“Burn tokens, not headcount.” It’s a slogan, and it comes from Silicon Valley — the same family as “move fast and break things”. Like all good slogans it carries something true and a lot of baggage that doesn’t translate cleanly to a UK accounting firm.
The true part: for the first time in the history of the profession, the choice between adding capacity by hiring a person and adding capacity by spending on compute is a real, specific, comparable choice. It is something a finance director can put on a slide. It is no longer a metaphor. That is what makes it worth taking seriously.
Where the slogan comes from
In a software startup, headcount is the dominant cost. A founder who can solve a problem with £5,000 a month of API calls instead of a £60k engineer has made an obvious decision. But a first-year trainee in a UK firm is not a software engineer. They are cheaper, they’re inside a three-year training contract that is also a regulatory pipeline to qualification, and they generate billable revenue early. You cannot simply not-hire them and rent GPUs instead — the training pipeline collapses, qualification numbers fall, and five years later there are no managers. So the slogan doesn’t work as a literal instruction. It works as a forcing question: what proportion of your capacity should come from each, what is each genuinely good at, and is the current ratio the one you’d choose if you were starting today?
What is actually comparable
Compute and headcount are now denominated in the same currency. A trainee costs a firm roughly £40–50k a year all-in, once you factor salary, qualification, study leave and supervision. A meaningful enterprise-grade AI deployment costs somewhere between a few thousand and a few tens of thousands per active user per year, and that cost is falling roughly an order of magnitude every couple of years. These are not equivalent units — a trainee does things AI can’t, and vice versa — but they are now in the same conversation, on the same spreadsheet, in the same monthly P&L review. That is the new fact.
The honest question: if next year’s plan is four trainees, two qualifieds and a manager, what would the same money buy in compute — and what would the firm look like at three trainees, one qualified, one manager and a serious AI deployment? For most firms the right answer is not zero of either column. But the current answer is last year’s plan with an AI line item bolted on top. That is the legion structure, expressed as a budget.
Drowning in manual, repetitive work? Tell us the task and we’ll show you what to automate.
What does not appear on the spreadsheet
The training question is the largest. If trainees spend less time on the preparation work AI now does, what are they learning by, and how do they reach competence in three years? A firm that simply removes trainee prep and assumes the pipeline sorts itself out is making a serious mistake; a firm that designs a deliberate alternative pathway is not. The client-trust question is second — SME clients hire a relationship as much as an output, so be honest about what stays human and make it visible. The regulatory question is third — the professional bodies and PI insurers are still working through what AI-assisted work means for review and liability. None of these is a reason not to ask the question. They are reasons to ask it properly.
The honest comparison
Look at next year’s recruitment budget and next year’s potential compute budget. Treat them as the same currency, because they now are. Ask which work should sit on each side of the line. For most mid-tier firms the answer is neither zero recruitment nor last year’s plan plus a tool — it is somewhere in between, and it depends on your mix of compliance and advisory, your client segments, your succession plan and your honest view of what the firm is for. Properly translated, the slogan is: “be deliberate about the mix of tokens and headcount that actually serves your firm and your clients, instead of inheriting last year’s mix.” Less catchy. More useful.
Part 2 of The Self-Improving Firm. Daniel Lawrence is the CEO and co-founder of Bots For That.
