AI for professional-services firms

Most industries have to go looking for places to put AI to work. Professional services don't. If you run a law firm, an accountancy practice, a consultancy or an agency, your raw material already is the thing AI is good at: documents, knowledge, research, and repeatable judgement applied at scale. The work is read, draft, summarise, advise, repeat — and that is almost a description of what these tools do well.

The numbers bear it out. Among accountants, 81% say AI has improved their productivity and 86% that it has reduced the mental load of day-to-day work, per Intuit QuickBooks' 2025 survey. In legal, the Thomson Reuters and Wolters Kluwer research puts adoption among legal professionals around 79%, with routine tasks like review and research the clearest wins. The opportunity is real. The trick is capturing it without giving away your margin or your professional standing.

Where it pays first

The fastest returns come from the work that is high-volume, document-heavy and low-stakes if a human checks it before it leaves the building. In practice, that means:

• Drafting — first-pass letters, engagement terms, file notes, standard clauses and routine correspondence.
• Summarising — long contracts, bundles, sets of accounts or discovery documents reduced to the points that matter.
• Research and discovery — finding the relevant authority, precedent or treatment quickly, then verifying it.
• Proposals and reports — turning a brief and your past work into a structured first draft you then sharpen.
• Knowledge retrieval — answering "have we done this before, and what did we say?" across years of past matters and engagements.

Notice the pattern. The wins cluster around tasks where AI produces a draft or a shortlist that a qualified person then owns. One legal managing partner put it neatly in the Clio research: the best gains so far are in summarising documents, while pure drafting is more uneven. That distinction matters more than any headline adoption figure.

The billable-hour question

Here is the uncomfortable bit for anyone who sells time. If your team can now produce a competitive analysis in two days instead of two weeks, the value to the client is unchanged — but on an hourly model your revenue for that piece of work just fell. AI rewards efficiency, and the billable hour quietly punishes it.

That is why the larger firms are already moving. McKinsey has said roughly a quarter of its fees now come from outcome-based pricing, and SPI Research's 2025 benchmark recorded billable utilisation slipping to around 69% — below the level usually considered healthy. The direction of travel is toward fixed and value-based pricing for standardised work — contract reviews, compliance checks, routine filings — with the time saved redeployed into the higher-value advisory work clients will still happily pay a premium for. AI doesn't have to shrink your practice. But it does change what you should be charging for.

Where to be careful

The same qualities that make these tools useful make them dangerous if you trust them blindly. Three risks deserve real attention in a professional-services setting.

Accuracy. Generative tools can produce confident, fluent, and wrong output — including invented cases, mis-stated figures or plausible-but-incorrect treatments. In a profession where being wrong has consequences for clients, every factual or legal assertion has to be verified against a source, not taken on trust.

Confidentiality. Client information put into a public, consumer AI tool may be stored, processed or used in ways you can't control. That is a problem for privilege, for data-protection duties and for the trust your clients place in you.

Professional and regulatory duties. Your obligations to clients and to your regulator don't change because a machine helped. The work still has to meet the standard of conduct your profession expects — and the responsibility for it stays with you.

This article is general information, not legal, accountancy or compliance advice. Check your own professional and regulatory obligations before adopting any tool.

Keep a person accountable

The principle that makes all of this safe is simple: a human owns the output. AI drafts; a qualified person reviews, corrects, and signs off before anything reaches a client. The tool is a faster junior, not a replacement for judgement — and it is never the name on the advice.

Two rules carry most of the weight. First, nothing goes to a client without a competent person reviewing it. Second, client-confidential or privileged information never goes into a public, consumer-grade tool — only into systems with appropriate confidentiality, security and data-handling terms that you have actually checked. Get those two right and you have removed most of the downside while keeping nearly all of the upside.

A sensible starting point

You don't have to reinvent your practice to begin. Start with low-risk, internal-facing uses where a mistake costs minutes, not clients:

• Summarising long internal documents and meeting notes for your own team.
• First-draft internal memos, briefs and file notes that a person then edits.
• Drafting standard, non-confidential correspondence from a template.
• Searching your own past work to find relevant precedents and prior advice.
• Tidying, reformatting and structuring information you already hold.

Prove the value on work that never leaves the building, build the review habit while the stakes are low, then extend it outward. The firms that win with AI aren't the ones that adopt fastest — they're the ones that keep a person accountable while they do.

Sources: Intuit QuickBooks Accountant Technology Survey (2025); Thomson Reuters, "How AI is transforming the legal profession" (2025); Wolters Kluwer, legal AI adoption research (2025); Clio Legal Trends Report (2025); McKinsey outcome-based pricing disclosure, reported 2025; SPI Research Professional Services Maturity Benchmark (2025). Figures vary by sector and by how "AI use" is defined.