If AI halved your delivery time and your prices didn't move, you gave the entire gain to your clients. The three ways to reprice speed — and the one thing you should never do with the saved hours.

The week's quiet question

All week we've built the system: the SOP, the voice-note plan, the consultant prompts, the agent rules, the competitor scan, the £0 stack. Here's the Sunday question the whole week has been walking towards:

If AI made you 30% faster, who's getting paid for that?

Because someone is. If the proposal that took a day now takes half a day and your price didn't move, the gain didn't vanish — it transferred, quietly and completely, to your clients. Nothing wrong with that if you chose it. Most founders never chose it. It just happened, one faster delivery at a time.

The decoded idea: efficiency isn't a pricing strategy — it's a pricing question. Speed gains have to be deliberately captured or they leak. And the standard reflex — "I'm faster now, I can take on more clients!" — is usually the worst of the three options, because it converts your efficiency into more work at the old margin, plus more admin, more context-switching and a fuller calendar. Volume is where speed gains go to die.

The three honest ways to capture the gain:

1. Reprice around value, not hours. If you bill by the hour, AI actively cuts your revenue as you improve — the absurdity that finally kills hourly billing. Move the offer to outcomes: the proposal, the month of management, the delivered thing. Your speed becomes margin instead of a discount you never announced.
2. Hold price, upgrade the deliverable. Same fee, visibly more: faster turnaround as a stated promise, an extra revision round, a monthly report that used to be unaffordable at your old speed. You bank the gain as retention and referrals instead of margin. Slower payoff, very durable.
3. Cut hours, not rates. Take the gain as time — the four-day week, the 3pm school run. Entirely legitimate, and the maths only stays honest if you actually protect the hours rather than backfilling them with unpaid busywork.

What doesn't work is the default: silently absorbing the speed while your pricing page stays frozen in the pre-AI economy. That's a strategy too — it's just one you never get to vote on.

The practical move. Thirty minutes this week: list what you sell, note honestly how long each took a year ago versus now, and mark where the difference went (price? deliverable? your time? nowhere?). Anything marked "nowhere" is the leak. Pick ONE offer and run option 1, 2 or 3 on it this month. One offer, one month, then read the result — price moves are tests, not vows.

Run the numbers with Wednesday's analyst prompt if you like — the system reviewing the system. And the usual line, meant sincerely rather than legally: this is a framework for thinking about your prices, not financial advice. Your market, your margins, your call.

The responsible-AI note. If AI is now materially involved in what clients receive, the trust question isn't only "should prices change" but "does the client know what they're paying for?" Our position is the boring, durable one: disclose where it matters, charge for the judgement and accountability that never got automated — because that's what they were always really buying. Speed came from the tools. The standard came from you. Price the standard.

The objection to answer before your market does. "But if AI does it faster, shouldn't it cost less?" Some clients will think it; a few will say it. The answer is the one professionals have always used when tools improved: the fee was never for the hours — it was for the outcome, the judgement and the accountability. The camera didn't make portraits worthless; it made the photographer's eye the product. Same here: what clients buy from a one-person business is that you stand behind the work — that someone vetted the tools, checked the facts and owns the result. That got scarcer in the AI era, not cheaper. If your positioning says "hours", this is the season to change what it says.

A note on sequencing, learned the careful way. Don't reprice everything at once — you lose the signal in the noise and you spook your own nerve. One offer, one change, one month, then read the result against a number you chose in advance (enquiry rate, close rate, margin — pick before, not after, or you'll grade your own homework kindly). If the test fails, you've bought precise knowledge about your market's ceiling for the cost of a few proposals. If it works, run the next offer. Pricing power, like the rest of this week's system, compounds one deliberate move at a time.

That's the week: one system, not more tools. One decoded AI move, every week — follow @aithebrief on Instagram for the daily version, or subscribe free on Substack for the deep one.