What changes
Most firms have never looked at their services the way a new competitor would. Taking stock means doing exactly that. You list everything the firm sells, what each service earns, how many hours it takes and which parts of it an AI tool could already do to a reasonable standard. Then you score the firm against the six principles of the New Model, so you know where you’re starting from and which move to begin with.
This is a desk exercise that takes about two weeks. It needs the fee data you already hold, a few honest conversations with the people who do the work, and an afternoon testing current AI tools on real, anonymised tasks. The output is a one-page map of the firm and a short list of priorities.
Why it matters now
If AI keeps getting more capable each year while staying cheap to use, the parts of your work that depend on expertise being scarce will come under price pressure first. You can’t plan for that without knowing which parts they are. Firms that skip this step tend to buy tools before they know what problem they’re solving, or they reprice services that clients still value highly and leave alone the ones that are quietly losing margin.
A clear map also helps with your regulator and your clients. For advisers, FCA Consumer Duty asks firms to show that their fees represent fair value. Accountants and solicitors face the same question from clients, even where the rules are framed differently. Knowing what each service involves, and what it costs you to deliver, is the evidence base for every later move.
Checklist
- Export last year’s fee income by service. Use your practice management or back-office system, group it into no more than 15 service lines, and add the hours recorded against each one.
- Rate each service line for AI exposure. Mark which tasks AI can already do well, which it can do with checking, and which still need a professional’s judgement or signature. Test at least three real tasks on a current AI assistant using anonymised material.
- Score the firm on the six principles. Give each principle a score from 0 to 5, with one sentence of evidence behind each score. Ask two colleagues to score separately, then compare the results and agree a final figure.
- Ask ten clients what they’d pay for. Hold a 15-minute call with each about what they value most, what they’d happily do themselves with a good tool, and what they’d like more of from you.
- Write a one-page summary and pick your first move. Name an owner for it, set a start date this quarter and book a review meeting 90 days later.
Illustrative example
An accountancy practice with four fee earners earns £600,000 a year across 12 service lines. When it groups its income, year-end accounts, self-assessment returns and bookkeeping together make up £330,000, or 55% of fees. The team rates most of the work in these services as “AI can do this with checking”. Advisory work, including tax planning and business support, earns £90,000, or 15%, and is rated as needing professional judgement throughout.
On the six principles the practice scores 2 for AI capability, 0 for its own digital offering, 1 for AI legwork, 2 for coaching clients, 1 for pricing and 1 for being found by AI, a total of 7 out of 30. The summary is clear. More than half of its income sits in the work most exposed to AI, and the team isn’t yet equipped to change how that work is done. Its first move is to build AI capability, with a plan to hand over the bookkeeping legwork straight after.
Common mistakes
- Scoring on hope. Every score needs evidence. “We use AI a bit” is a 1, not a 3, until you can point to a process that runs on it every week.
- Testing AI on made-up tasks. Toy examples either flatter the tools or undersell them. Use real, anonymised work from the last month so the ratings reflect your actual caseload.
- Treating the map as a one-off. AI tools change quickly, so the answers will move. Date the map and repeat the exposure ratings every six months.