Answering the AI discount request
The 'you use AI now, so charge less' conversation, with three scripted responses that hold rate and one rule for when a discount is genuinely the right trade.
The fear runs ahead of the event. Agency owners spend real energy bracing for the client who says "you're using AI, so this should cost less," and the call lands less often than the bracing suggests. When it does land, the request is almost never really about the tool, which is why it is usually answerable without touching the rate.
It is about one of three things. Diagnosing which one you are looking at takes about two questions and determines everything that follows.
What they are actually asking
| What they said | What it usually means | What to answer |
|---|---|---|
| "I saw a demo, it looked like ten minutes of work" | They cannot see the work that is not generation | A value question — reframe the deliverable |
| "Our budget got cut for Q4" | The AI framing is a polite door into a budget conversation | A budget question — re-scope, do not re-rate |
| "If it's faster, we should be getting more" | They want output, not savings | A scope question — sell the volume |
The tell is what happens when you ask one clarifying question: "is this about the rate, or about what you're getting for it?" A value objection produces an argument about the tool. A budget objection produces a number. A scope objection produces a list of things they wish they had.
Answer the wrong one and you lose either way. Defend your rate to someone whose budget was cut and you sound tone-deaf. Offer more volume to someone who thinks you are overcharging and you have confirmed the work is cheap.
Response one: it was never priced on hours
For the value objection. The client has watched a demo of a tool producing a clip and mentally priced your deliverable at the length of that demo.
"That's a fair thing to raise, so let me be direct about what you're buying. Your rate isn't built on how long the render takes — it never was, including when we were booking crews. It's built on the strategy work, the concepting, the fact that we know which of your claims legal will bounce, the revision cycles, and the accountability when something underperforms. The generation step is maybe five percent of the hours on your account. If we halved it you wouldn't feel it, because it was never the thing you were paying for.
What has changed is what we can do inside the same fee. Last year you got three concepts and one execution. This quarter you're getting three concepts and nine executions, because testing is now the cheap part. That's where the efficiency went."
The mechanics under that answer matter. If you cannot say what share of your hours go to generation versus everything else, you are guessing, and a client with a spreadsheet will find the soft spot. Cost per creative, agency versus generated is the arithmetic worth doing on your own book first, and per-minute versus per-deliverable versus retainer pricing covers which contract shapes invite this conversation.
Response two: same rate, more surface area
For the scope objection, which is the most common of the three and the easiest to close well. The client is not trying to pay you less. They are trying to get more, and "you use AI now" is the lever they reached for.
"You're right that we can produce more per cycle than we could eighteen months ago. So rather than take the fee down, let's take the output up — at the same rate, next quarter we go from four hero assets to four heroes plus twenty-four variants, and we run them as a proper test rather than shipping our favourite. You get a read on which hook works instead of one asset and a hunch.
If that volume turns out to be more than you can absorb, we revisit the fee at renewal with actual data instead of a guess."
This converts a price conversation into a performance conversation, which is the ground you want to be standing on. It also commits you to something real: the variants have to run as a structured test rather than land as twenty-four files in a Drive folder. Testing video creative properly is the difference between "more assets" and "more learning," and only the second justifies the rate at renewal.
One caution: do not offer unlimited anything. Volume expands to fill whatever ceiling you name, and an uncapped variant promise becomes an uncapped revision promise within two cycles. Name the number, put it in the scope of work, and pair it with a revision policy that stops scope creep.
Response three: the pass-through split
For clients on contracts that itemise production spend separately from your fee — common in larger accounts, media-buying relationships and anywhere procurement is involved.
"Your contract already separates production spend from our fee. Production spend has come down, and you'll see that on the next statement — the platform line is materially lower than the shoot and edit line it replaced, and it's passed through at cost as it always has been.
The fee sits on top of that and hasn't moved, because the fee was never the production spend. It's the strategy, the creative direction and the ownership of the outcome. What I'd suggest is we hold both lines for a quarter, look at what the savings on the production line actually bought you in test volume, and have the renewal conversation with that in front of us."
This answer is only available if your contract is structured for it, which is the argument for structuring it that way in advance. It hands procurement a genuine, documented reduction to report — one that costs you nothing, because it was never your fee. It does require that you can attribute platform spend per client. Tracking credits per client deliverable is the unglamorous prerequisite; set it up before you need it, because reconstructing three months of usage across shared workspaces after the fact is genuinely painful.
When a discount is actually right
There are three, and only three, conditions under which conceding is the correct commercial call.
The scope genuinely shrinks. The client starts supplying finished briefs instead of a phone call, delivers approved assets in a usable state, consolidates feedback into one voice, or drops a deliverable. Your work went down, so your price can go down. Write the scope change and the rate change into the same amendment, on the same page, so the relationship between them is documented rather than remembered.
You are buying something. A twelve-month commitment instead of rolling monthly. A second business unit. A case study with named results. A move from net-60 to net-15. Discounts in exchange for term, volume or reference value are trades, not concessions, and they should read as trades in the paperwork.
You are deliberately repositioning the account. Sometimes a logo, a category entry or a relationship is worth more than the margin, for a defined period. Fine — but define the period, write the step-up into the contract from the start, and know what you are buying.
The rule underneath all three: never discount for the existence of a tool. Discount only for a change in what you deliver or a change in what you get. A rate cut granted because software improved is permanent, unearned and impossible to reverse, because the tool is not going to get worse next year. A rate cut granted for a twelve-month commit expires with the commit.
If you are already underwater on an account, that is a different problem with a different playbook — raising prices on existing content clients.
Say it before they do
The strongest position in this negotiation is not having it. Agencies that get asked are usually agencies whose clients found out sideways — a stray filename, a visible artefact, a competitor's pitch deck.
Three things prevent it. Put your use of generative tooling in the scope of work in plain language, at signature, so it is a documented method rather than a discovery. Name the efficiency yourself and say where it goes, in a sentence you repeat consistently: faster iteration, more variants, more testing, same fee. And show the work that is not generation — the concept round, the rejected directions, the claim review, the test readout. A client who sees twelve rejected concepts does not think the work took ten minutes.
What to tell clients about using AI covers the disclosure language, and how to price AI video services as an agency covers the pricing architecture that makes the position defensible. For the delivery side, how agencies scale client video output and the broader agency workflow.
FAQ
What if the client says they can just do it themselves?
Agree, sincerely, then be specific about what "it" is. They can generate a clip. Producing thirty on-brand assets a month, on a calendar, cleared for claims, sized per platform, tested against each other and reported on is an operating function, and standing it up costs a hire plus a ramp period. Some clients should bring it in-house, and the honest answer builds more trust than a defensive one. In-house versus agency versus AI production lays out the real comparison.
How do we handle procurement asking for an "AI efficiency" line item?
Give them one, and make it the pass-through production line rather than your fee. Procurement's job is to produce a documented reduction. A production-spend line that has genuinely fallen satisfies that, is honest, and leaves your fee intact. What you should refuse is a percentage cut applied to the fee itself with no corresponding change in scope, because that number never comes back.
Does any of this change if a competitor is charging less?
It moves the conversation from pricing to positioning. If a competitor genuinely does the same scope for less, you have a differentiation problem, not a discount problem, and the fix is on the delivery side. If they are doing a smaller scope for less, say so precisely and let the client compare like for like. Most "competitor is cheaper" claims dissolve the moment the two scopes are laid side by side.