AI for Professional Services Firms: What Changes, What Doesn’t

AI for professional services is the use of AI inside firms that sell expertise: law, accounting and tax, audit, consulting, advisory and similar practices, where the product is judgment delivered under professional obligations. The tools are the same ones every company can buy. What differs is what a mistake costs, who answers for it, and how the work is priced, and those three differences decide almost everything about how a firm should adopt AI.

The profession has already adopted it, informally. Thomson Reuters’ Future of Professionals report 2026, a survey of 1,816 professionals across law, tax, audit, accounting, compliance and risk, found 74% use AI several times a week. It also found 34% use AI tools their organisation has not approved, in ways it cannot see, and 41% lack access to tools designed for professional work. Use is widespread. Control of it is not.

This is the index for that problem: what makes professional services a special case, what clients now expect, where AI genuinely pays in a firm, and the risks particular to the sector.

Why professional services is a special case

Four features separate a firm from an ordinary business using AI.

Confidentiality is a duty, not a preference. Client information in a law or accounting firm is protected by professional rules and, in several jurisdictions, criminal law. Where it goes when someone pastes it into a tool is a regulatory question before it is an IT question.

Accountability is personal. When AI-assisted work is wrong, the professional who signed it answers for it. In the same Thomson Reuters survey, 47% said final responsibility for an AI-assisted error lies with the individual professional, and respondents set clear conditions for tools they would trust: 96% require that client data is safeguarded, 94% that outputs are grounded in authoritative content, and 90% that the reasoning can be explained and defended.

Pricing is tied to time. Much of the profession still bills by the hour, so work AI makes faster can become less revenue rather than more margin. That tension is covered in depth in the repricing of consulting, and it applies across the sector.

Judgment is learned by doing. Firms have always trained juniors on the routine work AI now does quickly. Remove that work carelessly and you remove the apprenticeship. 48% of professionals in the Thomson Reuters survey said they were concerned about AI’s effect on developing independent judgment.

What clients now expect

The pressure is no longer only internal. Thomson Reuters found 78% of corporate clients say AI-enabled quality improvements from their professional firms are very important or essential, yet just 6% say most or all of their providers deliver them. Within twelve months, 32% expect to reconsider relationships with firms they see as falling behind.

On the legal side the gap is sharper still: 71% of in-house legal professionals expect their outside firms to change their commercial models as AI use grows, while only 28% of law firms say they have changed their pricing in response.

Clients are not asking firms to use more AI. They are asking for better, faster work, and assuming the firm has found a way to deliver it. The firm that cannot explain how it uses AI responsibly is becoming as exposed as the one that uses it carelessly.

Where AI pays in a firm

Across the sector the returns cluster in four layers.

• Research and first drafts. Finding relevant material, summarising long documents, producing a first version of routine correspondence, memos and reports. The largest time saving, and the layer where verification matters most.
• Review and checking. Comparing documents against standards, spotting inconsistencies, flagging clauses or figures that need a human look. AI raises a hand rather than deciding, which is the safest shape for it in regulated work.
• Client intake and communication. Structuring new enquiries, collecting documents, answering routine status questions and drafting updates for a professional to approve.
• Running the practice. Time capture, resourcing, billing and forecasting, the operational side that decides whether a firm makes money on the work it wins.

The Thomson Reuters data suggests strategy, not tooling, separates the firms getting value from the rest. Where a firm or department has a named AI strategy, 66% of professionals say AI is meeting or exceeding expectations. Where there is no strategy, 22% do.

The three risks particular to firms

Confidential data leaking into the wrong tools. The 34% using unapproved tools are rarely acting in bad faith. They are filling a gap the firm left. The fix is an approved option good enough to use, with contractual terms on training, retention and access that match the firm’s professional duties. The underlying questions are set out in data sovereignty, and the register of what is in use belongs to AI governance.

Unverified output reaching a client or a court. AI produces plausible text, and plausible is exactly what makes an error dangerous in professional work. Every output that leaves the firm needs a named professional who checked it against the source, not just read it.

A hollowed-out junior pipeline. If juniors stop doing the routine work and are not given something better to learn from, the firm will have plenty of capacity and too few people ready to exercise judgment in five years. The better design has juniors checking AI output against the underlying sources, which teaches the same judgment faster.

How this silo is organised

This briefing is the frame. Further briefings in this silo take it into specific professions and functions: AI for accountants and tax practices, AI for law firms and where it is safe to use, professional services automation and the operational side of a firm, and in time financial advisers and consultants. Each assumes the four features above and asks what they mean in that practice.

For firms, the adoption question is also a people question, which is why change management matters as much here as anywhere.

Where to start

Three steps, in order. First, find out what is already in use, including the unapproved tools, without punishing anyone for telling you. Second, provide an approved tool on professional terms and a one-page rule for client data, so the quiet workarounds have a better alternative. Third, pick one recurring piece of client work, redesign it with AI inside it and a named reviewer at the end, and measure the result in turnaround, quality and margin.

Then talk to your clients about it before they raise it with you. In Thomson Reuters’ 2026 AI in Professional Services report, about three-quarters of respondents said firms should take the initiative in that conversation, and 40% said they had been told both to use and not to use AI on client work, depending on the client. Clarity is scarce on both sides of the table.

If you want AI inside client work without losing control of it, tell Daniel which piece of recurring work you would start with, what is in use today, and what the firm needs it to do. Book 30 minutes with Daniel.

Frequently Asked Questions

How is AI used in professional services firms?

Mainly in four layers: research and first drafts, review and checking, client intake and communication, and running the practice through time capture, resourcing, billing and forecasting. Thomson Reuters’ 2026 survey found 74% of professionals in law, tax, audit and compliance use AI several times a week.

Why is AI adoption different for law and accounting firms?

Because of four features: confidentiality is a professional duty, accountability for errors is personal, pricing is often tied to time, and junior professionals learn judgment by doing routine work. Each changes how AI should be deployed, reviewed, priced and built into training.

Do clients expect their professional firms to use AI?

Increasingly, yes. Thomson Reuters found 78% of corporate clients say AI-enabled quality improvements from their firms are very important or essential, but only 6% say most or all providers deliver them, and 32% expect to reconsider relationships with firms falling behind within a year.

What are the main risks of AI in professional services?

Confidential client data entering tools without adequate protection, unverified AI output reaching clients or courts, and a weakened junior pipeline if routine work disappears without a better way to learn. Thomson Reuters found 34% of professionals use AI tools their organisation has not approved.

Does AI threaten the billable hour?

It puts pressure on it. When AI makes work faster, hourly billing turns time saved into lost revenue. Thomson Reuters found 71% of in-house legal teams expect outside firms to change commercial models as AI use grows, while 28% of law firms have changed their pricing so far.

Where should a professional services firm start with AI?

Find out what is already in use, including unapproved tools. Provide an approved tool on professional terms with a clear rule for client data. Then redesign one recurring piece of client work with a named reviewer, measure turnaround, quality and margin, and raise the subject with clients before they raise it with you.