AI strategy for professional services
How a law, accounting or consulting firm should approach AI: where it changes the work, and what it must never touch.
The short answer
A professional services firm sells judgement and trust. AI does not change that; it changes how the work around the judgement is produced. Research, first drafts, document review against checklists, summaries of long files, data extraction, internal knowledge search, proposals, admin and reporting all become faster, sometimes dramatically. The judgement, the advice, the sign-off and the responsibility remain with the professional, and the trust depends on clients knowing that. The strategy therefore has three layers in strict order. Confidentiality first: which tools may touch client material, under which terms, enforced. Internal productivity second: the work around the judgement, accelerated and always reviewed. Client-facing uses third, with disclosure and review. The largest risks are confidentiality breaches through consumer tools, fabricated citations or facts reaching professional output, and billing models that quietly assume hours AI has removed. Firms that get this right produce the same quality faster, price it deliberately, and can tell every client exactly how AI is and is not used on their matters.
The three layers
| Layer | What it covers | Non-negotiables | Measures |
|---|---|---|---|
| 1. Confidentiality | Approved tools; business terms with training excluded; processing agreements; staff rules; client statement | No client material in consumer tools; enforcement, not guidance | Zero incidents; audit of tool use; client questions answered |
| 2. Internal productivity | Research and source summaries; first drafts; document review against checklists; data extraction; knowledge search; proposals; admin; reporting | Professional reviews everything; citations and facts verified; output owned by a named person | Hours per matter type; turnaround; realisation; error rate |
| 3. Client-facing | Client portals with AI summaries; automated updates; intake assistants; document assembly for clients | Disclosure; review before anything reaches the client; a person reachable | Client satisfaction; response time; adoption |
Building the strategy
- Write the confidentiality rules and the approved-tools list; move everyone to business terms; enforce it.
- Tell clients in engagement terms how AI is used on their matters and how their material is protected.
- Map the work by matter type: which steps are research, drafting, review, extraction, admin, and which are judgement.
- Automate the admin and the knowledge search first; they are safe and they return hours.
- Introduce drafting and review assistance with mandatory verification steps built into the workflow.
- Measure hours and turnaround per matter type and decide pricing deliberately.
- Add client-facing uses only with disclosure, review and a reachable person.
- Train the professionals on what the tools do well, where they fabricate, and how to verify.
- Review quarterly: incidents, hours, quality, pricing, client feedback.
The pricing question
When a task that took four hours takes one, an hourly firm has lost three hours of revenue and a fixed-fee firm has gained three hours of margin. The strategy should decide which the firm wants to be for each kind of work: fixed or value-based fees for work AI has made fast, hourly for genuine judgement, and productised offerings where the process is now repeatable. Deciding this deliberately, and explaining it to clients, is better than discovering it in a falling realisation rate.
What this means for you
Approach AI in a professional firm in order: confidentiality rules and approved tools first, enforced and explained to clients; then productivity in the work around the judgement, with verification built into every workflow; then client-facing uses with disclosure and review. Decide pricing deliberately as hours fall, train the professionals on where the tools fabricate, and keep the judgement and the responsibility where clients expect them. The firm produces the same quality faster and keeps the trust that is its actual product.
Frequently asked questions
Can AI do professional work like legal research or financial analysis?
It can accelerate it: find and summarise sources, draft first versions, check documents against checklists, extract data from files, prepare analysis for review. It cannot be relied on for the professional judgement, it fabricates citations and facts with confidence, and the professional remains responsible for everything that goes out under their name. The right frame is a very fast junior whose work is always reviewed, never a substitute for the professional.
What about client confidentiality?
It comes first, before any productivity gain. Client material goes only to tools under business terms with training excluded, limited retention, known processing location and a processing agreement, and never to consumer tools. Firms should have an approved-tools list, staff rules that are enforced, and a clear statement for clients about how AI is used on their matters. A confidentiality breach through a convenient tool ends relationships and can end careers.
How does AI affect what we charge?
If the firm bills by the hour and AI removes hours, revenue falls unless pricing changes. Firms are moving towards fixed fees, value-based pricing and productised services for work that AI has made faster, keeping hourly billing for genuine judgement. The strategy should decide this deliberately rather than discovering it in the numbers, and should be honest with clients about what they are paying for.