There was a session at a conference last week that I haven’t quite stopped thinking about.
The CEO of Mindvalley was talking about the future of learning. He described a small bone-conducting device, worn behind the ear, that listens all the time. It builds a mental map of who you are; your habits, your goals, the person you want to become. It listens to your meetings. It knows your worst habits better than your best friend does. And in a relationship argument, it whispers to you: slow down, take a breath.
The rest of the event was good: measured panels, honest conversations. But that one image stuck. It stayed with me on the train home, the next morning, and I’ve been chewing on it since.
The bit that gave me the ick, however, was the relationship argument.
Slowing down in an argument is the thing you need to learn. Taking a breath – that’s a muscle you build, by failing at it and trying again. The minute we let AI do that work for us, we’ve handed over the part of being human that’s most worth keeping. And more importantly, we’ve stopped getting better at it.
I’m not anti-AI – quite the opposite
In fact, we’re using a lot of AI at Wanstor, and it will be big business for us. The shift in what’s possible inside a managed technology business in the last 18 months is real, and we’re moving fast on it. At a different event the same week, Dominic McGregor put a phrase on stage that I want to borrow. Made by humans. That’s the line I want to hold.
I keep coming back to social media. We adopted it before we made the design choices, and we’re still living with that. I don’t want us to do that again with AI. Not for our customers. And honestly, not for my one-year-old son, who’s going to grow up in whatever world we build now. By the time he’s old enough to ask questions, AI will be where he asks them – and unlike us, he won’t have grown up sceptical of the first answer. That’s the bit that worries me. Not AI itself, but what gets atrophied around it if we’re not deliberate.
Which raises the practical question. So what does “made by humans” actually look like in a managed technology business in 2026? And equally, what does it look like for the customers we serve?
Inside Wanstor: making AI adoption deliberate
The first answer is the one we control: the way we approach AI, both for our customers and inside our own team.
Ross Hale and our AI team have built our offer around a single conviction. The way Ross frames it: licences are easy, everyday value is hard. Most users have Microsoft Copilot now. However, far fewer have confident, everyday use. Most organisations don’t have a technology problem – they have an adoption problem. The whole point of our Copilot Adoption Managed Service is to close that gap – a year-round rhythm of practical enablement, not a one-off training day. And our AI Readiness Assessment leads with what most readiness work doesn’t: surfacing oversharing, label gaps and tenant risk before Copilot exposes them. The deliberateness shouldn’t start when something goes wrong. It should start before.
Similarly, the third strand, and the one growing fastest, is the agents and automations we build and deploy directly with customers. Copilot Studio agents for internal workflows. Power Platform automations for the repetitive jobs that quietly eat the day. Custom integrations where the off-the-shelf doesn’t quite fit. It’s the bit of the AI conversation that’s moving fastest, and it’s where the value question gets sharpest – because once you can build a thing, the next question is whether it’s the right thing to build.
Measuring AI value: direct, indirect and people value
The frame we try to hold is to be clear, at the outset, which of three kinds of value we’re going after.
- Direct value – the £s on the P&L. Cost out, revenue in, hours saved.
- Indirect value – the risk reduction, the time-to-decision, the experience improvements that don’t sit neatly in a budget line but compound over time.
- People value – capability, capacity, the time given back to do the work only they can do.
Most well-designed projects hit all three. The ones that drift are usually the ones that didn’t decide which mattered most at the outset. Going after all three by default tends to mean going after none of them well.
One we’re proud of: World Travel Holdings reworked their reconciliation process with us. As a result, the business achieved a 77% reduction in operating costs – from around £130,000 a year down to around £30,000 – and a process the business now owns. That’s the direct value, named and designed for. The capacity it freed up to do work the business couldn’t afford to spend time on before – that’s the people value, which usually compounds further than the cost line itself.
Meanwhile, inside our own team, the same logic applies. Henry Deacon runs a monthly AI Super Users series. The most recent session was called Stop Making AI Slop, with a game called Spot the Slop where the team practises telling when AI has produced something flat, generic, fabricated, or just wrong. Rich Kuczma has built a custom skill that teaches Copilot to produce Wanstor-shaped documents from the first draft, so our people aren’t reformatting AI output forever.
None of this is grand. But it’s the work – the deliberate building of the habit that says we use AI to be sharper, not lazier. Our own AI team puts it more concisely in their proposition language: human-AI interconnect, not human replacement. That isn’t a slogan we wrote for the website. It’s how we’ve designed what we sell, and how we want our own work to feel.
For our customers
We work across three sectors – professional services, not-for-profit, and hospitality. At first glance, they look like very different businesses. But the question I keep asking is more or less the same in each. Where does AI free people up to do the thing only they can do? And where does it quietly take the muscle away?
Professional services: protecting judgement
This sector has the sharpest version of the question. In May, Anthropic shipped ready-to-run agents for the most time-consuming work in financial services: building pitchbooks, screening KYC files, closing the books at month-end. The model providers are moving up the stack, from general tools towards the specific jobs a firm used to bill for. Work that used to need ten years in the firm is increasingly available in ten seconds.
So the question every professional services CEO is starting to ask is: if everything can be done from behind a laptop, what do we provide as value? My answer is two things. The first is change management – AI doesn’t land on its own, and the real value is in helping the firm actually adopt it. Which workflows to keep, which to redesign, how roles shift, how to bring partners and associates along, how to hold the culture intact while the operating model changes. That’s a service problem, not a technology one. The second is the part of the work AI can’t replicate. Judgement. Taste. The specific way this firm serves its clients. The institutional knowledge that makes you worth picking. AI should be sharpening that, not flattening it into something the next model release can do for free. Which also means being careful about what we feed in. If we hand over the bit that makes us distinctive in exchange for a productivity bump, the bump won’t last.
Not-for-profit: creating capacity where it matters most
This is the one I feel most strongly about. Social workers, charity teams and the people propping up our public services do some of the most important and most undervalued work in our economy. Society has under-resourced these roles for decades, and the people doing the work have absorbed the gap by stretching themselves thinner. The work that drains the day in these roles is the admin tail. A social worker typing up visit notes between calls. A charity team finishing the day with another evening of grant reporting. A caseworker keying the same data into a different system for the fourth time. That isn’t the job. The job is the person on the other end of it, and the hours given to the paperwork are hours not going to them.
AI is the first opportunity I’ve seen to level that playing field – to give a small charity team genuine resourcing it couldn’t otherwise afford. But it only lands if we lean into the human, not replace it. The grant report can be drafted in fifteen minutes. The conversation with the beneficiary can’t, and shouldn’t. And the data underneath has to be honest, in this sector the cost of a wrong answer isn’t a missed quarter, it’s a person. That’s a service problem before it’s a model problem.
Hospitality: protecting the guest experience
Here the question flips on its head. Hospitality is fundamentally a human business. The value lives in how a member of staff makes a guest feel looked after – the unexpected detail noticed and remembered, the conversation at check-in, the read of the room. That’s the bit that turns a booking into a returning regular, and a returning regular into a recommendation. The work that drains the day in hospitality isn’t the work the guest sees. It’s the rota juggling, the stock checks, the post-shift paperwork, the same enquiry answered for the fortieth time, the food cost reconciliations, the rolling staff churn. The backstage maths.
The opportunity is to lift those things off the day, so the front-of-house team spends more time with guests and less time behind a screen. But the same rule applies. The minute AI starts trying to replicate the human bit, the value collapses. A bot that answers the front desk faster but coldly is a worse hotel, not a better one. A check-in process optimised for throughput is a check-in process that doesn’t notice the family on their anniversary trip. The technology should be lifting weight off the operation behind the guest, not stepping in front of it.
Where I’ve landed
I’m not arguing against AI. I’m arguing for being deliberate about it – for making the design choices on purpose, not by drift.
For our customers, this looks like more time with us, not less. More understanding of their business. Quicker response, sharper thinking, better-informed advice. AI under the bonnet. Humans where it matters.
For Wanstor, it looks like investing in the parts AI can’t easily replicate: judgement, taste, relationships, vertical depth, while letting AI take the operational weight.
And for my son? It means growing up in a world where he still has to do some things the hard way. Have difficult conversations, work through disagreements, solve problems that don’t have instant answers, read books all the way to the end, build the muscles that matter. Because technology will keep getting better. The question is whether we keep getting better alongside it.
Made by humans. That’s the bit I’m holding on to.
FAQs
What does successful AI adoption actually look like?
Successful AI adoption is when people genuinely change the way they work, not just when the licences have been deployed.
Most organisations can buy Copilot. Far fewer build the habits that turn it into everyday value. The organisations seeing the biggest returns are usually the ones investing in practical use cases, ongoing enablement and helping people become confident using AI day after day.
Ultimately, the goal isn’t AI usage. It’s giving people more time for the work that requires judgement, creativity and human expertise.
Why is Microsoft Copilot adoption harder than organisations expect?
Because technology is usually the easy part. The harder part is behaviour change. People need confidence, practical use cases and reinforcement over time. That’s why so many organisations have deployed Copilot, but far fewer have seen consistent everyday use across their workforce.
Will AI replace professional expertise?
I don’t think so. It will change it. AI can generate outputs. What it can’t replicate is judgement, context, experience and the specific way an organisation serves its customers. The firms that win won’t be the ones that replace expertise. They’ll be the ones that use AI to amplify it.
How can AI help not-for-profit organisations?
By creating capacity. Many not-for-profit teams spend too much time on administration and not enough time on the people they exist to help. If AI can reduce reporting, documentation and repetitive tasks, it gives that time back to where it matters most.
How can AI improve hospitality operations?
The best use of AI in hospitality is usually behind the scenes. Rota management, reconciliations, stock reporting, repetitive enquiries. The goal isn’t to replace human interaction. It’s to give staff more time to create the guest experiences that people actually remember.
What is the biggest risk of AI adoption?
Drift. Not making deliberate decisions about where AI should be used and where it shouldn’t. The technology will keep improving. The challenge is making sure we don’t lose the human skills, relationships and judgement that create real value along the way.