If you sell hours, AI is a problem before it is an opportunity.
Every hour AI removes is an hour you cannot bill. Get faster at a task you charge by the hour and you have cut your own revenue.
That tension is why professional services firms move slower on this than anyone else. Worth naming rather than tiptoeing around.
Where the work is landing first
Across consulting, accounting, legal, engineering, and advisory practices, the same jobs go first.
Research and preparation. Everything before the meeting. Background on the client, the market, the regulation, the precedent. Hours of reading compressed into a briefing you sanity check.
First drafts. Proposals, engagement letters, scopes, reports, file notes. The blank page disappears. The senior review stays.
Document review. Reading a large volume of material and surfacing what needs a human. Contracts, submissions, statements, applications.
Meeting capture. Recording, summary, actions, and a draft follow up before you have left the car park.
Knowledge retrieval. Finding what your firm already knows. Most practices have twenty years of excellent work sitting in folders nobody can search.
The pattern holds across all five. AI does the gathering and the drafting. Judgment, relationship, and accountability stay where they were.
The billing problem, honestly
Three responses. All of them are in use right now.
Do nothing and lose slowly. Keep billing hourly, keep working the old way. This holds until a competitor prices the same outcome lower because their cost base is smaller. It is a decision, even when nobody decides it.
Keep hourly, take the margin. Your team does the same work in less time. You bill the same rate for fewer hours and take on more clients with the same headcount. Revenue per person rises. Most firms land here first, and it works while it lasts.
Price the outcome. Charge for the result. Fixed fee, value based, or a retainer. Your efficiency becomes margin.
The third one is where this ends up. It is also the hardest, because it changes how you scope, how you sell, and how you pay people.
Move one service line to fixed price. Learn on something you understand well enough to scope accurately, then extend.
What Australian firms need to watch
Four local issues, and they matter.
Privacy. The Privacy Act applies to client data going into any tool. Where does that data go? Where is it stored? Is it used for training? Get answers in writing before client information touches anything.
Professional obligations. Legal, accounting, financial advice, and engineering all carry duties that stay in place when a tool writes the draft. You are accountable for the advice. A confident model is not a defense.
Client consent. Some clients will want to know. Some contracts will require it. Decide your position and put it in your engagement terms before a client asks you first.
Where data sits. Government and financial services clients will ask which country holds their information. Know the answer for every tool your team uses.
All four are cheaper to sort out early.
What separates the firms getting value
Everyone has access to the same models. The gap is fluency, and it shows up in three places.
Knowing what to ask for. People who get good output ask differently. They give context, examples, constraints, and a clear picture of what good looks like. Learnable, and most firms never teach it.
Knowing when to stop trusting. Experienced people spot a wrong answer instantly in their own field. That instinct needs retraining, because AI fails differently. It fails confidently, fluently, and in the middle of otherwise correct work.
Knowing what to keep human. The client conversation. The judgment call. The moment you tell someone the thing they did not want to hear. Firms that hand these over lose the reason clients pay them.
A sensible first ninety days
- Pick one service line. Something you do often enough to measure.
- Map where the hours go. Preparation, drafting, review, delivery, admin. Real numbers.
- Automate preparation and drafting. Leave review and delivery alone.
- Sort out privacy and consent. Written answers on data handling, and a position in your engagement terms.
- Train fluency. Sessions where people work on real matters with real supervision.
- Measure the same thing twice. Hours before, hours after, quality either side. Without this you have anecdotes.
- Then decide on pricing. With real numbers in hand, the fixed fee conversation stops being theoretical.
What clients still pay for
Judgment, accountability, and the relationship. Someone who has seen this before and will be there when it goes wrong.
AI removes the gathering and the typing that used to sit around all of that.
Firms that thrive will spend the recovered hours on the parts clients value. Firms that struggle will keep billing for typing.
Find your AI fluency level.
The fluency check takes two minutes and shows where you and your team sit today, plus what to work on next.
This is general information rather than legal or compliance advice. Check your own professional obligations and your firm's privacy position before putting client data into any tool.