The Operator Letter

This letter is adapted from a keynote given at Beyond Active in Madrid on 24 June 2026.

Right now, somewhere within ten kilometres of one of your clubs, a prospective member is looking for a gym. She is not on Google and she is not scrolling Instagram. She has opened ChatGPT on her phone and typed: find me a gym within 15 minutes of my home and 15 minutes of my work, that has a HYROX training area, evening reformer Pilates classes, and isn’t known for teenagers taking selfies in the squat rack.

The agent is now reading your website, your competitors’ websites, your Google reviews, and your class schedule. It is forming an opinion about your business that you have not been consulted on. In about ninety seconds it will produce a shortlist of three gyms. Yours may be on it, or it may not. She will join one of those three. And at no point in the entire process did anyone from any of those gyms have any influence over the outcome.

The website you spent tens of thousands redesigning last year? She never saw it. The tour your team would have walked her through — that carefully calibrated thirty-minute journey from front desk to changing rooms to the moment of close? That tour does not happen. She is not coming for it. Her agent has already done that work, faster, comparing all of you in parallel, before she has even put her phone down.

That agent does not work for you. It works for her. But it belongs to the same wave of technology that is about to be working inside your business, this time on your side of the desk.

What we mean when we say “agent”

When I say agent, I do not mean a chatbot, or automation, or a slightly cleverer version of the software you already have. An agent is software that you give a goal to. It then decides how to achieve that goal, takes actions across whatever systems it needs, and reports back when it is done. This is completely different from anything we have had before. Software used to wait for you to tell it what to do, step by step. An agent waits for you to tell it what you want, and then it works out the steps itself.

Let me give you two pictures.

The first is an agent that handles every prospect lead the moment it arrives, twenty-four hours a day, in any language, on any channel. Not by sending a templated email, but by reading what the person actually said, working out what they actually need, checking your live class schedule, suggesting a relevant trial, booking it, and following up if there is no response. All done at the point of maximum intent — the moment the lead has decided to enquire — rather than three hours later when someone gets to the inbox. Every lead, every time, every language, consistently.

The second is an agent that runs through your member base each morning and finds the people who are quietly on their way to cancelling, often long before they know it themselves. Not the ones already heading for the door — by then it is too late — but the ones who still look completely fine on the surface. It has learned what that early drift looks like by working through a vast history of member journeys, weighing millions of combinations of behaviour against who stayed and who eventually left. It reads dozens of signals at once: how often they come, how that is changing, how they use the app, how they book, how they pay. No single signal gives anyone away. It is the pattern across all of them, weighed together, that lets it say — with a precision no human could match — which members will still be here in twelve months and which will not. Then it builds a personalised intervention for each one, routes the cases that need a human to a human, and handles the rest itself.

Both of those exist. Both are deployed in fitness operators today. Neither is hypothetical.

Digital debt repayment is not transformation

Which brings me to where I think the industry is confusing two very different things. Much of what is currently being sold as AI transformation is not really transformation. It is digital debt repayment.

Look at how AI is being talked about across the industry: automate your sales process, streamline your member engagement, optimise your operations, reduce front-desk workload, help your team do more with less. Every one of these headlines describes real value, and if you have twenty years of accumulated operational debt sitting in your business, paying it down matters. But it is not a new way of running a business. It is plumbing. Automating a follow-up email is not transformation. Neither is putting a chatbot on your website to answer the same ten questions all day, or writing a member newsletter in twelve seconds instead of two hours.

Transformation is when the agent stops patching your function and starts performing it. It does not write the email for someone to send — it runs the whole prospect conversation, decides who to escalate, learns from how each one lands, and does the next one better. It does not hand you a retention report — it finds the at-risk members, designs the intervention, runs it, and tells you afterwards what worked. That is a different category of thing. And the gap between operators who get to that and operators who do not is not going to be a 10% efficiency gap. It is going to be the difference between businesses that can run at fundamentally lower cost, with fundamentally more responsiveness, with fundamentally better consistency. Once that gap opens, it does not close.

What 901 audits taught us

Three months ago, I built an agent to find out where the market as a whole actually stood. The brief was simple: audit the AEO readiness of every meaningful fitness operator globally. AEO — Answer Engine Optimisation — is the next version of SEO, measuring how your business shows up when an AI like ChatGPT or Perplexity is asked to recommend a gym.

In two hours, the agent audited 901 fitness operator websites across 23 countries. For each one, it analysed the website, evaluated the content structure, checked the underlying schema, and wrote a personalised readiness report. Then it built each report into its own web page, published everything to unique URLs, pushed the data into the CRM, and triggered a personalised outbound campaign to leadership at every operator audited. Two hours, start to finish, audit to inbox. Ten years ago, that would have cost a fortune and taken a team of consultants months, if it were possible at all. I did it on a Tuesday, alone, with one agent, in two hours. I am telling you that not because I am special, but because the tools have changed.

The speed is actually not the headline, though. The headline is what the audit found. Of 901 operators across 23 countries, the number that scored as ready to be found and chosen by agents was in single digits. Single digits out of 901. What that number tells you is that the whole industry is standing at the same start line. A new way of being found and chosen is arriving, and almost no one has built for it yet. That is not a threat. It is the opening.

When the buyer stops being human

Let me come back to the woman looking for a gym, because we left her halfway through her decision. The agent has given her three options. She now has answers to all of her questions. She picks one.

Sit with what that means, because it is genuinely different to how things have operated. For your whole career, your customer has been a human you could influence. You could move them with a photograph, with a feeling, with a room that smelled right, with a salesperson who read the moment. The agent is a different kind of customer. It does not respond to atmosphere. It does not get talked round. It compares every option in parallel against an explicit set of criteria, and it does not forget what it found. You are no longer persuading a person. You are satisfying a system. Almost no one in this industry has built for that.

Today, she clicks through to your website, fills in a form, provides payment, and signs up. Very soon, she will do none of those things. She will just say to her agent: go ahead and sign me up at the second one. And the agent will. Credit card authorised. Membership active. Welcome email received. Induction booked. She will not have visited a website, spoken to a human, or gone through your funnel. Your brand may well have played a role in her decision — she trusts the name, she has heard good things — but the transaction happens through her agent. The first time anyone at your gym becomes aware of her existence is when she walks through the door for her induction.

You may think I am describing 2030. I am not. The capability is already live. Stripe, which handles a significant share of the world’s online payments, has built infrastructure that lets AI agents be issued payment cards and make purchases autonomously on their owner’s behalf. They have already partnered with Google, OpenAI and Meta to do exactly that inside their AI products. Visa and Mastercard are just launching. There are now browsers being built specifically for agents to use, because the browsers we designed for humans are slow and clumsy when an agent is at the wheel.

So the infrastructure exists — the agent infrastructure, the payment infrastructure, the transaction infrastructure. The only thing missing is operators being ready to receive a non-human buyer.

Three things your business needs

Here is what your business needs to be, when the buyer is no longer human.

It needs to be visible. An agent has to be able to read your website, understand what you actually offer, and form an accurate picture of your business. But this is where a decade of good instincts can work against you. You have spent the last ten years making your websites more visual, more cinematic, more beautifully designed — hero videos, animated transitions, lazy-loaded imagery, JavaScript-rendered content that materialises only when a human scrolls past it. Agents are far less effective at interpreting any of this than a human is. They cannot reliably watch a video, wait around for JavaScript to render, or read meaning into your photography. What they read well is structured text, clean schema, clear copy. None of what you built was a mistake — it won you human attention for a decade — but the rules are changing, and the things you spent the most on are now the things the new buyer struggles to see.

It needs to be credible. When the agent checks your reputation, your reviews, your citations, your presence in trusted sources, what it finds has to confirm what you claim. Agents are very good at spotting the gap between marketing and reality. If your website says “best gym in Manchester” and the rest of the internet does not agree, the agent will notice, and you will not be recommended.

It needs to be accessible. When the agent decides to buy from you, it has to be able to complete the transaction. If your signup process requires three forms, a phone call, and a tour booking, no agent will get to the end of it. The friction you have spent years building into your funnel — to qualify leads, gather data, set up the upsell — is the friction that stops an agent completing the sale.

Most operators will not score badly on any one of these three. They will score badly on all three together, because each represents a habit the industry has spent a decade reinforcing in the opposite direction. And remember: there is no page two of an agent’s recommendation. The agent does not return ten results. It returns one, or two, or three. The funnel just collapsed into a podium.

The people inside your business

This is the part where conversations about AI often become careful. I am not going to be careful, because I do not think operators need me to be.

The biggest mistake people make here is to treat this as a cost story. It is not a cost story. It is a capacity story. The same people, producing more, responding faster, personalising at a level they could never reach by hand. It changes the shape of your workforce in three ways.

The first is that some functions will need fewer people over time. This is observable, not controversial. The operators who deploy agents into early-stage sales, first-line retention, and content production will need fewer humans in those functions. In practice, this does not necessarily mean layoffs. The fitness industry has high natural turnover. Most often, a vacant role is simply not refilled because the work it used to do is now done by an agent.

The second is that the people you keep become far more valuable. Right now, very often, your best people are doing your worst work — admin, chasing, data entry, the things they quietly hate and were never hired for. Hand that to an agent and you get them back for the job you actually hired them for. Your best membership consultant did not get into this industry to send follow-up emails at 9pm. She got into it to help people change their lives. The agent does the emails. She does the conversations. More of them, better, because she is not drained by the admin that used to sit underneath them. And she stays longer, because she is finally doing the job she came for.

The third is the one that can be disorienting. Your managers are about to start managing agents alongside people. You will brief an agent the way you brief a junior team member. You will review its output, set objectives for it, add it to projects and take it off others. This is not a metaphor. It is the actual work of a manager in 2027. Your org chart will reflect this — agents will appear on it, with names, with responsibilities, with reporting lines. Some of you will find that uncomfortable. Get past it quickly. An org chart is a division of labour. If part of your labour is being done by an agent, the chart should say so, or it is not an honest picture of how your business runs.

And one more shift, which might be the deepest. By 2028, and probably sooner, your people will start bringing their own agents to work. When you hire someone today, you hire them for what is in their head — their experience, their judgement. Increasingly, that value will live in the agent they have built alongside themselves: their methods, their decisions, their voice. Their professional operating system becomes portable. When they leave, they do not just walk out with what is in their head. They walk out with a system, and it becomes their next employer’s advantage. What does a non-compete even mean when half of someone’s value sits in an agent they own? Nobody has the answer yet.

How ready are you, really?

Let me finish with something every operator should be able to answer: how ready is your business, actually? Not how much you talk about AI. Not how often it appears in your board deck. How ready you are in reality.

The agentic workforce I have described is where this is going, but you do not get there in one jump. You get there by nailing the basics first — the foundations everything else is built on. In my experience, most operators have not yet nailed the basics. So here are five questions. Answer them honestly.

One. Has every person in your organisation been given a paid subscription to a capable AI tool? Not permission to use the free version on their phone. A subscription, paid for by the business, treated as essential equipment like a laptop.

Two. Has the CEO — and if that is you, has the version of you that walks into the office on Monday — publicly and repeatedly told the entire organisation how AI is expected to be integrated into every function? Not a memo. A repeated, visible, unmissable expectation.

Three. Does your business have a forum where AI wins are shared, celebrated, and learned from? A meeting, a channel, a recurring moment where the people doing the best AI work in your business are publicly recognised and their methods are spread.

Four. Are you measuring the impact of AI by function — time saved, quality improved, growth enabled — broken down by department? Or is “we are doing AI” the entire reporting line?

Five. Is there at least one person in your business whose explicit responsibility, with an explicit incentive, is to find out how AI can change what you do? Or is it the side project of whoever happens to be curious?

Count your yeses.

If you have four or five, you are genuinely ready. There are perhaps two operators in Europe I would put in that bracket today. They will not necessarily be the biggest, but they will be the ones quietly pulling away from everyone their size.

If you have two or three, you have started. Be honest with yourself about how deep it goes — for most operators in this bracket it is a licence here and a pilot there, not yet the way the business runs.

If you have one or none, you have not really started. The good news is that puts you alongside much of the industry, not behind it. The foundations I have described are a few months of intent, not a moonshot. The start line is right in front of you. The harder news is that the gap between the leaders and everyone else is widening fast, and the woman at the start of this letter is going to choose her gym next week. Her agent will not wait for anyone to catch up.

Your future agentic workforce is not a future state. It is a decision. Either you are designing it, actively and deliberately, or someone else is designing it for you.

The members coming through your doors over the next three years will be the most AI-native consumers in history. They will arrive having had agents do the searching, the comparing, and increasingly the buying for them. And here is the opportunity in that. Fitness is still a profoundly human industry. What agents strip away is the friction that gets in the way of it — the admin, the chasing, the slow replies, the work nobody ever joined this industry to do. The winners over the next decade will not be the operators with the most technology. They will be the operators who use it to put more time into people, more attention onto members, and better outcomes for both.

That is a future worth building.

I started this letter by telling you about a woman looking for a gym. I will finish by telling you the rest of her story. She joined the gym her agent recommended. The induction was excellent. The trainer was attentive. The facilities were exactly what she had been told they would be. She referred her sister. She referred a colleague. The operator never knew they had won her on the strength of being machine-readable.

But they had.