Few businesses are as densely packed with automatable work as an accountancy practice: recurring, deadline-driven, rule-governed, and spread across systems that do not talk to each other. That density is why the generic AI conversation underserves accountants, and why this page is organised by process rather than by product.
The processes, ranked by fit
Document collection and chasing leads every list: records requests, missing bank statements, unsigned letters of engagement, approaching deadlines. An agent watches the state, drafts each chase in your voice, and a person approves the sends; the pattern is described in the workflows guide and it lands hardest in January. Extraction is second: client paperwork, bank exports and receipts read into structured data for your working software, with provenance kept so review is fast; the form-filling and extraction pattern generalises across ledgers. Reconciliation third: two systems that should agree, compared continuously, exceptions surfaced with evidence. Drafting fourth: engagement letters, standard advice paragraphs, working-paper summaries and client explanations of technical points, always as first drafts under review. And meeting capture feeds all of it: calls transcribed and turned into file notes and action lists, with the consent rules from the meeting tools guide settled first.
The list against the calendar, because accountancy pain is seasonal:
| Process | Worst in | Automation shape |
|---|---|---|
| Document collection & chasing | December-January | The collection loop, approvals on sends |
| Extraction from client records | Year-round | Paperwork to structured data, human reviews |
| Reconciliation | Month-end | Exceptions surfaced with evidence |
| Drafting | Year-round | First passes under review |
| Meeting capture | Year-round | Consent-first, feeds everything above |
The tools that fit
Most practices live in one of two suites, which decides the everyday layer: Copilot for Microsoft firms, Gemini for Workspace ones, with the by-task table covering the rest. The practice-management and bookkeeping platforms you already pay for are sprouting AI features too; judge them with the AI-washing guide before paying again for what a licence already includes. The processes that cross systems, which is most of the list above, are where agents built to your shape beat any single product.
The rules, briefly
The January test is the honest scoping question for any accountancy automation: will this measurably change the firm’s worst month? Builds commissioned in autumn arrive proven, via a first month of approved-drafts-only running, exactly when the volume hits. Builds commissioned in January arrive as a lesson about autumn.
Client financial data is exactly what the approved-tools discipline exists for: business tiers, training excluded, DPAs, and one enforced rule about where client data may go. The templates section has the one-page policy and the DPIA starter; the professional-services governance guide covers the failure modes that reach regulators and insurers.
Where I fit in
I build exactly these systems: the chasing agents, the extraction pipelines, the reconciliations, under written specs with a human gate on everything client-facing. An Automation Audit maps your practice’s version of the list above in an afternoon; a Kick-starter ships the first process. The whole picture in practice terms, proof included, is at AI and automation for accountants. Bring your January pain to a call in October, and January will feel different.