The ICAEW published the third edition of its mid-tier firms research last week, and one pair of numbers stuck with me. Of the 35 firms surveyed, 40% plan to hire fewer graduates and 49% plan to hire more school leavers (Accountancy Today, 2026). Read properly, that is a firm quietly rebuilding its cost base around AI, and it is a move I can't copy, because I don't employ anyone.
TL;DR: Mid-tier firms are absorbing the AI productivity gain by changing who they hire - fewer graduates, more school leavers, same headcount. A solo practice has no junior pyramid to rebalance, so the same shock arrives as a service-mix decision: take the gain as time, as revenue, or move up the value chain. I'm taking the third.
The headline numbers, and the one that matters
The research surveyed leaders at 35 mid-tier UK firms in February and March 2026. AI now sits in the technology strategy of 86% of them, and 95% expect to use more of it over the next three years (ICAEW, 2026). Two-thirds, 68%, think AI will cut demand for some early-career roles, yet 83% say that won't mean fewer roles overall, and 71% say AI will let them move up the value chain.
The figure I keep returning to is a quieter one: only 17% feel they can actually assess what AI will do to their workforce. The other four in five are acting on a guess. Against that uncertainty, the graduate-to-school-leaver swap is the most concrete bet in the whole report. It tells you what these firms believe even when they say they can't model it.
Why can mid-tier firms hire around AI when a solo can't?
Because a mid-tier firm has a junior pyramid and I don't. The lever they're pulling is labour-mix substitution - replacing one type of worker with a cheaper or differently-skilled one to do the same volume of work. A firm with fifty staff can swap 22-year-old graduates for 18-year-old school leavers, train them alongside AI from day one, and hold output flat while the wage bill falls.
That works because graduate cost, senior cost and AI cost all sit on the same profit and loss account. The firm can reshape the mix between them. I am the entire labour input in my practice. There is no junior layer to trade down, no pyramid to reshape. When the same productivity shock reaches my desk, it can't express itself as a hiring decision, because there is no one to hire or not hire. It has to come out somewhere else.
So what lever does a one-person practice actually have?
Service mix, not hiring mix. When AI makes the mechanical work faster, a solo has three honest ways to take the gain. Take it as time: same clients, fewer hours, protect the evenings. Take it as revenue: more clients at the same fee, fill the freed hours with new work. Or take it as repositioning: same hours, but shift the work itself from low-value compliance to higher-value judgment.
The first two are comfortable. The third is the one that compounds. When I tested an AI close agent earlier this month, the categorisation and reconciliation it did well was exactly the work I bill least for. The work it couldn't touch - should this cost be capitalised, what's the dividend timing for a March year-end, does this director's loan need writing off - is the work clients actually keep me for. The shock didn't threaten my practice. It told me which half of it to grow.
What does "moving up the value chain" look like when it's just me?
It means drawing a deliberate line between the work I hand to AI under supervision and the work I keep. In my practice that line sits at judgment. Bank-feed coding, draft reconciliations, first-pass variance analysis - AI produces a credible junior's draft, and I review it the way I'd review a junior's. Anything that touches a tax position, lands on a CT600, or carries my name to HMRC stays with me from start to finish. That line matters more now HMRC is reading across returns with its own AI, where thin file notes get spotted faster than they used to.
The pricing has to follow the work, or none of this pays. Hourly billing punishes me for getting faster: if AI halves the time, hourly billing halves my fee for the same result. Fixed and value-based pricing for advisory work breaks that link, so the productivity gain lands with me rather than leaking back to the client as a smaller invoice. I've written before about reaching for the simple tool before the clever one, and the instinct is the same here. The tool earns its place by freeing me for the judgment, not by quietly doing the judgment unsupervised.
What the survey leaves out: most of the profession isn't mid-tier
The research is sound, but it's a portrait of 35 firms that look nothing like mine. Around 80% of chartered accountancy firms are small businesses employing four people or fewer, and roughly 18,000 sole practitioners work in the UK, against fewer than 200 large or mid-tier firms (ICAEW, 2025). The mid-tier playbook assumes scaffolding a solo doesn't have: a pyramid to rebalance, the option of private equity (46% of mid-tier firms now carry PE backing, up from 25% a year earlier, per Accountancy Today, 2026), the headroom to consolidate.
None of that makes the research wrong. It makes it incomplete for the bulk of the people reading it, since the firms outside its sample frame are the ones who can't pull its main lever. The strategic question for a mid-tier firm is "what should our hiring mix be". For me it is narrower and harder: which half of my own work am I still the best person to do, and what happens to my fees when a machine can draft the other half overnight?
Frequently Asked Questions
What did the ICAEW mid-tier firms research find about AI and hiring?
Surveying 35 mid-tier UK firms in February and March 2026, the ICAEW found 86% have AI in their technology strategy and 95% expect to use more over three years. On hiring, firms forecast a 40% fall in graduate recruitment and a 49% rise in school-leaver hiring, while 71% expect AI to let them move up the value chain. Only 17% felt able to assess AI's actual workforce impact (ICAEW, 2026).
How does AI affect a solo accountant differently from a mid-tier firm?
A mid-tier firm can absorb the productivity gain by changing its hiring mix, trading graduates for school leavers while keeping output steady. A sole practitioner has no junior pyramid, so there's no hiring composition to adjust. The same gain has to come out as a service-mix decision instead: fewer hours, more clients, or a deliberate move from compliance work toward advisory and judgment work.
What does "moving up the value chain" mean for a small accounting practice?
It means letting AI handle the mechanical work under supervision - coding, reconciliations, draft analysis - and concentrating your own time on the work it can't do safely: tax positions, planning, anything requiring judgment about a specific client. In practice it usually requires moving off hourly billing toward fixed or value-based fees, so getting faster doesn't simply shrink your invoice.
Will AI reduce the number of accountants in the UK?
The mid-tier firms surveyed don't think so: 83% expect no reduction in overall roles even as 68% expect lower demand for some early-career work, and ICAEW chief executive Alan Vallance noted demand for accountants remains high while the nature of junior roles changes (ICAEW, 2026). With around 90% of UK micro-businesses already using an accountant, the work is shifting rather than disappearing.
Should a small business worry that its accountant is using AI?
Not if the accountant is clear about where the line sits. AI doing first-draft bookkeeping and reconciliation under review is no different from a junior doing it, and arguably leaves a better audit trail. The thing to ask is whether anything touching your tax return or a real judgment call is being signed off by a person who understands your business, rather than waved through by a workflow that doesn't.
If you're a small business owner wondering what your accountant should and shouldn't be handing to a machine, get in touch - I'm happy to walk you through where I draw that line and why.