The moment clients learn AI is involved, the pricing conversation changes: “shouldn't this cost less now?” Answer it wrong and every efficiency you gain becomes a discount you gave away. How to charge clients for AI work is ultimately one decision — price the outcome, not the hours — plus the proof structure that makes the decision stick.

The discount spiral, and why hours cause it

Hourly billing turns every improvement in your process into a pay cut. AI cuts a six-hour job to forty minutes; on hours, you just cut your invoice 90% for making yourself better. The client didn't buy your hours — they bought a working site, a clean audit, a fixed problem. Bill the thing they bought.

What actually changed in your cost structure

Four models that survive the AI conversation

  1. Fixed-price outcomes. “Landing page: €X, delivered Tuesday.” Speed becomes margin, not discount. Anchor to the value of the page, not the effort.
  2. Productized retainers. Care plans with named deliverables — scans, updates, monthly report. The checklist runs on autopilot; you bill for the outcome and the accountability. The care-plan playbook maps tiers to automation.
  3. Value-priced projects. Migrations, rebuilds, SEO recoveries — priced against what the result is worth. AI moves your cost, not the client's value.
  4. White-label wholesale. Selling through other agencies? Flat per-site rates work because your marginal cost per site is near zero — the white-label model in full.

“But AI did the work” — the script

When a client pushes, don't defend the hours — reframe the offer: “You're not paying for typing time. You're paying for the result, the judgment about what to change, and the fact that every change on your site is reviewed, approved and logged by us. AI made us faster; it also made the work more accountable than any agency you've hired before.”

That last claim is literal: the audit log records every action and approval, and the month-end report is generated from it. Show it once and the conversation usually ends — the full argument is in is it safe to let AI edit a client's site?

What not to do

A worked example: the care-plan P&L

Numbers make the argument. Say your Standard care plan bills €150 per site per month. Manual delivery: two to three hours of checklist, updates and reporting — €100+ of senior cost, thin margin, quality hostage to busyness. On autopilot: jobs run the checklist, playbooks run the passes, the report generates itself, and the human cost is roughly twenty minutes of approvals — call it €25. Same invoice, margin transformed, and the deliverable got more consistent, not less. Multiply by thirty sites and the retainer book quietly became your best product line.

Notice what you did not do: cut the price. The client's alternative — hiring it done manually elsewhere — still costs what it always did. Your efficiency is not their discount.

Migrating existing clients without drama

  1. Grandfather nothing silently — at renewal, move hourly arrangements to the productized plan with named deliverables.
  2. Lead with the upgrade, not the mechanism: more checks, monthly reports, faster turnaround. AI is the how, not the headline.
  3. Show the first month's report side-by-side with what they used to get. The report usually sells the renewal itself.
  4. Keep genuine consulting hourly — judgment-heavy work never fit fixed pricing anyway.

What to charge more for now

The bottom line

AI didn't shrink the value of a working site — it shrank your cost of delivering one. Keep pricing outcomes, let the audit log carry accountability, and the retainer doesn't die; it gets better. The delivery machinery is here: maintenance on autopilot · white-label delivery · pricing.