AI for Professional Services Firms: Where It Pays First

AI pays first in a professional services firm on the repeated work around the expertise, not the expertise itself. That means proposals and scoping, meeting notes, first drafts of routine documents, finding past work and reporting. Keep advice, sign-off and the client relationship with people, and keep confidential client information out of public AI tools.

This is general information, not legal advice.

Where does AI pay first in a professional services firm?

It pays first on work that happens every week, follows the same pattern and can be checked by a qualified person in minutes. In most firms that is proposals and scoping, meeting and file notes, first drafts of routine documents, finding past work and preparing reports. None of it is the advice clients pay for.

The test is the same one used in any business: frequent, predictable and checkable. The guide to where AI pays first sets it out in full. What differs in a firm is the cost of a wrong answer, so every one of these produces a draft for a qualified person rather than a finished output.

What should stay with people?

Keep the advice itself, the professional judgement behind it, sign-off on anything the firm puts its name to, and the client relationship. AI can prepare research, structure an argument and draft the letter. A qualified person decides what the client is told and checks every source the draft relies on.

Verification matters because general AI tools still invent things. According to Stanford RegLab and HAI, state-of-the-art language models hallucinated on 69% to 88% of specific legal queries in their 2023 testing. Those were general-purpose models working without a firm's own sources. So a useful draft cites the firm's own material, and a person can check each reference quickly. The guide to AI hallucinations at work covers the fixes that hold.

Clients also hire a firm for the people they speak to. A partner's call after a bad result, or the judgement to turn down an engagement, is the relationship itself. Automating it saves little and risks a lot.

What do the professional rules say about AI?

Existing duties apply: confidentiality, competence, supervision and fair billing. Australian legal regulators have said so directly, and privacy law applies to any personal information put into an AI tool. Accounting, engineering and other professions have their own codes, so check your own body's guidance before setting firm policy.

According to the joint statement issued on 6 December 2024 by the Law Society of NSW, the Legal Practice Board of Western Australia and the Victorian Legal Services Board and Commissioner, "Lawyers cannot safely enter confidential, sensitive or privileged client information into public AI chatbots/copilots." The statement also suggests "Limiting the use of AI tools in their practice to tasks which are lower-risk and easier to verify." In addition, lawyers should make sure AI "does not unnecessarily increase costs for their client".

Privacy law reaches every profession. The OAIC recommends, as a matter of best practice, that organisations do not enter personal information into publicly available generative AI tools. In practice that means business AI plans, not personal accounts.

Anti-money laundering rules now reach many firms too. According to AUSTRAC, from 1 July 2026 these obligations apply to certain services typically provided by lawyers, conveyancers, accountants and trust and company service providers. AI can help assemble due diligence files and flag gaps. However, the compliance officer the program requires still makes the calls.

How does AI change the billable hour?

It shrinks the time some tasks take, which squeezes hourly billing on that work and strengthens the case for fixed fees. Firms that bill by the hour should decide how AI-assisted work is priced before clients ask, and never bill for time the work did not take.

The legal regulators' point about costs applies to every profession that bills time. Checking an AI draft is real work, but it should not cost the client more than doing the job without AI would have.

The practical answer is to measure. Time a proposal or a monthly report before AI and after it, then price from what you find. Firms that already quote fixed fees keep the saving as margin, so for them the case is simpler.

How should a firm start?

Start with one workflow the whole firm runs weekly, such as proposals or monthly client reports. Move staff onto business AI plans first. Then write the method down and run it for a month, with a partner reviewing every output. Measure the time saved before choosing the next.

1. Business plans first. Client data in personal AI accounts is a risk the firm cannot see.
2. One workflow. Pick the one with the most repeated hours.
3. Written method. Steps, checks and an example of good output.
4. Full review. A qualified person checks each draft for the first month.
5. Measurement. Hours before and after, and how much each draft needed changing.

Resist starting with the hardest, most valuable work. A first workflow that partners trust earns the time to tackle the next one.

Why does shared knowledge matter more in a firm?

Because a firm's product is its people's know-how, and most of it sits in senior heads, past files and inboxes. If each person's AI starts from nothing, every answer is generic. If every AI reads the firm's own precedents, methods and approved positions, juniors start closer to how the partners work.

Write down how the firm does its core jobs: how a scope is set, how a review is run, what a good file note contains. A method written as a skill can then be run by anyone's AI the same way. Keep client specifics separate, in each client's own matter or workspace, so one client's details never surface in another's work. For a firm, that separation is a professional duty, not just good housekeeping.

Skein builds a Company OS for professional services firms: one shared system underneath everyone's AI, holding what is true about the firm and how each job is done there, with every connection starting read only; the Company OS page shows the three layers. That suits a firm of 10 to 200 people whose staff already use AI on their own. A sole practitioner, or a firm that wants AI to advise clients without a person checking, is not a fit.

Frequently asked questions

How is AI used in professional services firms?

Most firms use it first for the work around their expertise: drafting proposals, summarising meetings, writing first drafts of routine letters, finding past work and assembling reports. A qualified person reviews each output. The advice itself, professional judgement and sign-off stay with people, because the firm carries responsibility for them.

Can lawyers use ChatGPT with client information?

Not in public AI tools. Australian legal regulators said in December 2024 that lawyers cannot safely enter confidential, sensitive or privileged client information into public AI chatbots. Commercial tools need their contract terms reviewed first, and lawyers remain responsible for confidentiality, verification and supervision whichever tool they use.

Will AI replace accountants or lawyers?

Not the judgement clients pay for. AI takes on much of the drafting, searching and summarising that surrounds professional work, so the same people can handle more. Advice, sign-off and accountability stay with qualified people, and their professional duties mean checking anything AI produces before it reaches a client.

Does AUSTRAC regulation apply to accounting and law firms?

From 1 July 2026, AUSTRAC's anti-money laundering obligations apply to certain services typically provided by lawyers, conveyancers, accountants and trust and company service providers. Firms providing those services must enrol, run an AML/CTF program and appoint a compliance officer. AI can help prepare due diligence files; people make the decisions.

Where should a small professional services firm start with AI?

Start by moving staff onto business AI plans so client data stays inside the firm's accounts. Then pick one weekly workflow, such as proposals or client reports, write down how it is done well, and have a partner review every output for a month before measuring the time saved.

Related reading