AI for accounting firm partners: practical ways to work the margin problem

Accounting firm margins are under pressure in 2026 from AI-driven compliance commoditisation, salary inflation, and seasonal office overhead. The same AI tools that compress your fee can lift your margin if you apply them to internal work. Five use cases below.

Which service lines AI helps most

1. Client deliverable first drafts

Tool: Claude Projects with engagement context

Create a Project per engagement. Upload: engagement letter, prior year working papers summary, key issues list:

Draft the management letter for this audit engagement. Include: key findings, three recommendations with supporting rationale, management responses section (blank for client to complete), and a summary paragraph for the board pack.

2. Time and billing analysis

Tool: Claude or ChatGPT Code Interpreter

Export your WIP and billing data as CSV:

Analyse this billing data. Identify: the five clients with the highest gap between budgeted and actual hours, the three service lines with the worst margin, and the partner whose engagements most consistently overrun. Present as a table with one recommended action per finding.

3. Tax technical research

Tool: Perplexity or Claude with web search

Summarise the current HMRC position on [specific tax issue]. Include: the relevant legislation, any recent HMRC guidance or tribunal cases in the last 18 months, and the three practical implications for a client in [sector]. Flag any areas of uncertainty that require specialist review.

First-pass technical research in five minutes rather than 45. Always validate against the primary source before relying on it in client work.

4. Staff utilisation reporting

Tool: Claude with time recording exports

Analyse this staff utilisation data for the current quarter. Identify: which staff members are consistently below 70% billable utilisation, which client types are generating the most non-billable time, and what the utilisation impact would be if we moved two clients to fixed-fee arrangements.

5. AI-assisted seasonal capacity planning

Tool: Claude with historical billing data

Using this three-year billing history, model our expected capacity requirements for the coming busy season (January to April). Identify: the peak weeks by staff grade, which service lines create the most capacity pressure, and whether our current headcount plan will cover peak demand or require temporary resource.