aren’t present in the online spaces where a lot of information is created and shared. Where do you see that absence having the biggest consequences?

I think the most acute examples are in health, science and in a lot of the conversation about climate. People are going to trust what’s familiar to them.

It’s really interesting to look at how brands have built consumers’ trust; they’ve invested a lot in figuring it out. I think there’s a lot of lessons we could take from the private sector into trust-building in the public sector.

For example, the marketing rule of seven states that a customer needs an interaction with your brand seven times to recognise your brand and to internalise your message. A successful campaign at Meta would be two impressions — the number of times your ad was displayed on a user’s screen — per week for four weeks which is about the same as this rule of seven.

For me it’s about trying to change the culture within institutions so that they feel that participating online is a safe and legitimate way to build trust.

What are the barriers to getting experts like scientists, researchers and doctors to participate more effectively online and how might those be overcome?

Most doctors and professors and public servants who work within very traditionally hierarchical work environments are not going to become full-time content creators. One, it takes a lot of time to make thoughtful content. And two, being a spokesperson is often delegated to comms teams or reserved for leaders or designated spokespeople.

In that kind of work environment, the implication to experts is that public communication is someone else’s job. And when organisations lack a clear policy on how individual employees can talk about their expertise or work, two things can happen. First, many won’t speak or write publicly about their work, either out of fear that it’s not condoned by their employer or because they don’t view it as part of their scope of work. Second, those who do engage publicly may cause situations that pose risk to the institution, like if they misrepresent its position on a topic or engage in inappropriate behaviour.

That’s why I advise every institution to create a clear policy for how their staff can talk about their work online. It gives permission and acknowledges online engagement as normal professional behaviour, while also building frameworks and recourse that protect the institution.

I think providing training to enable people to learn about the information ecosystem is also helpful. For example, some doctors I’ve spoken to feel very frustrated that their patient doesn’t want to vaccinate their baby because they were listening to some kind of wellness influencer. When I asked them what kind of training they get on digital communications, or why they think people would trust an influencer rather than their doctor, they didn’t know, because they don’t get any kind of education on this.

This has such big ramifications on their ability to do their work. Public understanding and public sentiment impacts public policy, it impacts funding decisions and shapes whose expertise gets taken seriously.

So I think it starts with recognising that digital literacy and an understanding of the information ecosystem are core professional skills for any public servant. It needs to be incorporated into graduate school programs and medical programs as part of professional development, alongside content skills training.

What are the different ways in which individuals can intentionally participate to improve information quality on the internet?

People often understand their role in the ecosystem as binary: you’re either in front of the camera or you’re in the audience. But now, in the digital world, all of our roles are much more fluid than that, and participation exists on a spectrum. Even people who never post content shape which creators and content get amplified through their actions online.

Participating doesn’t mean you have to spend more time online or become a full-time creator; it’s about what platforms you use and how you engage with them. For example, Meta counts watching a video for more than three seconds as ‘watched’ — that’s a signal to the platform that the content is worth sharing with other people.

Every action has different values. Liking, sharing, commenting are all high value signals to platforms because you’re generating new content that’s creating more engagement opportunities with that post. So, if you are following someone who is making content you find useful, engage with it.

You may not view yourself as a creator, but you’re still contributing to the ecosystem, even if you’re not creating content yourself.

So for you, it’s all about empowering the individual in the end?

Yeah, and reframing what participation is. There’s so much conversation right now about technology and AI being forced on us. I really want people to feel like this is not something that’s happening to you — you have a lot of power and control over your relationship with these platforms.

Also, it’s perfectly legitimate to be spending time on these platforms; there’s a lot of value that can be derived from them.

Generative AI is already tilting the balance of the information ecosystem, making it harder to discern between what’s real and what’s not. But it’s also provided creators with new ways of creating content. What risks and opportunities do you see there?

AI is for sure disrupting the creative class and how people create content. Like with food or fashion, I think we’re going to end up with tiers of content. We’re going to have fast content which is cheap to make, easily consumed and quickly discarded, and we’re going to have slow content — that is more like an artisan craft with an original perspective, but is more demanding to produce and consume. I think there is legitimate space for both.

There’s a lot of snobbery about where people are getting their information. I hear a lot of people lamenting about how people are getting their news or getting their research from Facebook or from Instagram. And I understand that. But at the same time, there are other forces at play here.

We don’t work in a time anymore where everybody can be home at 6pm for dinner with the family and watch the evening news. People are working multiple shifts. We’re busier than ever. And people are having to get their information in small snippets while they’re on a break between shifts or when they’re in the bathroom.

There’s also a major literacy issue in a lot of countries. In the US, over half of American adults read below a sixth-grade level and nearly a fifth of adults in Germany read at literacy level 1, which means they’re able to read short pieces on familiar topics.

When we’re dealing with these bigger structural challenges, it’s not very productive to shame people for engaging in content that’s simpler when they just don’t have the time, the mental energy, or the literacy skills to be able to engage with lengthier or more complicated things.

However, it’s really on the institutions and the experts who care about serving the public to make sure that these different content classes are not resulting in different information classes. We have a responsibility to ensure information equity for everyone.

I believe that information is life-saving. If institutions and experts remain absent or passive on digital platforms, it actually causes harm because it makes credible information harder to find by the people who need it. That goes against the purpose of entities and professions that exist to serve the public.

“I learned … that most of us are a living, breathing soup of memory and imagination — and that we may not be the best arbiters of which is which. So read this book as you would a novel. It makes no larger claim. But then there can be no larger claim. Fiction is that strange smoky thing that writers don’t entirely own, even if they think they do. Where does it come from? Our past, our present, our reading, our imagination - yes. But perhaps from premonitions of our future too?”

— Arundhati Roy, in Mother Mary Comes To Me

1. Standing at the underground station waiting for the train to arrive and resisting checking my phone. Yes, it’s warm, it’s loud, and I long to listen to music to drown out the sound of children squealing, dogs barking and machinery clanking. But if these are the precious five minutes I have to myself today just to think my own thoughts, I’m going to take them. The empty space is often where my best ideas flow.

2. Wandering into a bookshop to prevent the decision overwhelm of searching for a book in a large space with several floors and sections. I see a book cover catch my eye and stop to read the blurb. People always say they love the old smell of books. I pick up a book and put it down again — maybe I’ll come back to it later, or another day. I find a memoir my friend recommended that made her cry. I think of those human stories that pierce your heart in places you thought were healed — that remind you of life’s light and shade and the grey, uncomfortable space in between. I buy the book and start reading it on the train home.

3. Rawdogging a walk in your neighbourhood without your phone is underrated. I walk the same walk I’ve walked a thousand times. There’s a small woodland nearby in between housing estates which merges into an old graveyard. Vines curl and wind their way over the gravestones; the names of the dead barely recognisable. Trees line the pathways creating a lush green hue overhead. I breathe in the summer air. A big German property developer wants to bulldoze over this little haven, where people walk their dogs and children play, but the community is fighting back. The local Green Party has taken up the issue with the European Union as there is endangered wildlife living in the wood. I like to listen to the sound of the birds here. It gives me a sense of safety, anchoring me in time and place. Bird song is scientifically proven to lower cortisol in humans and shift our nervous systems out of fight or flight. I try to remember this feeling of calmness as I ponder the future of the forest. Can it prevail over the powerful claw of capitalism?

4. Procrastinating. It feels defiant in the technological age, where every part of our day is optimised to reduce time-wasting, do more work and, I say cynically, maximise profit. I procrastinate big cognitive tasks I have to do on my laptop by washing dishes, doing laundry, and attending to the plants on my balcony. There’s safety in the mundane tasks of ordinary life.

5. Writing — and maybe in a notepad rather than on a laptop to not disrupt my flow. As I write this, Grammarly, which I have installed on my computer to check for spelling mistakes and typos, puts a big blue line underneath certain colloquial turns of phrase that haven’t made it into its dictionary: “decision overwhelm”; “rawdog”. I won’t be removing them. Writing in a notepad unlocks another level of thinking; like a trap door opening in my brain into another universe. The corporeal act of picking up a pen and writing ink into the page is at once more intimate than typing on keys.

6. Being spontaneous. I’m not good at it. I try to resist the urge to plan every minute of time into my calendar and set endless digital reminders. Items on my to-do-list pad careen off the page. My little brother told me on the phone recently to stop being in such a hurry with life: “I think you need to slow down a little,” he said. But what does it mean to slow down, I ask? How can you slow down when the world around you is spinning at dizzying speed?

7. Cycling through the city (without headphones). It’s the best method of transport for watching the world go by. I stare as two men walk through traffic carrying a large, rolled-up rug on their shoulders, while the cars beep in irritation. I laugh at the absurdity. A little girl with a pineapple of blonde hair on her head loses the top of her yellow ice-cream to the pavement; her mother gently consoles her. On the other side of the street, people queue in a line that snakes down the street for falafel wraps glistening with tahini sauce. Business as usual in Berlin Neukölln.

8. Listening to an album all the way through. I forgot how much I love this, now that I mostly stream music on Spotify, and make playlists - a patchwork quilt of songs. When you listen to an album start to finish, it reveals the story the artist is trying to tell; the intention they had in putting the songs in that specific order. The imagery, colours and shapes of feeling.

9. Singing in a choir, learning music by heart in languages such as Bulgarian, Polish, Thai, French, and Latin. The best tool I have for preserving my memory. A member of the choir uses the words and sounds of the song to make up stories to help others memorise it. They’re funny, light-hearted, mostly random — and they work a treat. When I sing, I’m away from my phone, listening only to the sound of the people next to me harmonising in unison. The mind wanders free.

10. Staring blankly into space. No, really. It’s the only thing to do when my brain won’t stop whirring. A moment spent looking out of the window, observing the world, is a forgotten pastime. Maybe people think I’m weird — but the quiet is a privilege these days. It’s the last vestige of a time when our heads weren’t pounded daily by notifications and attention-grabbing headlines; where we didn’t wake up every day thinking about AI (maybe that’s just me?) or how to raise children — or even oneself — in a world where we’ve all forgotten how to just sit in silence for a while.

11. Thinking of stories that lead to nowhere, like driving without a direction, hoping the roads lead you to a place you want to go. The cliffhanger at the end of a film or book that leaves you questioning — so unsatisfying, but so full of potential.

📚 Books I’ve loved this year:

• Mother Mary Comes To Me — a memoir by Arundhati Roy, mentioned above

• Love in Exile — Shon Faye

• Between Two Kingdoms — a memoir by Suleika Jaouad

• Crying in H Mart — Michelle Zauner

• Heaven’s Coast — a memoir by Mark Doty

• Wintering — by Katherine May

• Sunburn — Chloe Michelle Howarth

Subscribe now

One of the guests told a tale of ancient wine being discovered by people marooned on a desert island; another shared how the wine they brought had once been an elixir that saved a town at the foot of a mountain from a plague. But when it emerged that some people had used ChatGPT to concoct their stories, the magic instantly faded. One person scoffed that people couldn’t use their brains anymore, even just for party games.

There's a growing body of research indicating that the use of AI tools can negatively impact critical thinking, memory and attention span. From conversations I’ve been having recently with tech and communications professionals, people using AI regularly are starting to be seriously concerned that their skills, honed over a lifetime, are waning.

While AI is readily accessible for many knowledge workers, and its use is often mandated by companies eager not to fall behind, some who do have the freedom, some might say the privilege, to push back against AI are starting to do so.

Some individuals I’ve spoken to have spent time outlining the tasks AI can help with, and ruling out the things it shouldn’t touch: for example, “I use it to write emails; I don't use it to summarise client calls.” Some have drawn strict boundaries, restricting AI use to their desktops, or only using it at work and not in their leisure time. I've heard examples of people only allowing themselves to use it in the morning, or at night, or giving themselves at least one AI-free day per week.

Others are trying to resist the pull of using AI at all, afraid of where their data is going, and how far their AI use is lining the pockets of Big Tech CEOs, and harming the environment.

Research indicates these small acts of resistance against AI are becoming a wider trend. A new study published on Friday by the Digital Futures Institute at King’s College London and Responsible AI UK shows that 42% of UK individuals are restricting their use of AI, primarily due to worries over privacy and security. The second most common reason cited was that people want to continue working the way they always have.

“Anti-AI sentiment seems to be on the rise,” said Professor Kate Devlin, chair-director of the Digital Futures Institute at King’s College London, in a statement.

“Employers and tech companies should be mindful that people are anxious about using AI tools […]. Our findings also show public attitudes are not split between a clear ‘pro’ and ‘anti’ AI. People acknowledge potential uses and benefits while simultaneously worrying about possible downsides.”

Similar attitudes towards AI are emerging in the US, home to the companies that created popular chatbots like ChatGPT and Claude.

According to a recent survey by Pew Research Center, just under half of US adults use AI chatbots. Yet just 16% of respondents think the technology will have a positive impact on society.

As you can see from the graph, a large proportion of respondents are also ambivalent about what an AI future could bring.

It seems people are increasingly questioning how AI fits into their daily lives, and whether it should at all.

What I hear repeatedly in my reporting is that people want to choose how they use this technology, rather than having it thrust upon them by employers, peers, or the tech CEOs predicting an imminent jobpocalypse.

While I don’t spend too much time with AI chatbots, I do worry about the time I spend on my phone and laptop and what impact that’s having on myconcentration and mental health.

As a child, I used to tuck myself away in the corner of a room and read for hours. I remember being so immersed in a story, I could hardly wait to turn the page and see what happened next. I realise now how lucky I was to have had books at such an early age; to have experienced what pure concentration felt like before email, notifications and social media. As an adult, when I feel irritated, sad or hopeless, I turn back to reading to portal into another world.

Recently, I’ve been trying to train myself to concentrate and absorb information better by reading challenging texts that force me to stop, consider, reread a page or chew on an argument for days to understand it fully. I’m not about to read Tolstoy’s epic novel War and Peace (though I hope one day I will have the guts…), but perusing literature from other eras is a great mental exercise, and a pleasant reminder of how much language has evolved.

I recently read Virginia Woolf’s A Room of One’s Own, an essay published in 1929 which argues that women need financial independence and their own private space to be able to write. It’s a fantastic reflection on what literature would have looked like, and how the literary canon would have evolved differently, if women had had the same opportunities to experience the world as men did and the freedom to creatively express themselves.

“I would venture to guess that Anon, who wrote so many poems without signing them, was often a woman,” — Virginia Woolf

It's also quite hard work. Woolf writes in the modernist style, which is characterised by long meandering sentences, meant to mirror the unfiltered, continuous flow of thoughts. The style is widely considered to be a reaction to the destruction and disillusionment brought about by World War One; authors traded in traditional rigid forms for looser, fragmented prose that seemingly holds multiple lines of argument at once.

In A Room of One’s Own, you really feel like you are inside Virginia Woolf’s brain: you witness one thought appear and then morph swiftly into something else. Thoughts and ideas toss and turn like leaves in a salad bowl. It can be difficult to follow, but her writing feels like a nice counterweight to the sharp, condensed sentences generated by AI that we are all, regrettably, used to reading by now.

If AI is to become a significant part of all of our lives, we have to find ways to be continuously curious, talk to each other, debate, test and challenge ourselves, and, conversely, find time to be still and let the mind wander.

As someone who regularly struggles with feeling overwhelmed and overstimulated, I’ve started to put together a list of things I do to counter that feeling, which you can read here. (I’ve played around with style and released myself of traditional form for a bit of fun. Virginia inspired me).

Establishing your own personal code of conduct for how to engage with technology, and not allowing it to overrule you, your mind and your sense of self, feels like one of the most radical things we can do right now.

🌱 One more thing:

I’ve really enjoyed looking through the AI Resist List which documents the ways communities around the world are fighting back against harms committed by AI. It makes me hopeful that the future hasn’t yet been decided for us, and that we all have agency to shape AI development.

It’s put together by journalist Karen Hao who wrote Empire of AI, (the next book on my reading list!), alongside researchers, reporters and technologists around the world. Enjoy!

If you have any thoughts, ideas, hot takes on any of today’s ramblings, reach out to me. I love reading your emails and DMs.

Companies are downsizing, layoffs are continuing (though how much they are actually related to AI is a different matter), and many entry-level engineers are struggling to find jobs. The AI era has also ushered in a new brand of tech bro-ism, as my former colleague Freya Pratty wrote in Sifted, as if the lack of diversity in the tech industry wasn’t bad enough already...

To find out more about how the startup workforce is changing, and to discuss pragmatic solutions for handling the shift to AI, I chatted with Liv Parry, talent partner at US venture capital firm General Catalyst.

We spoke about why generalists are in top demand, why “diversity of experience” is needed in teams more than ever — and how the graduate rotation scheme model could help talent at risk of AI-related layoffs stay in organisations.

Liv also made a case for staying positive. AI has lowered the barrier for non-technical people to create companies, and as jobs are reshaped, new roles are emerging.

This interview has been condensed for length and clarity.

Many companies have been restructuring their teams since AI was introduced. How are you seeing this play out in startups?

There’s a healthy rise of anxiety in founders where they’re being told they need to compress teams with AI and they don’t know how to do it. Founders have always been anxious about hiring the right people, but now they want to understand even more about what assembling a good team looks like.

As talent partners, we’re spending most of our time redesigning organisational structures and hiring roadmaps: first by understanding how the business operates, then evaluating existing teams and then starting to identify what the gaps are.

But whereas before it would be clear that a company needs to hire three account executives or a product lead, we’re now having to rethink what these roles are called as the nature of the jobs are changing so rapidly.

I’m quite anti-titles because I believe a lot of people miss out on great roles because they’re pigeonholed. So I think AI has brought about a great opportunity to redesign how we think about great people and smart generalists. It feels like quite a creative and experimental time.

There’s a real fear that entry-level roles in startups are being automated out with AI. To what extent are you seeing this happening, and in which functions?

In early stage startups, engineering is coming under heat the most. Teams are compressing, and there’s a lot of emphasis on hiring engineers who can orchestrate AI systems. You don’t just need people who can code anymore.

It’s not an entirely new phenomenon, but we’re finding there’s a preference for the more experienced senior engineers that can make decisions on their own, over junior people who need training. I do worry about what will happen to junior engineers.

The good news is that new roles are emerging as well. Because AI has made non-technical people more productive and faster at making decisions, I think we’ll need people that have really good judgment and taste.

Internally we talk about ‘tastemakers’, which I know sounds a bit like an album cover… But this role is hard to define because, how do you judge someone’s taste and judgement? How do you define that as a skill to hire for?

Are ‘tastemakers’ likely to be generalists who are cross-functional in an organisation? So if they’re in the tech team, they wouldn’t only know how to build a product, but also how to sell it, how to market it?

Yes, definitely. I think tastemakers maybe won’t end up being a single role or a function. I think this is a behaviour many people will have across the organisation as AI gets further embedded — you need people with good judgement. So it’s changed the criteria of how we evaluate people across different functions.

For example, go-to-market is significantly evolving. Sales reps are now needing to be much more technical to be able to understand the product they’re selling and help customers to use it. Ultimately, if the customer isn’t using your product, your revenue is shrinking — and so you need someone who sits between sales and customer success whose sole job it is to make sure deployment is successful.

I’m unsure what this particular role is going to be called. Different organisations have been calling it different things and all kinds of different people are coming to the role, like ex-product managers and consultants. But it’s definitely one to watch.

Do you think team structures will be increasingly more fluid as AI speeds everything up and creates new tasks and temporary roles that then morph into something else?

Totally. There are a few places where this isn’t happening — for example in engineering, tech and product.

I think a great example where I have seen this happen a lot is in the talent and people function. There are some great examples across our portfolio of great product or commercial people that are given the people function as a project to build up and run for eighteen months — then they move into a marketing role.

There’s also a lot of fluidity in sales, customer success, product — anywhere that is hovering around the customer.

We’ve also had requests from founders to have a commercial person running the people function. To some extent, the people function has always been commercial — a really good and well-run people function is super data-driven and owns and runs the performance of the company. The reason it’s having its moment right now is because performance is everything; teams are leaner, and retention is really important.

I think we’re also going to see more ‘head of people’ positions become ‘head of performances.’

How is this changing how founders assess who to hire — if this level of fluidity among people’s roles and skillsets is so much in need?

How to assess people is a free for all at the moment because you are having to really focus on behaviorisms and how people operate, which again, isn’t necessarily a new phenomenon.

We encourage founders to assess how people think, how they work under pressure and how they make decisions. You then layer that with the skill set: what do they actually bring from a domain perspective?

Behaviourisms are being prioritised because the need for smart generalists is massive. Think about the chief of staff role. That was new three years ago, and now everyone has one.

That is an indication of a need in a company that hasn’t gone away — a need that AI hasn’t disrupted, but instead amplified — which is that organisations need someone who can sit in and work across various different functions.

There’s no doubt that layoffs are happening as companies of all sizes adopt AI. What’s your take on how this can be handled as a society?

AI transformation and displacement is real, and we’re all waiting to see how this is going to be solved. I think a lot about how to ensure companies can see the gains of AI reinvested into new work and new teams. The more concrete examples we have of companies navigating this thoughtfully and carefully, the better.

One of our earliest-stage businesses hires by taking on entry-level graduates or people with experience in tech or academia or what have you. These people join as generalists and work in all the different functions of a company for six months — just like a good old-fashioned graduate rotation scheme. At the end, they get selected for a job, but then get rotated again at another point in time.

I’m wondering whether organisations that need to reduce headcount in certain teams can think about introducing a similar rotation scheme — like a repurposing program, where great performers get reallocated to another area of the business.

There have been huge examples in the press where businesses like Klarna made massive headcount reductions and replaced people with AI, which didn’t work. We know this direct replacement is careless. So how can we ensure great performers stay in organisations?

There’s a worry that senior profiles and slightly older people in the workplace are also having difficulty getting hired as there is a bias in startups now for younger, ‘AI-native’ talent. How should startups be thinking about this?

It’s hard to make a conclusive point on how far senior profiles are being hired over entry-level profiles and vice versa, as there isn’t much data, and each company has a unique situation.

I think one thing startups need to be mindful of is having a diversity of experience in teams. If you don’t have that, and you’re just hiring purely for work ethic or AI fluency, you can end up with quite a homogenous room full of people.

Having up and coming talent is important, but you also need the experience of people who have worked in various roles and companies — they provide the ability to look around corners and predict what might happen next.

I know that teams are leaner and founders are trying to be more thoughtful around headcount, but they need to hold themselves accountable for ensuring their teams are representative of people from all backgrounds, be it in terms of ethnicity, gender, age or education. What we don’t want is exclusively young “cracked” AI engineering teams that are more male heavy being the only ones to create companies.

Passive hiring is a great skill for founders to adopt. It just means when you meet great people, you just find a way to include them. That’s the wonderful thing about this age: you can design new roles and call them anything you want.

Amid all the concerns about AI we discussed, what is one glimmer of hope you see in terms of how AI is being embedded into companies and teams? What’s the big opportunity?

I think it’s freeing up time to think and create more. What I’m hoping is to see a rise of creativity, of new ideas, and founders — especially in Europe — starting out on their own because you don’t need big teams anymore to build things.

I’m also hoping that what we don’t lose is the authenticity of relationships and looking out for and caring for each other. That might sound wishy washy, but the downside of AI is that it can create isolation.

I’d love for people to focus on connecting with people more, doing things in person, bringing people into the journey of their companies early on, referring their friends to great roles — just being more intentional in connecting with the people who give meaning to their lives.

I met up with a few academics working at various universities in London to give me an insight into how AI is reshaping education. One of the current quandaries is how to prepare students to use AI sensibly, while still ensuring they develop skills like critical thinking, judgement and empathy — things that will be even more important as AI becomes further integrated into our lives.

Another concern was around assessment: students are using AI to come up with ideas and write essays, so how can universities ensure they’re still passing exams on individual merit? Blanket bans aren’t really that effective, said some of the people I spoke to, and some universities are struggling to articulate what the parameters could be for acceptable AI use.

Below are a few thoughts on one particular risk of using AI that — while seemingly obvious — is still good to be reminded of.

There are so many stories out there at the moment about authors who accidentally misattributed quotes, or simply made them up, by using AI in their research. I doubt in most cases that this was intentional; rather it was too much blind faith that AI is a reliable, infallible research partner.

One of the most pointed-out risks of using AI is that it could impact our ability to think and express ourselves authentically (queue the rants on LinkedIn about AI-generated posts). But a greater risk is that we forget to check where information comes from in the first place.

The challenge of using chatbots like Claude and ChatGPT is that it’s not immediately transparent how they arrive at a specific answer, a professor of neuroscience at King’s College London I spoke to last week reminded me.

As a test, he asked Claude to compare three separate degree curriculums and give them a grade out of five. When Claude produced its assessment, the professor asked how the AI had come to its conclusions: what data it had pulled from and what criteria it had used.

Claude answered with an apology, stating that it couldn’t find the right information to compare these curricula and so simply filled in the gaps. In short, it made stuff up. These machines are optimised to give plausible answers, not to guarantee that every claim or data point they serve up is true — and this can have severe consequences.

The professor told me that he encourages his students when using AI to ask chatbots to give an explanation of the ‘thought’ process it went through to produce an answer to a query — as he puts it, “show your working”. That way, you can spot whether the AI simply fabricated something.

This resonated as I remember doing Maths exams in secondary school, where examiners would still give you points for showing how you arrived at an answer, even if it ended up being wrong. That principle of rewarding the process rather than just the final output will likely become even more important in the age of AI.

Part of the problem with assessing students at university today, said a professor of biology I spoke to, is that AI-generated writing, at least in the scientific field, is getting quite convincing; sometimes it can be hard to distinguish whether a student has used AI or not. That ambiguity makes it even harder to know whether students have actually thought through and understood the material they’re being taught.

Schools and universities might have to find new ways of assessing students that encourage them to ‘show their working.’ Some universities are toying with the idea of using oral exams to supplement written ones; if students can speak confidently about a subject and answer questions on it, it’s usually a good sign that they’ve absorbed and understood the information and formed their own critical judgements. But setting up oral examinations — or even other ways to combat ‘improper’ AI-use — requires resources that some UK universities don’t have.

The bottom line: anyone using AI systems should ask to see the working. We have to get better at questioning AI’s claims, interrogating its sources and weighing up evidence to form our own judgement. It’s not enough to read the AI summary (though I can see how seductive this might be).

I’m interested in learning more on how AI is affecting school and university education — and I wager that not all use cases are bad. I want to know more about how AI is being used in classrooms or lecture halls; how universities are balancing teaching AI while ensuring students use it mindfully; and how they’re thinking about regulating the use of AI in coursework and exams, if at all?

If you have any thoughts, please get in touch with me here.

Subscribe now

As tech leaders and politicians rush to roll out AI as quickly as possible into every facet of our lives, many of us are questioning how our jobs and our lives are going to change — and what will be left for humans once machines take over (if that will indeed happen).

I’ve been interested in speaking to people who spend time thinking about what it means to be human, and how to maintain that humanity as AI develops. It feels like a necessary counterweight to the “AGI-pilled” talk we read about in the headlines.

Olivia Gambelin is among the thinkers I’ve been following on this theme. Olivia is an AI ethicist who has spent the last decade helping companies develop and deploy AI responsibly — splitting her time between San Francisco’s tech world and Brussels, where she advises on AI policy and regulation.

When we spoke, Olivia reflected on the growing anxiety around AI, the pressure people feel to adopt it, and why she believes there are human qualities that AI simply can’t, and won’t, replace.

I also got a sneak peak into her forthcoming book, ‘Not Every Solution Has a Problem,’ where she argues that many of the AI products being pitched to us are just “solutions looking for a problem”, rather than actually addressing human needs.

This interview has been condensed for length and clarity.

When we talk about ethics in AI, what do we actually mean practically?

The ethical questions underpinning AI, such as moral agency or what ‘consciousness’ means, have existed for millennia. But AI ethics has only been an industry practice for about a decade.

To put it simply, ethics is split into two camps. There’s responsible AI — which is the ‘how’ do you actually do this? It’s the infrastructure and operations that allow people to make responsible decisions around AI and mitigate the risks that come with it.

Ethics is the why. It’s about looking at specific human values and principles and trying to reflect those in the technology we build so that we benefit society as a whole.

An example is curiosity. An AI ethicist would say: I want to be more curious. I don’t want AI to be more curious for me. I want my use of AI to help me be more curious.

I can imagine that the emergence of modern generative AI, ChatGPT and so forth, in 2022 was one of many major turning points for the field of ethics.

100%, because all of a sudden, people could actively engage with AI as if it was a physical thing.

We’ve had AI on our phones for years — but unless you work daily with AI and were actively seeking out a specific application, you weren’t really cognizant of when or how you were interacting with it. But with ChatGPT, people could understand, here is a chatbot I’m actually conversing with.

Generative AI catapulted the AI ethics conversation really far forward, when we’re still trying to get basic governance for things like computer vision. Now, we’re having to look a step ahead and figure out how to deal with things like vibe-coding and agentic AI.

I guess that is the inherent challenge: building an ethical code for technology that is quickly progressing. When a company puts something like ChatGPT out into the world, it’s perhaps hard to anticipate the multiple hundreds of thousands of use cases that it’s going to be used for and how to implement certain guardrails.

I don’t think that’s inherently bad — though I would definitely say there were better ways to release this kind of technology into a world where everyone’s a guinea pig now.

What is bad is what happens when ethical challenges around a technology meet a media frenzy.

There is this general narrative around AI that you’re falling behind if you don’t use it, and it’s become such a polarised topic that it makes it really difficult to navigate the ethical questions about how it should be used.

I can’t sit down with people and ask them what feels like good AI use to them, because the person themself is like: ‘well, this influencer told me that I need to use it in these ten ways.’ And ‘Mark Zuckerberg and Sam Altman are telling me I need to do this and this.’

We don’t have the space to be able to experiment with the technology in a way where we are the agents of that experimentation — where we are the ones figuring out what works and what doesn’t. It’s really being forced on the general public in ways we’re not yet equipped for.

For example, teachers are expected to be able to teach the next generation about how to use AI, but they haven’t had the chance to figure it out themselves. That’s really the problem here: there’s no time for questioning.

We’re reaching a point where people feel they can’t keep up, and that fear — for themselves, their children, their friends — is driving pushback.

We’ve seen in the US that fear and anxiety around AI is leading to anti-tech protests and even violence, like when a Molotov cocktail was thrown at OpenAI CEO Sam Altman’s home in San Francisco earlier this year. That’s an extreme example, but what do you think is driving a lot of public outrage about AI?

I think what people want is the agency back over their lives. With the pace of how AI is being developed, people think: when is it coming for me next? And the way it’s positioned is that AI will replace you — but that’s not really what we are worried about.

We’re worried about how AI is going to take over our purpose in life: whether it’s an artist worried about AI threatening their creative expression (though I have many arguments as to why that will never happen), or an engineer who writes beautiful code. I think a lot of the frustration and anger comes from feeling like we don’t have a say in it.

Again, the positioning is there’s an AI code to crack — meaning there’s one way to do it and you aren’t smart enough yet to know how. If you did, then it would unlock all of these amazing things in your life, and you’d be part of this elite club in society. That’s a bunch of bullshit.

You just mentioned human ‘purpose’ there, which is a theme at the heart of your new book Not Every Solution Has a Problem. Tell me more about the core idea behind it. What are you trying to challenge or reframe in some way?

I started researching for this book on the premise of: we have all these AI solutions that don’t really solve any problems. Businesses, for example, are spending so much money on AI tools, but many of them multiply complexity rather than reduce it.

We have clearly hyper fixated on the solution side to this equation — so what if we were to look at it the other way: how do you find and define problems worth solving?

AI is often pitched as a way of making us more human; it automates the things we don’t want to do so we have more time to do the stuff we do want to do. But what does being human actually mean in this context? What are the intangible parts of being human that are undeniable — that we can’t erase or ignore?

What are the human qualities that we can lean into more in this digital age, where we’re starting to reinvent ways to think and interact? That’s what the book has become.

Part of your research has been interviewing people across varying walks of life. Tell me about them.

I had two rules for the people I interviewed. One, they can’t come from tech — we’ve heard too much from tech. What we’re faced with right now is a cry for humanity, a cry for: what does it mean to be human? I didn’t want to hear more about how to build an AI system, I wanted to hear from the average everyday person about how they go about being human.

And the second is that they have to love what they do — because this implies they’ve found a reason behind what they’re doing. They’ve clearly found a problem they like solving.

I’ve spoken to a fun handful of people. I interviewed a married couple who are artists in Rome who invited me to their studio. I interviewed an ex-BBC radio producer; two Belgian lawyers who are good friends and just gave each other shit the whole interview with their French bulldog sitting next to them; a former Finnish cycling champion who I spoke to while cycling. She literally rocked up to the interview and pulled two bikes out of her trunk and was like — let’s go.

I interviewed a Buddhist prison chaplain. I talked to a financial analyst, a bartender, a wine sommelier, and a breathwork specialist. I interviewed a military widow who told me her story which had me crying by the end. She’s a beautiful person.

So are you essentially trying to find out what the human skills, values or emotions are that shouldn’t, or perhaps can’t, be replaced by AI?

Exactly, and this is something that has come out in the research too. In the interviews, I often ask, what is most sacred to your work? For some it’s trust, or the ability to sit down and talk to someone. A magician said suffering, which was interesting.

When I followed up with the question: can AI ever replace these sacred things? Most would say no — but then some would backtrack and would say, ‘I don’t know, as who knows where this tech is going.’ But why can’t we have that same faith in the parts of what makes us human?

That’s where I’m trying to flip the narrative. I want to empower people to say: no, AI cannot replace this.

What are some of the deeply human experiences that should be safeguarded from AI?

What I’ve been finding is that there are different aspects of the human experience that feel like problems at first — but they’re not actually problems, they’re needs. There are things that we need to experience because through experiencing them, that’s where we find purpose, development, growth. When we try to solve them with AI, we basically shut ourselves off from that path to growth.

So to give you an example, vulnerability feels like a problem. People are uncomfortable with it. We try to use AI to solve for these moments of vulnerability — like in dating. Dating is just uncomfortable, you have to be vulnerable. You have to be able to craft that lovely first message or a breakup text, right? Being vulnerable allows us to connect with each other.

So, if we try to use AI to solve what feels like a problem, we are cutting ourselves off from that very human purpose of connection. Vulnerability actually isn’t a problem to solve.

To play devil’s advocate, are there any areas where AI can actually help us be more human?

AI can help us be more human, but only if we first respect what it means to be human. That means respecting our sense of humanity and our lived reality — those sacred things that the people mentioned in my interviews with them. If we start there, we can ask ourselves how AI can support us in becoming more human.

The problem is, we often don’t respect those things.

Do you have an example?

One of the elements I’m looking at is imperfection — our need to be imperfect. AI is designed to do the opposite; AI is designed to be perfect and we tend to default to it because we feel pressure to get things exactly right.

I did a workshop with a big PR agency in Stockholm where we talked about how AI could assist them in their work.

The agency’s chief strategist said that imperfection is the whole point of the agency’s work, and is what makes something feel human and creates connection. She suggested that what they could use AI for is to get as close to perfect as possible, and then they could add their own imperfect human flair.

She even suggested reframing their company motto as ‘creating imperfections for a hundred years’, because it’s those slight imperfections that have always made their work stand out.

That kind of shifted everything for them. The conversation went from how can AI solve this, to what are the human qualities we want to portray in our work, and how can we get AI to help us do that.

Is this how you would see, in a utopian world, humans being able to live comfortably side by side with machines?

Yes, my hope is that we’ll start the conversation with having respect for what humans can offer and then thinking how AI can support us — not the other way around.

But over time, that lens started to feel a bit too narrow. What is often missing in tech reporting is a way of looking beyond the methods the market uses to assess companies — money raised, valuation, potential for scale — and instead asking how technology is shaping the human experience. That’s what I want to explore with this newsletter.

Technology has crept up on us in a way that makes many of us feel uncomfortable. Our phones have created a parallel world that sucks us in like a vortex, distracting us from reality and from the people around us: from a neighbour needing help with their shopping, to an elderly person looking for a seat on a crowded train. At the same time, headlines abound about the (very real) dangers of social media use in young children and about the difficulty of getting a job as a graduate. Another growing concern is the slow erosion of our mental and physical health, shaped by an addiction to our phones and inboxes, and our need to feel constantly connected.

On a broader scale, technology is being used to incite hate speech, rally support for parties and individuals with extreme and dangerous ideas and scam people out of money. Everywhere we look, technology is asking new moral questions of us, like whether AI should be embedded into weapons, or whether certain individuals or institutions should even be investing in defence at all.

This is not to say that all technology is inherently “bad” — many of us can point to ways in which tech has made our lives easier, richer and more connected — but rather to start a conversation about what’s not working and how we can do better. The questions I seem to return to on a daily basis are: what does it mean to be a human living alongside rapidly evolving technology? What would it mean to regain control of the parts of our lives that have been handed, in some sense, to algorithms and the (predominately) men who run them? And, how can we experience, interact with and build technology in a more thoughtful way?

Those are the questions I’d like to explore with this newsletter. I’ll be looking at AI and its impact on the workplace, creativity, education, and our daily habits; the ethical challenges of building technology; and the ways companies are increasingly shaping the political and social systems we rely on. I want to talk to people from a range of disciplines who bring thoughtful, philosophical, idiosyncratic takes on technology and its place in our world. There’ll also be the odd interview on AI being used in creative ways that we may never have deemed possible.

If you’re curious to explore these questions with me, join me on Substack here. I’ll be in your inboxes every other Monday.

I started journalling in the hope of finding inspiration. Every so often, I managed a few lines — but I ended up scrapping them. I couldn’t shake the feeling I was writing for an audience, as I’ve been trained to do throughout my entire career. The words scribbled in a notebook were not worthy of my own scrutiny, let alone that of the public. So, I decided to write this Substack article on a topic I know intimately: that awful feeling of being completely lost when sitting down to write.

When I was at university, an insecure literature student struggling with long reading lists and the rather grandiose, academic style I was expected to write in, I took comfort in the great authors and poets who have also felt the anxiety of creating.

Sixteenth-century English poet Philip Sidney wrote in Astrophil and Stella about the difficulty of finding the perfect words to describe his love for the woman he desires — concluding, rather cheesily, that anyone struggling to put pen to paper should simply “look in thy heart, and write.” Romantic poet John Keats lamented in his 1848 sonnet ‘When I Have Fears I May Cease To Be’ that he might die young before being able to write all the poems in his “teeming brain.” American writer Sylvia Plath wrote powerfully in her poems about the often debilitating act of creating, comparing her lack of inspiration to barren landscapes, and her unsuccessful poems, in a more harrowing metaphor, to stillborn babies.

Creative block, while frustrating, is an essential part of the artistic process: a platitude, but a true one. It allows writers to pause and reflect, providing space for a change of direction. The eureka moment that often follows a fallow period of ideas can feel like ecstasy for a writer.

Though I fear that many people today will never experience the joy of being stuck and then unstuck. In this new era of AI-generated content — from emails and social media posts to blogs and articles — people are outsourcing their voices and ideas to machines, diminishing their unique way of expressing things in the process. The result is just… average.

I get why people do it; using AI removes the fear of creative paralysis. Unlike a person, a chatbot doesn’t have any notion of self-esteem (or the lack thereof), nor does it have premature worries about the battering its reputation might take if it produces something inadequate. With just a few words of a prompt, AI can produce streams of content that many people think do a better job of writing than they could themselves.

The thing about AI is it’s trained to recognise patterns; using it in drafting can help solve grammar mistakes and make an argument more structurally sound, but it doesn’t always produce riveting stories. It simply can’t replace the words that flow from a human’s embodied experience living in the world. What AI does is smooth out language, often stripping away the beautiful human chaos that gives a person’s writing its voice and personality. It’s like being lost in a desert surrounded with beige-coloured sand.

Looking at the internet today with all those same-y sounding blogs and LinkedIn posts, somewhat controversial writer Charles Bukowski, who believed in the rawness of the creative process and decried writers who spent too long being hung up on grammar rules and form, would be turning in his grave: “Let’s allow ourselves space and error, hysteria and grief. Let’s not round the edge until we have a ball that rolls neatly away like a trick,” he wrote in a 1959 letter.

My writing is full of spiky edges that editors often need to refine and polish to make it more comprehensible and enjoyable for the reader. When I draft an article, the structure is often messy, there’s probably a few awkward metaphors and turns of phrase, and a handful of long, sprawling sentences that most editors would chop up immediately. But those quirks give my writing its own cadence and edge. If I was to deconstruct my own writing style, it’s probably a mix of academic-style phrasing left over from my university days, some decorative, but often redundant language, and my journalistic training in clarity and getting to the point. It comes from my unique mix of experience and my personality shaped by moving through the world, something I struggle to see large language models, no matter how good they get, being able to replicate.

For writers in today’s world, what’s more challenging than the act of writing itself is thinking about how to continue doing it in a way that is financially viable and creatively fulfilling. The job of a writer is enormously more pressurised than it once was as newsrooms downsize, audiences turn to social media for news and writers with meaningful things to say have to push harder to cut through the noise — especially with platforms swilling with AI slop.

Additionally, it seems like it’s not enough to be a published author; one has to have a podcast, a LinkedIn brand, an events series and a knack for creating Instagram reels or TikToks. It all seems so exhausting.

I say this as a person who is now entering the fray as a freelance writer, unsure of how to do this writing thing anymore. I ask myself regularly: how do you get your voice out there anymore? Should I even try? Does my opinion matter? Will anyone actually read this all the way to the end? Is it better to just do … nothing? Put down the pen and let someone braver than me contend with the algorithms?

But I do still have faith in the written word and the human capacity to enjoy it, simply because I know how I feel when I read a book and feel the person’s personality shine through. I just finished reading ‘Between Two Kingdoms,’ a biography of Suleika Jaouad who writes about battling cancer at a young age, and I feel her life-force leaping off the page. It moves me in a way that only human stories can. That shared space of vulnerability between writer and a reader, who brings their own history, dreams and suffering to the fore to understand and relate to the writing in front of them, is nothing short of magical. If that kind of writing still exists, it gives me the confidence to keep returning to the blank page.

“The power of story is to heal and to sustain. And if we are brave enough to tell our own story, we realize we're not alone, again and again,” — Suleika Jaouad.

Subscribe now