As AI takes on more execution, the useful question is quieter: what still belongs to the person who owns the outcome?
The answer turns out to be the part that was always the actual job. Deciding what's worth doing. Judging whether the work came out right. Standing behind it when a client, a partner, or an examiner asks who decided. Execution is being commoditized. Judgment and accountability are not.
The job that doesn't transfer
An agent may be able to draft a memo, prepare a reconciliation, collect research, or assemble a filing package, depending on the sources and review available. It cannot own the result. Ownership doesn't move to software. It stays with a person — and that person's job shifts from doing the work to managing the work.
That's not a smaller job. Anyone who has managed people knows it's often the harder one. You stop being measured by what you produced and start being measured by what your team produced under your review. The team is just made of agents now.
We call that person an AI agent manager. It is the role this firm is designing for.
Three rungs of one ladder
The progression has several rungs, even if you have not named them yet.
On the first rung, AI is a better search box. You ask it a question, you get an answer, you move on. Useful, but it's still you doing every step. On the second rung, you've made it do real work — drafts, analysis, a workflow you cobbled together. It works because you are still operating it. The setup lives in your head, your screenshots, and three chat threads. It does not survive you being out sick on a Tuesday.
The third rung is a work system: defined inputs, boundaries, review points, and a record. The manager job becomes real when the workflow no longer depends on one person remembering each step.
Why clear boundaries can accelerate delegation
AI work becomes hard to defend when its sources, review, and decision record are missing. "The AI did it" is not enough.
Each action class gets explicit permissions. Drafts may move differently from money movement, ledger posting, or client communication. Consequential steps stop at a named review point, and the record shows what happened. That lets you delegate more without hiding who decided.
Clear authority can be an accelerator because it makes the limits visible before the work moves.
What I'm not claiming
This is where I'd normally oversell. I won't. The tooling is early. The systems that do this well are still being built, ours included, and we say so on the page. Plenty of the work still needs a human in the loop who knows the domain cold. We do not promise a hands-off firm or a fixed date when that becomes possible.
Our thesis is that more professional-services work will be AI-managed. A deliberate transition names the owner, boundaries, review points, and record; an accidental one does not.
Where this goes
I'll keep adding to this as the ground shifts. It isn't a newsletter and it isn't on a schedule. It's the place I think out loud about the one transition I'm betting the firm on: the move from doing the work to managing the agents that do it.
If you're somewhere on that ladder and not sure which rung, start by naming one workflow and seeing what the Start Here guide gives you.