On Thursday 12 November 2026, in the classic hotel rooms of The Chesterfield Mayfair in London, Turing held "Turing in London: An evening with OpenAI and Granola". The event brought together members of Turing Frontier Fellows, an invite-only programme for senior operators, leaders and researchers in engineering, medicine, law, finance and STEM, to build with Codex, OpenAI's coding agent, and to talk about how AI gets adopted at work.

Turing in London: an evening with OpenAI and Granola, event cover

The event cover. Photo: Turing

The room where the agent meets the org chart

It was a weekday evening in November in Mayfair. Charles Street was quiet, the hotel lobby was warm with lamplight, and upstairs a room of people with fifteen or twenty years of expertise in other fields was opening laptops to hand work to a piece of software.

That was the scene Turing built, and the format said more than any slide could. The first half of the night was an OpenAI workshop on Codex. The second half was a fireside chat with Nick Taylor, Head of Customer Success at Granola, on why people keep using an AI product, or quietly stop.

I came away with a simple thought: the capability problem in AI is close to solved for most knowledge work. The adoption problem is not. The companies that do well over the next two years are the ones that treat AI as a change in how people work, not as a software purchase. This evening put both halves of that sentence in one room, which is why I was glad to be there.

Lobby of The Chesterfield Mayfair hotel in London

The lobby of The Chesterfield Mayfair, a Red Carnation hotel on Charles Street. Photo: The Chesterfield Mayfair

Why this evening, why now

Start with the numbers, because they explain why OpenAI sends a team to run workshops in hotel function rooms.

Codex has grown fast in 2026. Sam Altman announced on 8 April that Codex had passed 3 million weekly users. On 2 June, OpenAI said Codex had crossed 5 million weekly active users, a few months after the desktop app launched in February. By late July, reports put Codex and ChatGPT Work together at about 10 million weekly actives, a combined figure, not Codex alone. Those are self-reported numbers, and I treat them as such. The direction is still clear: agents that take a task, not a prompt, are now a mainstream product.

Granola is the other half of the story. The AI notetaker raised $125 million in March 2026 in a Series C led by Index Ventures, at a $1.5 billion valuation, per Bloomberg's report. Its previous round, in May 2025, was $43 million at a $250 million valuation. That is roughly six times in under a year. Named customers include Vanta, Gusto, Asana and Mistral AI, and the company is moving from personal meeting notes to shared team workspaces and enterprise APIs.

Put those two curves side by side and you get the tension the evening was built on. Usage is going up. Spending is going up. But inside most companies, the hard question is not "can the tool do it". It is "will my people still use it next month".

Then there is the host. Turing, the company led by founder and CEO Jonathan Siddharth, runs Frontier Fellows as a fellowship for domain experts who want to learn how AI is built and apply it in their own work. Per the programme page, fellows need five or more years in their field and experience leading cross-functional teams, and they commit one to two hours a week. The programme runs from September to November 2026, with in-person stops in the Bay Area, New York, London and Bengaluru, and future cohorts listed for cities from Boston to Tokyo.

The perks tell you who Turing wants in the room. Fellows get hands-on workshop series on Claude and Codex, firesides with AI leaders, and tool access that includes three months of Granola Business for up to five team members. Turing puts the total value at over $500,000. The programme's published speaker list includes Aaron Levie of Box, Kanjun Qiu of Imbue, Tanay Kothari of Wispr, Mary Liu of Runway and Carina Hong of Axiom, with Nicolas Bonamy from OpenAI's Codex team on the agenda.

This is not a developer meetup. It is a room of senior people from medicine, law and finance being taught to direct agents. That audience is exactly who AI Central writes for.

London image from the Turing Frontier Fellows programme page

London is one of four in-person stops for the first Frontier Fellows cohort. Photo: Turing (Frontier Fellows page)

Inside the room

The invitation framed the night in one line: "Join Turing Frontier Fellows in London for an evening of hands-on building and candid conversation." Laptops were encouraged but not required, a small detail that matters. Turing designed the workshop for people who might never have opened a terminal.

Part one: Codex, hands on

The OpenAI team led the first half. The session covered putting agents to work on coding tasks, from writing and fixing code to running larger projects in parallel.

That last part, parallel work, is the shift most people miss. With a chatbot, you ask, you wait, you read, you ask again. You are the bottleneck. With an agent like Codex, you scope a task, hand it over with a goal and a definition of done, and start the next one while the first runs. The work looks less like typing and more like managing a small team.

For a room of senior operators, that reframe is the whole lesson. Nobody there needed to become a software engineer. They needed to learn to write a brief an agent can execute, to check what comes back, and to decide which jobs they will never hand over.

Part two: Granola, and why people stay

The second half moved from the machine to the humans. Nick Taylor runs customer success at Granola. Before that, he spent four years scaling customer success at Synthesia, the AI video company. Few people have watched more teams try an AI product, and then either fold it into their week or drift back to the old way.

The conversation covered four themes:

  • How to get people to adopt AI in their work

  • What makes people keep using an AI product after the first try

  • How to build a customer success team inside a fast-growing AI company

  • How AI is changing customer-facing roles themselves

The themes alone are telling. A company that had just raised at a $1.5 billion valuation chose to send its head of customer success, not a product demo, to talk to this audience. In 2026, the scarce skill in AI is not building the model. It is getting a busy professional to change a habit.

The hosting team from Turing was Sanjit Kotwal, Josh Halliday, Mariana Raso and Dhanasree Rao. Turing has not published an attendance figure, so I will not guess one.

What it means

In my years in strategy and operations, and now at AI Central with an audience of 300,000+ readers, I have seen the same pattern again and again. People sign up for an AI tool, try one thing, get a good result, and then fall back to the old way within a month. Not because the tool failed. Because nothing in their week changed to make room for it.

This evening was a thoughtful model of the fix, and I think it applies well beyond Turing's fellows.

For senior professionals

The good news first. The skill an agent like Codex rewards is delegation: scoping a job, setting the standard, reviewing the output. That is a management skill, and senior people already have it. Fifteen years of knowing what good work looks like in law, finance or medicine is exactly what you need to direct an agent well. A junior person with perfect prompt tricks and no judgment will get worse results than you.

The harder news: you have to actually change your routine. Pick one recurring task you dislike. Hand it to an agent every time for a month. Measure whether you still do it that way in week four, not whether it impressed you on day one.

For founders

Your adoption plan matters as much as your model choice. Granola's growth curve and its decision to put customer success on stage say the same thing: retention is the product. If you roll out an AI tool to a team without a named owner, a first use case and a check at the one-month mark, you are buying a demo, not a capability.

There is also a hiring signal. Customer success at an AI company now looks less like support and more like change management. The people who can sit with a team and redesign how it works are the people who will make or break your numbers.

For media and community builders

Look at the format Turing chose: a small, senior, invite-only room, a hands-on half and a candid half, in a hotel rather than a conference hall. That is the opposite of the big-stage AI keynote. It works because the audience leaves having done something, not just having heard something. For AI Central, that is a clear note for how we run our own events and how we cover others: the value is in the doing.

Key takeaways

  • Give agents tasks, not prompts: Codex is built to take a job with a goal and a definition of done, so write the brief as you would for a colleague

  • Parallel work is the shift: the gain comes from running several agent tasks at once while you review, not from faster single answers

  • Delegation is the new AI skill: scoping and reviewing work beats prompt tricks, and senior professionals already have it

  • Adoption is the hard part: plan how people will change their routine before you roll out a tool

  • Measure week four, not day one: retention after the first try is the number that tells you a tool works

  • Customer success is now change management: AI is reshaping customer-facing roles, not only engineering

The close

What stayed with me most was the shape of the evening. One half showed what an agent can do. The other half asked whether people will let it. Most companies invest heavily in the first half and very little in the second, and I suspect that is where much of the value of AI is quietly being lost.

Thank you to the Turing team, and to OpenAI and Granola, for a generous and thought-provoking evening. If you want a place to start tomorrow, pick one task, give it one owner, and check after a month that people still do it the new way. I'd love to hear how it goes.

New to Codex? Start with our step-by-step guide How to Set Up Codex in ChatGPT: 10 Steps. We publish practical AI agent tutorials every week. Join 300,000+ readers.

AI Central covers AI, tech and startup events, from global stages like London Tech Week to local meetups and founder dinners. AI doesn't have to be the only topic on the agenda. If there's an AI angle worth sharing, we bring it back to our readers

Running an event, big or small? Write to [email protected] and tell us about it