Leading change through the AI shift: what firm leaders are actually doing

At the Digital Accountancy Show this year we hosted a fireside panel on the hardest part of adopting new technology, which is rarely the technology itself. As Henrik Grim, our CEO, put it in the opening: almost anyone in the room can buy the tools on the stands. What you cannot buy is leadership, setting a direction, making the call, and bringing a team with you on the journey. Three firm leaders joined him to talk openly about how they are doing exactly that: Lee Maughan, CEO of Green & Purple, now part of the Affinia Group; Amy Thrower, who heads outsourcing at Albert Goodman; and Cory Mason, who heads innovation at BK Plus.

Here are the takeaways worth keeping.

Everyone is using AI, just in very different places

Henrik's first question was blunt: cut the hype, what are you actually doing? The answers showed how uneven the picture still is, even among firms that are clearly ahead. Lee has embedded ChatGPT across a team of 40, with people using it every day, the result of a decision made two years ago and stuck with since. Amy's firm has gone wider, automating bookkeeping, using UiPath to extract data from repetitive sources such as NHS doctors' statements, and using Claude to pull together tax packs for clients. Cory's firm is split: early days on accounts, bookkeeping and tax, but well ahead on audit, where all 200 auditors now use AI in their day-to-day work. The honest summary is that adoption is real, patchy, and still has a long way to run.

The decision still has to be someone's

Lee has gone from running an independent firm to being part of a group that has grown from 500 people to 1,600 in ten months. His point was that consolidation does not stop you making decisions, it just means more people need to understand them. The risk is not that you cannot get an answer, it is working out who should be involved. His approach is to keep acting as a firm owner would, and to move in smaller batches. Green & Purple is a Mimo user, and Lee's own example made the point: he would not roll our VAT review out across 1,600 people in one go, but he can prove it with a team of 60 first. Decisions, he said, should still land within a week or a month, not drift for longer.

Adoption lives or dies with the team

This was the thread running through the whole session. Amy was refreshingly honest that her first big change, automating data entry, was a decision she made largely on her own. Some of her bookkeepers do the work because they enjoy data entry, so telling them it would be automated was uncomfortable, and buy-in was slower as a result. Her lesson has reshaped how she works: a small group now drives each trial, feedback goes into a central log, and the team meets weekly to push the software forward. By the time something reaches the wider team, it comes from their colleagues rather than from her. Cory built a network of digital champions across every office to do the same, and learned that champions without a clear plan drift, in his case heavily towards tax. Both now set the goal before they pick the people.

A good pilot is not an extended demo

Amy shared the clearest filter of the day. Anything new has to be scalable first, because inconsistent systems across clients undermine the service. After that it has to answer at least one of three questions: does it reduce risk, does it make us more efficient, or does it add value to a service line we do not offer today. If it does not, it does not get her time. Cory's test is operational: a good pilot has an owner, a plan for who uses it and on which clients, and an honest read on your own book of work. His sharpest signal is the simplest one. When the team quietly stops using something, the pilot is dead. He described pulling ten licences for a tool that added a step to the client workflow and getting none of them back.

Why rollouts stall after the decision is made

Even a good decision can fail to land. Cory's view is that it usually means managers have not built the tool into the workflow, and that accountability matters: if someone has a licence and is not using it, they should not have it. Amy's cautionary tale was a six-figure investment trialled in one department, hers, from one department's perspective, which then did not fit the rest of the firm. Lee was candid that one of his own rollouts stalled over the summer despite his backing, and what fixed it was a concerted half-day effort to force it into real client files. Sometimes, he said, you simply have to shove it in.

Training for a job that keeps changing

The hardest question had no neat answer. If software starts doing the review, how do you train people to review? Amy's position was that nobody knows yet what the next three to five years look like, so the real skill is adapting quickly, and training people for the role they are actually doing now. Lee expects a chunk of future training to move in-house, and for regulators to rethink how the basics are taught. Both agreed the work at senior and manager level will grow as data entry falls away, which makes bringing people through more important, not less.

One thing to do in the next 30 days

We closed by asking each panellist for a single concrete step. Cory: come to events with your own problems to solve rather than letting vendors tell you what you need, and back the people already doing the work well, because they will go further and others will follow. Amy: share the strategy with your team, openly and often, because people who are anxious about their role will not come with you unless you bring them in. Lee: take the next easiest thing, a note-taker, payment processing, whatever moves everyone along, and just have a go, because very little of this is right forever anyway.

If there is one line to leave with, it is Amy's: if the team does not come with you, it will not work. The technology is the easy part. The leadership around it is where change is actually won.