Clients have started asking agencies for an AI discount. The logic sounds simple from their side of the table: if your team uses AI, the work should cost less. From your side it is not simple. Your rate pays for judgment, accountability, and people who know the account. None of that got cheaper.
So you hold your rate. Most agencies do. Holding it is the easy half. The hard half is finding the hours somewhere other than your margin.
What the numbers say
In Productive's 2025 survey of around 180 agencies, 27% had already been asked by a client to lower fees because the agency uses AI, and nearly half expected the question soon. Only 13% of those asked cut their rates. The rest held the line. Productive's six-month pulse report (2026) found the question no more common.
The question is not spreading fast. It is not going away either. If you hold the rate and change nothing else, the next client who asks gets the same answer with nothing new behind it. The time has to come from somewhere.
Start where the hours are senior and repeatable
Look at proposal preparation. Three things make it the sensible first place to look.
- It is repeatable. Discovery notes, your template, service descriptions, and past examples, assembled again for every prospect.
- It is senior-heavy. The people who write proposals are usually the people who also sell and run accounts.
- It is built from material you already approve. Your services, your approach, and your best past work already exist. Someone rebuilds them by hand each time.
Published vendor benchmarks put the volume at 8–12 proposals a month and 4–6 hours each (Pitchsite, 2026). Used as assumptions, not as GradeCircle data, that is 32–72 staff hours a month. Your own figure may sit below or above that range. What matters is that it can be counted.
An illustrative example, not a client result: an agency of 25 that sends ten proposals a month, with a senior strategist spending five hours on each, puts 50 senior hours a month into assembly before anyone judges the content. Those are the hours behind the discount question.
"Just use ChatGPT" is not a process
Your team probably already uses a general AI tool to start drafts. That is a fair place to begin. It is not the same as having a workflow. Proposal software and general AI tools can both produce a draft. What neither comes with is your process: who feeds it, who checks it, and what counts as done.
Four things separate a habit from a process.
- Approved inputs. The draft is built from your discovery notes, your proposal template, your service descriptions, and approved examples. Not from an AI's idea of your offer.
- A named reviewer. Who checks accuracy, scope, pricing, and client commitments before a proposal leaves the agency is written down.
- A baseline. You know how long preparation takes today, so you can tell whether the new way is faster or only feels faster.
- A decision. At the end of a defined test, you expand, revise, or stop, based on what you measured.
Without those four, the time one person saves on a first draft can come back as someone else's corrections. Nobody sees the trade, because nobody measured the before.
What the draft does when it does not know
A proposal draft that invents a budget or a timeline is worse than no draft. In the workflow GradeCircle proposes, a missing budget or timeline is flagged for the owner to resolve, not filled in. Your review checkpoint confirms it was flagged, before anything goes out.
Your team stays in control. The workflow prepares a first draft from approved inputs only. Your team reviews scope, fees, claims, and commitments before anything goes to a client. The agency's judgment stays where it is. What shrinks is the repeated assembly.
For how that runs with a real team over a fixed period, see Inside a Four-Week AI Workflow Pilot.
Test one workflow on your own numbers
The AI Opportunity Review is where this starts. It costs $1,500 and takes 10 business days. We review representative work samples and up to three workflows, interview up to three people, and assess benefit, effort, and data suitability. You get a business case, human review requirements, and a pilot brief with a fixed implementation quote.
The business case sets setup and ongoing costs against released capacity, in conservative-to-optimistic scenarios. Capacity is reported as hours, not as payroll savings. Released capacity does not automatically reduce payroll costs, and the review does not pretend it does. The benchmark assumptions above are only a starting point. The review replaces them with your own evidence: your volume, your hours per proposal, your people. What your senior people do with any hours released is your call.
The review ends in a decision meeting. You leave with a recommended first pilot at a fixed price, or a recommendation to defer AI and what would need to change. Deferring AI may be the recommendation. If the review finds your team can get there on its own, it says so. Either way the findings are yours.
You can buy the review on its own. The AI Workflow Pilot, $5,000 over four weeks, is a separate decision you make after the findings. Scoping costs nothing and commits you to nothing. Paid work starts only after you agree the scope and fixed fee.
Holding your rate is the right first answer. The Review tells you, on your own proposal numbers, where the hours behind that answer can come from.
GradeCircle builds and runs six products of its own: AfroSynergy, CallerBridge, HealthSquire, Heyzl, Lexfora, and LukeAPI, two of them in beta. The assumptions in our models come from operating those systems, not from a framework. See what we run.
If your agency writes proposals every month, scope your proposal workflow: five questions, and you see a preliminary fit summary before sharing contact details.



