We're entering a new era in accountancy — one where efficiency is no longer the competitive edge, because it's assumed. The real edge? Blending cutting-edge AI with high-touch human insight. It's not about replacing accountants; it's about giving them superpowers.
That's the optimistic framing, and I believe it. But there's a deadline attached to it now, and most firms haven't done the arithmetic on what it does to their capacity. So this piece covers both: the future worth building, and the maths that's arriving in the meantime.
Goodbye manual. Hello meaningful.
In the past, bookkeeping was about data entry, reconciling transactions, and chasing receipts. It was necessary but repetitive — and it drained the talent pool of time that could be spent advising clients or identifying opportunities.
Now, with tools like Xero, Apron, and AI-driven document readers, that manual labour is vanishing fast. Compliance runs in the background, bank feeds update in real time, and software flags anomalies before humans even log in. This isn't the death of accountancy. It's the rebirth of it.
The deadline nobody has costed — a worked example
Making Tax Digital for Income Tax starts on 6 April 2026 for sole traders and landlords with qualifying income over £50,000. It extends to £30,000 from 6 April 2027 and £20,000 from 6 April 2028 (GOV.UK, checked 28 July 2026). Qualifying income means gross income from self-employment and property before expenses — turnover, not profit — which is why more clients fall inside the threshold than firms expect.
Here's what that does to a practice. The firm below is illustrative, with invented client numbers, but the mechanics are exactly right.
A firm has 300 sole trader and landlord clients. Today each one produces a single Self Assessment return a year: 300 filing events.
April 2026. Say 120 of them have qualifying income above £50,000. Each moves to four quarterly updates plus a final declaration — five submissions instead of one.
- New events: 120 × 5 = 600, replacing 120. A net increase of 480 filing events a year.
- At 25 minutes of preparation, chasing and review per quarterly update, that's 200 hours of new work annually.
- At a loaded staff cost of £28 an hour, roughly £5,600 of new cost — with no new fee income unless the firm reprices.
April 2027. The £30,000 band pulls in perhaps another 90 clients. Another 360 net events, another 150 hours, another £4,200.
April 2028. The £20,000 band catches most of the rest.
By the third year that single firm has added something like 350–400 hours of pure compliance admin — two months of a full-time person — to serve exactly the same clients for exactly the same fee. Nobody grew. Nobody got better advice. The bill just went up.
There are three honest responses. Reprice — put the increase in front of clients and defend it. Automate — get records flowing from bank feeds so a quarterly update is a review rather than a rebuild. Refuse — decide which clients you no longer serve. Most firms will end up doing all three, and the ones who choose deliberately will do far better than the ones who find out in the January after.
That's the real argument for AI in accountancy. Not that it's clever. That the alternative is 400 hours you can't bill.
AI handles the routine. Humans handle the relationships.
The rise of AI in accounting isn't just about automation — it's about augmentation. AI does the heavy lifting so our people can do the higher thinking. That means more proactive conversations with clients; real-time forecasting and scenario planning; deeper advisory work around pricing, profitability and growth; and most importantly, a better client experience — one that feels personal, not transactional.
Applied to the example above, augmentation is specific rather than vague. If a quarterly update drops from 25 minutes to 8 because the records are already clean, 480 events cost 64 hours instead of 200. The work doesn't disappear — it changes shape, from assembling to checking. That's the whole game.
What the modern accountant looks like
The accountant of the future won't be buried in spreadsheets. They'll be strategists, sounding boards and growth partners. They'll understand the numbers and the person behind the business. To make that possible, we need to build systems that remove friction (seamless onboarding, smart proposals, zero data duplication), empower advisory (with AI surfacing insights, not just data), and prioritise service (delivered with empathy, speed and clarity).
In practice, the modern firm has four things the traditional one doesn't:
- Clean data at source. Bank feeds, receipt capture and rules that run without a human, so the ledger is right on the 5th rather than reconstructed on the 25th.
- A defined advisory product. Not "we're happy to have a chat" — an actual thing with a name, a scope and a price, that a client can buy.
- A cost to serve low enough to say yes. This is what makes it possible to serve the 3.2 million sole traders in the UK at all, rather than leaving them underserved.
- A named human per client. Every automation gain gets spent on relationship, not on headcount reduction. Firms that spend it the other way get very efficient at losing clients.
Where firms get this wrong
- Buying tools before fixing process. Automating a messy workflow produces mess at speed. Standardise first, automate second.
- Passing the tool cost to clients without the benefit. If your fees go up and their experience doesn't change, you've sold them your problem.
- Letting AI be the last check. Nothing generated by a model should reach HMRC, a client or a lender without a named person having reviewed it. That's not caution, it's the job.
- Cutting juniors because AI does their old work. Then nobody learns to spot when the output is wrong, and in five years you have no seniors.
- Waiting for certainty. The thresholds and dates are published. A firm that starts moving clients onto digital records in the year before its band bites has an easy transition; one that starts in the month it bites does not.
- Ignoring the client's side of it. Your client is the one who has to keep digital records. If they're still handing you a carrier bag, your software doesn't matter.
Do this in the next seven days
For firm owners, five concrete steps. The first one takes an afternoon and is the one everything else depends on.
- Segment your client list by qualifying income — gross self-employment and property income, before expenses. Three buckets: over £50,000, £30,000–£50,000, £20,000–£30,000. That tells you exactly who lands in 2026, 2027 and 2028.
- Count the filing events your 2026 bucket creates and multiply by your honest minutes-per-update. That's the hours number. Do not estimate it optimistically.
- Cost those hours at a loaded staff rate. Now you know what MTD costs your firm, in pounds, before you talk to anyone about price.
- Pick the ten clients with the messiest records in the 2026 bucket and get them onto bank feeds now, while there's no deadline pressure.
- Write your repricing letter this quarter, not next. Clients accept a fee change explained six months ahead far more readily than one explained six days ahead.
Where we're headed
The firms that thrive over the next decade will be those who embrace AI and elevate humanity. The ones who see technology not as a threat, but as a tool — one that enables better decisions, richer conversations and more meaningful work. The individual version of this argument, for anyone who isn't running a practice, is how AI helps brilliant people shine brighter. And if the honest answer is that your firm needs an operator in the engine room to make the change actually happen, that's the work I do.
We're not just making accounting better. We're making it matter more.