The quiet question
Owners keep quietly asking me if they can use AI to cut staff.
It is almost never asked in the meeting. It comes at the end, when the laptop is shut and we are walking to the door.
The instinct is understandable. Margins are thin, capacity is tight, AI looks like a cheaper way to do the same work, and wages are the biggest number on most service-firm P and Ls.
But here is the problem. You can only cut so much expense before you hit bone.
Why cutting has a floor
Past a certain point, every cut takes capability with it. The team gets smaller and weaker at the same time. The owner picks up the work no one else can do. The bottleneck gets worse, not better.
There is also a ceiling on it. Cost cutting is finite by definition. You can only get to zero once, and nobody has ever grown a firm by arriving there.
Headcount is rarely the real constraint
Most service firms are held back by demand or delivery quality, not by labour cost. Cutting people solves the wrong problem.
Run your own firm through it honestly. If you are turning work away because you cannot deliver it, cutting heads makes that worse. If you are delivering the work but the quality is uneven, cutting heads makes the unevenness worse. If your real constraint is that you are the bottleneck, cutting heads makes the bottleneck deeper, because the work lands back on the one person who was already the choke point.
In three years of these conversations I have not once found a firm whose genuine constraint was that it employed too many good people.
The move
The right move is to lift your people up to the work only they can do, and let AI take the rest.
Take the low-value work your best people are doing. Inbox triage, basic client comms, repetitive data assembly. Move it to AI, with governance. Free them into the work only they can do, the work the client pays a premium for, the work that compounds the relationship.
Output goes up. Margin goes up. The team gets more interesting work, not less of it.
We ran this on ourselves before we recommended it. Halving our own tickets and our own time to resolve did not mean fewer people. It meant the same people moved from firefighting to prevention, from closing tickets to improving the systems that were creating them. That is where the durable value came from, and it is not something a competitor can buy off a shelf.
Someone still owns the decision
AI can recommend all day. It does not carry the liability or the client relationship.
When the call is hard, when the client’s position is delicate, when the data is incomplete, the human owner or senior adviser still owns it. AI supports the decision. It does not replace the decision-maker.
Human value moves up the stack. Judgement, relationships, context, the calls that need someone who understands the client. Point your AI at everything below that line and your people at everything above it.
The prospect assessment that changed the team’s mind
A prospect came to us worried about where AI could help. Their people were worried AI would take all the work.
We mapped their week. Most of their best people were buried in low-value tasks. Meanwhile, there was higher-value work they were not doing. Strategic reviews with their biggest clients, advisory conversations, complex problem-solving.
The team did not need to be cut. They needed to be freed into the work that was waiting for them. AI took the low-value tasks. The team moved up. Nobody lost their job. The owner got back the weeks they had been spending as the catch-all on the busywork their best people had been drowning in.
The part I did not expect was the mood in the room. The fear in that team was not really about redundancy. It was about spending another year on work they knew was beneath them.
The question to ask about every role
Here is the one to take away, and it works on every seat in the firm including yours.
What is this person doing that a governed AI workflow could do, so that this person can do the thing only they can do?
Ask it honestly, role by role, and write the answers down. That question is the whole game. It is also, notably, a question about your people rather than about software.
The standard for this series
Every week is something we have run inside our own firm or inside a real New Zealand client firm. No theory I have not run myself. No invented case studies.
What to do next
If you want to map where AI would lift your team rather than shrink it, book a 15 minute call with me. If it is not a fit, I will say so. The AI Readiness Audit is the lighter first step.
Week 5 of The AI Divide lands next week. The margin moat, and why cutting cost is the shallow win.
