Transaction coding and bank reconciliation
Suggested nominal codes and VAT treatment learned from your own historical decisions per client, surfaced for review rather than posted silently.
Industries
AI for practices where the file still has to stand up
Practices are being squeezed from both directions: Making Tax Digital pushing compliance volume up, and recruitment making the junior capacity to absorb it harder to find. AI helps most in the preparation layer, where the work is high-volume and a reviewer already checks it.
The pressure right now
What we hear from accountants
- Making Tax Digital for Income Tax expanding the volume of routine filings
- Junior recruitment and retention limiting the capacity that absorbs compliance work
- Fixed fees on compliance work colliding with rising cost per file
- Client onboarding and AML checks consuming senior time
- Advisory work - the part clients value - repeatedly deferred by compliance deadlines
Where it pays off
Six places AI earns its keep here
Not everything on this list will apply to you. Most organisations start with one and extend once it has been measured.
Year-end file preparation
Assembling the working papers, chasing the gaps and drafting the schedules, so the manager opens a file that is 80% built instead of empty.
Client query drafting
The "please explain this transaction" emails written from the actual ledger context, batched, and sent once a person has read them.
Onboarding, KYC and AML
Document collection, identity check assembly and risk assessment drafting, with the MLRO decision and the record-keeping unchanged.
Management accounts commentary
First-draft narrative from the numbers - variances, trends, the things worth asking the client about - for the partner to sharpen and sign.
Inbox and deadline triage
Routing client correspondence to the right person with the deadline context attached, which is where most practice time quietly leaks.
Where we would start
The first three moves
1
Pick the highest-volume compliance line
Usually bookkeeping or year-end prep. We baseline the current hours per file before changing anything.
2
Redesign around the review point
The AI drafts, the qualified person reviews. We design the review to be fast and evidenced, not a second full pass.
3
Measure per-file economics
Hours per file, rework rate and realisation, before and after. Fixed-fee work makes this measurement unusually clear-cut.
Risk and regulation
The part most suppliers skip
Where the risk sits
- Client confidentiality and the terms of any tool processing client records
- Professional ethics obligations under ICAEW, ACCA or CIMA codes
- AML obligations: the decision and the record must remain demonstrably human
- Reliance on an unchecked output finding its way into a filed return
- Engagement letters that predate AI use and do not disclose it
How we handle it
Nothing is filed, submitted or sent to a client without a named person approving it, and the audit trail records who approved what and when.
Where client data cannot go to a third-party service, we deploy privately - inside your tenancy or on your own infrastructure - rather than compromising the constraint.
We update the engagement letter wording and the practice's AI policy as part of the work, so the disclosure position is straight before the tooling goes live.
Where this fits
Where to go next
Questions
Questions from accountants
Yes - the assistance sits around the practice software you already run rather than replacing it. Which integration is practical depends on the product's API, and we establish that before scoping.
It can prepare them. We do not build unattended filing: the reviewer's sign-off is both a professional requirement and the thing that makes the rest of the workflow safe to speed up.
Exactly what happens: which processor, in which jurisdiction, under what terms, with what retention. We produce that disclosure alongside the build, and where clients will not accept a third party, a private deployment removes the question.
Start with an audit of what you already run
Two to four weeks to an evidenced picture of your AI use, spend and risk - and a ranked list of what to do first.