As with the other case studies on this site, this is a composite scenario built from patterns across accounting and bookkeeping clients rather than one named practice — client financial data is confidential, so the specifics here are illustrative.
The firm: a seven-person accounting and bookkeeping practice serving around 180 small business clients across Yorkshire, split between statutory accounts work, VAT returns, and ongoing bookkeeping support. Like a lot of practices this size, they had already bought into several AI tools as part of their software stack — and were using almost none of them properly.
The Problem: The Tools Existed, the Value Didn't
The practice had Copilot licences sat mostly unused, a couple of AI features inside their practice management software nobody had configured, and a general sense that they were "behind" on AI without a clear idea of what "ahead" would actually look like. Meanwhile, the actual daily friction was mundane and specific: new client onboarding involved chasing the same list of documents by email, one client at a time, often multiple times. And the office fielded a steady stream of client queries that were, almost without exception, some variant of the same ten questions — when's my VAT due, what do you need from me this quarter, has my return been submitted yet.
What Was Actually Built
Onboarding and document chasing. New client onboarding now runs through a structured checklist that automatically emails and, where appropriate, texts clients with exactly what's needed, in plain language, with a clear deadline. Follow-up reminders go out automatically to anyone who hasn't responded, at sensible intervals, without anyone in the office needing to track who's chased and who hasn't. The office team's job shifted from manually chasing every document to handling the exceptions — the clients who need a phone call, not another email.
Query triage and FAQ handling. A simple AI-powered assistant, trained on the practice's own client FAQs and general (non-advisory) information, now handles the first response to the most common client questions — deadline dates, what documents are needed for a given service, general process questions. Anything touching actual financial advice, tax position, or client-specific numbers gets routed straight to a qualified team member; the assistant is explicitly scoped to never answer those.
What Changed
Document chasing went from a manual, easy-to-drop task to something that happened reliably in the background, which mattered most during the run-up to filing deadlines when the office had previously been firefighting. The office manager estimated getting back four to six hours a week that had been going into repetitive chasing emails.
Query handling didn't replace the team's client relationships — if anything, it protected them. Common questions got answered faster (including outside office hours), and the team's actual conversation time with clients shifted toward the questions that needed a real accountant's judgement, rather than repeating deadline dates for the hundredth time.
What Stayed Firmly Human
No AI system touches a client's actual figures, gives tax advice, or represents anything as the firm's professional opinion. The assistant's scope was deliberately narrow — general information and process questions only — and every escalation path defaults to a human. In a profession built on trust and professional liability, that boundary isn't optional.
The Real Lesson: Configuration, Not More Tools
The most telling part of this pattern is that the practice already owned most of the tools needed before we started. The change wasn't buying new AI software — it was configuring what existed around two specific, well-defined problems instead of leaving generic AI features switched on and unused. That's usually where the real waste is in accounting practices already "doing AI": subscriptions paid for, value not captured.
If you want an honest look at whether your practice is in that position — paying for AI tools that aren't actually doing anything — a free AI audit is a good place to find out.