ChatGPT's Deep Research and thinking about thinking with AI.

Why your human network will become the ultimate source of competitive advantage.

ChatGPT’s new Deep Research tool triggered a real ‘holy f*ck’ moment for me recently.

Just a few weeks ago, I wrote about my friend Evgeny’s brilliant article on how we’ll react when we realise that machines are more intelligent than us.¹ But I certainly didn’t expect that, just a few days later, I’d experience a machine doing my job – my not-in-any-training-data, seemingly-random-dot-connecting, intellectually creative job! – as well as I can. For ‘just’ $200 a month (update: as I was writing this piece, Open AI made Deep Research available to Plus accounts, costing just $20 a month 🤯).

I’ll get to sharing the prompt and its output that triggered this existential moment below, but since then I’ve spent the last few weeks diving into its capabilities and reflecting on where this is heading and what it means.

The immediate implications of Deep Research’s expert worker-level analysis are fairly obvious:

• Asking the right questions > getting answers;

• Validation, curation, and judgement will become the new sources of value;

• Agency – taking action – will trump analysis.

You’re undoubtedly already on top of these (and if not, here’s ChatGPT to unpack them further ;)

But, having burned through nearly my entire quota of 100 queries since getting access to Deep Research, I can see that the bigger, non-obvious implication of ultra-intelligent AI research agents will be to make your human network the ultimate source of competitive advantage.

Which isn’t something I’m hearing from the AI evangelists, but aligns well with my core belief that AI will make human connection far more, not less, important.

Let’s dive in…

---

Before we get into it: a note from our sponsor (me!). Many readers here discover me via my keynotes. If you’re planning your next strategy session, sales kick off, or customer conference – 🚨 I’d love to help.🚨

Typically I inspire clients with:

1. Your Future Normal: a trend keynote, bringing you non-obvious yet relevant and practical insights from outside your industry.

2. VisuAIse Futures: anew interactive, ‘multiplayer’ AI-powered creative experience. Watch the 2-min highlight video.

Read more about these at the bottom of this email, or reach out to Renee Strom on renee@ideapress-speakers.com if you’d like to bring me to your next event.

And if I’m not a good fit, then do check out the best speakers I saw in 2024.

---

The Good, The Generic, and the Existential.

Ironically, it wasn’t Deep Research’s research skills that blew my mind initially. It was its writing.² Most AI copy is obvious – bland, equivocal, and packed with ‘the true meaning of X isn’t just Y, it’s Z’.

Yet on my first Deep Research query, about how AI will reshape B2B events, I got back a 3,300 word article that blew my mind.

You can read the full piece here, or just check out these first couple of paragraphs:

From Algorithms to After-Parties – An Opinionated Take on AI and the Future of B2B Events

The cocktail-hour chatter in the convention center isn’t just between humans anymore. In one corner, two CEOs compare notes on an AI-matched meeting that felt eerily spot-on. In another, an attendee consults her phone, where an AI assistant whispers that the person she just met might be a perfect business partner – their conversation data has been cross-referenced in seconds. Welcome to the not-so-distant future of B2B events, where artificial intelligence doesn’t just streamline experiences, but fundamentally reshapes them. In this thought piece, I’ll venture beyond the obvious efficiency gains and explore some non-obvious implications of AI-driven events and where current trends might lead us.

The Augmented Attendee: Your Personal Event Sidekick

One underappreciated change AI brings is the rise of the “augmented attendee.” We often talk about how events have AI-powered apps, but consider it from the attendee’s perspective: increasingly, each participant will have their own AI sidekick. This goes beyond a static event app – I’m imagining a future where you, as an attendee, come to a conference with a personal AI agent (perhaps integrated in your AR glasses or phone) that knows your preferences, your schedule, and even your social anxiety level. This agent might suggest, “It’s a good time to take a break and get coffee – your energy seems low, and there’s a networking game starting in the lounge that fits your goal of meeting marketing experts.” In essence, every attendee could have a kind of coach/concierge in their ear, nudging them toward the most fulfilling experience.

What’s non-obvious here is how this might change attendee behavior. With an AI optimizing our time, we might take fewer random walks through the expo hall and instead follow more curated paths. We might also become less stressed at events – no more FOMO (fear of missing out), because your AI will catch anything you miss and brief you later. Perhaps people will be more present in conversations, knowing they can rely on their sidekick to summarize the last session or remind them of that person’s name and bio. On the flip side, will we become too dependent on our digital chaperones? One can imagine attendees almost in autopilot, dutifully going where their AI recommends. The trick for future event design will be making sure these AI sidekicks enhance rather than replace the joy of exploration. The best sidekicks might occasionally encourage, “Go somewhere completely new, just to spice things up,” injecting curated serendipity into an optimized schedule.

When I first read this, it was the most ‘human’ sounding AI copy I’ve ever read. And beyond the tone, its insights weren’t just solid, they were creative. Here are five that I would have been happy to write:

If AI can successfully shrink the degrees of separation, a 5,000-person summit might feel like a tight-knit workshop because everyone you encounter seems relevant or familiar. That could fundamentally change the vibe of large conferences – less roaming a sea of strangers, more “these are my people” cohesion. It sounds wonderful, but it raises a provocative question: Will we miss the strangers?

---

Essentially, events could have storylines customized per attendee. This is the gamification/Disneyfication of B2B events, and AI is the game master behind the scenes. Unexpected partnerships might emerge: the event industry could collaborate with video game designers or theme park imagineers, with AI as the common tool bridging the gap between fun and function.

---

if attendees trust the AI, they’ll lean in and benefit; if not, they’ll disengage and all the fancy tech goes to waste. Over time, positive experiences at events could actually help people warm up to AI in other parts of life, because it’s AI tangibly helping them network, learn, and have fun, rather than, say, AI just showing them ads.

---

One intriguing speculative idea: micro-events orchestrated by AI. Instead of relying solely on big conferences, an AI could determine that 20 people should come together for a very targeted one-day meetup, and it basically organizes it – sends invites, finds common available dates, books a venue, sets an agenda. In other words, AI might spin off bespoke events dynamically out of larger communities. This could redefine networking and learning – very efficient, need-based gatherings.

---

We should also watch out for over-automation. There is a risk of events becoming too data-driven, where everything is quantified – from an “attention score” for each session to a “networking efficiency rating” for each attendee. If taken too far, this could introduce a Black Mirror-esque competitiveness or stress (“Oh no, my networking score is low, I need to talk to more people or the boss will know I didn’t maximize ROI!”). While measurement is good, events are fundamentally social and human. The value of a hallway chat can’t always be captured in metrics, and participants shouldn’t feel like they’re being micro-measured to death. The future should preserve spaces for unstructured, no-data fun too.

---

I’m not the only one who’s had such an existential moment. Paul Schrader, the writer of Taxi Driver, recently posted:

Source

It’s undoubtedly extremely impressive. But as I continued to engage with Deep Research, I realised a few important caveats.

My work focuses on exploring ideas. While I remain in awe of Deep Research’s creative qualitative insights, this is a very forgiving use case. I’ve written extensively about why I believe creativity is the killer app for gen AI, because it makes hallucinations a feature, not a bug. Most researchers and analysts focus more on numerical data, which can be problematic.³

This is also a topic I’ve thought deeply about. My prompt included ideas that have emerged from discussions with some of the world’s leading event professionals over recent months. And, as a few people commented when I posted these paragraphs on LinkedIn, its answer echoed my writing style. That was deliberate, as you can see below:

When I used Deep Research on topics that I was less familiar with, I struggled to get nearly as impressive outputs. They were okay. Generic inputs give generic outputs. They can still be useful, but they don’t feel like an existential competitor.

So, the Deep Research paradox is that, far from democratising intelligence as initial AI studies suggested, these tools will help the rich get richer:

• the more deeply you have thought about a topic, the better you can direct the AI’s research, and the more insightful your results will be.

• the more prolific you are in your own voice, the ‘better’ (i.e. more uniquely you) your AI output will be.

Or, because it’s 2025, let’s summarise this in a meme:

Source unknown (yup, even to Google Images, Grok, Perplexity, and ChatGPT!). I’ll credit if / when I can :)

---

Deep Diversity; and living in a world where no two answers are the same.

At the same time as going down the Deep Research rabbit hole, I saw a tweet from the venture capitalist Paul Graham questioning the subtle ways AI might reshape society, complete with the answer that ChatGPT gave him, as you can see in the screenshot below.

His answer was thought provoking. But it also inspired me to see what I’d get if I asked ChatGPT the same question. You can see my answer in the screenshots below, or here.⁴

AI's Subtle Implications

---

Now, this isn’t a rant about how “AI will destroy our shared reality!” (which was, ironically, one of the points in my ChatGPT answer!). But a reflection on what it means when AI guides us all in slightly different directions.

Because the most interesting aspect of this new, diverse reality is that we will be keenly aware of it.

It took me a few minutes to experience the differences between Graham’s answer and my own. And I got different answers again by asking different models within ChatGPT, let alone Claude, Gemini and Perplexity.

Most times we won’t care that our answers are different to others’. They’ll be good enough. But when tackling especially wicked problems, creative endeavours, or any important, open-ended questions with no ‘right’ answers – we’ll be able to gain an edge by collaborating with others in order to escape the semi-invisible straitjackets that AI places each of us in.

---

Liquid collaboration

MC Escher’s “Bond of Union”

Economists would say that in a world of deep diversity, creative people will be incentivised to collaborate, in order to expand their thinking. That’s the demand side. But what about the supply side of the equation?

The beauty of AI is that it also makes it exponentially easier for those being asked to help to do so. Doing the research and analysis used to be hard, and therefore costly. Now, as we said right at the top of this piece – the research element has become abundant and near-instant. Which means the source of value has shifted to the questions you ask, and whether you have the judgement to identify the key insights from the reams of analysis you get back from an AI.

Deep Research now means that smart people with deep expertise can deliver huge amounts of value, in a fraction of the time. In minutes, experts can bring decades of experience and insight onto a particular question. For example, instead of taking 100 hours to do the research, they can now spend <1 hour thinking about the prompt, and a further hour reviewing, curating and contextualising the output (it may even be less!).

This is a profoundly different approach to today – where most people, when faced with a thorny problem, default to thinking about how and where to find answers rather than who to ask.

But it’s a profoundly optimistic, human-centred mental model. Because it’s one where far from being shunted aside, human insight and expertise becomes more valuable.

---

Despite being commonly quoted, while drafting this piece I learned that this quote is almost certainly apocryphal. I guess misinformation pre-dates AI…?!

---

A few years ago, Packy McCormick wrote a great piece on The Cooperation Economy and the rise of what he called ‘liquid superteams’. In it he explored how social media had empowered individual creators, and combined with new tools and economic models this was enabling the smartest, most empowered individuals to collaborate in more fluid ways, beyond traditional company or project models.

In the future normal, expect this to become even more commonplace as AI supercharges individuals’ ability to deliver valuable thinking faster, and therefore more cost effectively.

This has profound implications on the shape and culture of companies. Corporate leaders should be encouraging their teams to ask more questions – tapping into internal and external networks of experts far more than they do today. And for those freelancers whose work revolves around ‘thinking’, time-based pricing will become even more nonsensical when AI does the work in minutes.

---

Four ways to thrive if your job revolves around ‘thinking’

Kandinsky's "Several Circles"

This piece is now more than long enough for post-TikTok attention spans.

And yet there’s so much I haven’t discussed. What is lost when you outsource a big part of your thinking? Indeed, as I was writing this I saw a study showing that heavy AI use is linked to an erosion of critical thinking skills. It’s why, somewhat ironically, I didn’t turn to Deep Research to write this piece. I wanted to let its ideas develop and evolve. Similarly, I’ve not touched on what this means for younger, less experienced professionals – how will they build up the expertise to be able to use these Deep Research tools effectively?⁵

But my main point here is a simple one – AI analysis won’t replace human analysis. Far from it. But in the age of Deep Research, success won’t come from what you know but who knows you.

With that in mind, here are four tangible ways to become someone others seek answers from:

Become an X-shaped connector, not just a T-shaped expert. I no longer work for a single organisation any more. Yet over the last 5 years, I’ve worked with many people – I’ve co-authored a book, collaborated on illustrated guides to AI, and joined a number of liquid superteams (Group of Humans, Transformation Architects and Workathon), as well as a number of other more informal online and offline communities. As the saying goes, “your network is your net worth”.

Trade in ‘secrets’. As Ben Thompson notes, there will always be information that isn’t online. And as AI-powered research and analysis becomes abundant, it will exponentially increase the value of data and insight that lies outside AI’s knowledge base. He’s talking about pure information, but with my ‘people-focused’ hat on I’d argue this will apply as much to understanding the specific relevant internal and external cultural contexts too.

Tell accessible, inspiring stories. We’ve talked a lot about extracting insights from AI’s research. But your work should never be just about cherrypicking AI’s most relevant insights, but about ‘selling’ them to your audience. Analysis was never the point of research. Its ultimate job was, and still will be, to make people believe.

Build a brand. Talking of trust – when answers are abundant, it’s as much about who shares them. This is what so many AI startups which promise to ‘kill McKinsey’ fail to understand – clients aren’t simply buying answers, they’re buying the McKinsey brand as protection.

---

Finally, I’d be remiss if I didn’t wrap up this piece with a question to you, my human readers:

What do you, with all of your insights and experiences, think?

Have you tried Deep Research? How has it changed your thinking around how you’ll work in the future normal? What have I missed?

---

Huge thanks to all the humans who helped me develop and refine the arguments here. Evgeny Shadchnev, Peter Nixey, Arunjay Katakam, Mark Hadfield, Alessandra Lariu, Suki Fuller, Jason Mesut, Alain Thys, Alessia Clusini and Joe Pine, to name a few.

---

Next: Inspire your team to thrive in this Brave New World

In the last 12 months I’ve delivered 30+ sessions, both live and virtually – from Brazil to Saudi Arabia, Las Vegas to London.

My regular trend & innovation keynotes bring fresh, cross-industry, people-first perspectives to your audience.

VisuAIse Futures takes it one step further, turning a keynote into an interactive, ‘multiplayer’ creative experience that gets your team excited at how they can use AI to accelerate your innovation culture.

Here’s what people are saying about it:

“It was so refreshing to hear how AI can be used to power human imagination, rather than replace it. And then it was even better to actually experience it”

“Fantastic session! Hugely insightful and fun, too!”

“Brilliant. The feeling in the room was positively intense whilst the images were coming through!

Feel the optimistic vibes it will bring to your event in the 2-minute video below (or watch it here).

If you’d like to discuss bringing me to your next meeting or event then please do reach out directly to Renee Strom or check out my speaking site.  

Thanks for reading,

Henry

1. tl;dr, we won’t care all that much – the economy will reorganise around new, uniquely human qualities and we’ll celebrate them instead of intelligence.

2. Even more fascinating (imho), was that I didn’t even ask for a narrative article, it suggested it independently!

3. Although as Benedict Evans alludes to, this is as much a problem about research in general, than unique to LLMs. Having led a research team for a decade, I’m familiar with the challenges of finding credible, robust sources, and understanding what a statistic really means.

4. Despite multiple conversations with various AIs and even Google (gasp!), I still can’t figure out how to capture the whole answer in a single screenshot. Sigh.

5. My early thinking here is that this will end up being less of an issue than we fear today. The paradox of expertise in an exponential world is that the smart will use AI to leverage their smarts, until suddenly the world has changed and their expertise is outdated. And then they’ll be swept away in a gale of creative destruction. This is great on a macro level (AI won’t lead to cultural & professional stasis!), although on a personal level, it means I’m thinking hard about what expertise and skills I have that will protect me from suffering this fate!