The 10x question

At Xerocon last week, Alexis Kingsbury, a non-accountant and author of Accrual Intentions, the story of building what he describes as the world’s first 100 per cent AI accountancy practice, highlighted an important point around the use of AI. The real benefit of AI is not the 10 per cent efficiency gain on work you already do. It is the 10x solution: the ability to deliver something of a different order to the people you serve.

The distinction matters because a significant portion of the AI conversation happens at the 10 per cent end. A faster draft, a quicker reconciliation, an hour saved here and there. Those gains are real, and they arrive almost on their own, because every software product is now shipping them. On their own, though, they change very little. The same processes run, slightly faster, and the week fills back up.

The 10x arrives in two ways, and the first is what you do with the time. Kate Hayward, Xero’s UK managing director, put a number on it from the mainstage: only 6 per cent of UK accountants and bookkeepers think AI will reduce headcount, while 94 per cent believe it will “unlock capacity to deliver higher value services”. The same logic holds inside a founder-led business. The routine that gets automated is not the prize. The prize is the attention it releases, spent deliberately on the bigger questions: where the margin is actually made, which channel deserves more investment, what the forecast cash can allow for the next quarter.

The second way is solving problems that could not be solved before. A session on the second day, The Rise of the Builders, showed what that looks like. Members of the Xero product team built a working app live on stage in minutes, using natural-language tools and no code that anyone in the audience had to understand, in a room full of accountants and bookkeepers rather than developers. The framing was buy, advise, build: check the app store first, advise on what already exists, and build only when nothing does. What has changed is that the third option is now open to people who could never have taken it before.

The proof came from a practitioner on stage: a firm with no engineering background that built its own payroll tools, starting with ChatGPT and working through the errors, and now runs client work through software that could not be bought. The value sits in the problem being removed, and the cost of removing it has collapsed.

Neither route works without a person in the middle. The slide behind the opening keynote put it plainly: automate the routine, expert in the loop. These are probabilistic models, and what they produce is a draft until someone qualified has checked it. The session’s other discipline holds too: start small, deploy early, and improve in use. But the human review is part of the build, and the judgement about what the output means stays with a person.

For a founder, the lesson is in the questions you ask of the technology. The 10 per cent question is how to do the current process faster. The 10x question is really two: what could be made unnecessary altogether, and what the freed time should be spent on. In the finance function that might be systems that agree because they are connected, rather than because someone spends a day each month making them agree. And it might be the afternoon that used to go on crunching the numbers, spent instead on what the numbers mean. That’s where the true value comes from.

The gains at the 10 per cent end will keep arriving on their own. The 10x change starts with someone asking the bigger question of each part of the business, and for most founder-led businesses the finance function is a great place to start, because it touches almost everything else.