AI Changed What Delivery Costs You. Your Rate Card Didn't Notice

By Craig S. Cody, CPA, Certified Tax Coach.

Here's the answer up front. The fastest way AI hurts a marketing or PR agency isn't by replacing it. It's by changing the cost of delivery while the agency keeps pricing the old way. The fix isn't a discount and it isn't a price increase. It's taking your fee apart into the four things it actually pays for, so you can see which one AI collapsed and which three it didn't touch. You can't defend a number you've never taken apart.

I work with more than 70 marketing and advertising agency owners every month. When one of them tells me AI has made their delivery faster, I ask the same thing back: did that contribute to a better bottom line for the agency, or did that margin evaporate?

Those are the only two places it can go. The speed is real either way. Which one you got isn't automatic and it isn't visible from your top line, which is why the question has to be asked out loud. The rest of this is how you answer it.

What the Keyword Data Says, and Why It Misleads

Free Agency Media published an analysis on August 29, 2026 with a finding worth sitting with. By their read of the keyword data, search demand for phrases like "AI creative agency" is still barely there, while agencies are already rebuilding roles, production, testing, and pipeline around AI-assisted work. They put the gap at roughly a year.

Their examples are concrete. Briefs that used to take a strategist an afternoon coming back in about a minute. Small teams producing sixty creative variants in the window a larger shop needed for six. One agency dropping a twenty-thousand-dollar-a-month creator retainer because the tools covered it.

Take those as their reporting, not as verified benchmarks. I haven't audited their keyword set or their examples, and you shouldn't repeat the numbers as if you had. What matters isn't the precision of any one figure. It's the direction, and the direction isn't in dispute.

Because here's the thing about that year-long gap. It's the wrong lag to worry about.

Why the Keyword Lag Isn't the Expensive Lag

A keyword lag is a marketing problem. It means buyers haven't settled on the words yet, so the search traffic for "AI creative agency" isn't there to capture. That resolves on its own. Language always catches up to behavior, and when it does you write the page and you rank for it.

There's a second lag underneath it, and nobody's writing about that one.

It's the gap between the day your delivery cost changed and the day your rate card changed. That gap isn't waiting on the market to find its words. It's open right now, on live engagements, and every month it stays open is a month of margin that quietly went somewhere other than your bottom line.

One lag costs you traffic you never had. The other costs you money you already earned. Only one of them is urgent.

The Four Choices You Actually Have

Say a process that used to take several humans several days now moves through AI-assisted research, production, and variation in a fraction of that. You have four options. Not three, and there's no fifth.

Keep billing the old hours until the client notices. This works until it doesn't, and the ending is bad. You're not being paid for the work anymore. You're being paid for the client's incomplete information, which is a wasting asset and a terrible thing to build a relationship on.

Cut the fee because the labor disappeared. This is the reflex, and it's the most expensive reflex in the business. I've said this before and I'll keep saying it: AI is not a discount strategy. Clients didn't ask for cheaper work because you got faster. Once you cut, you've reset the anchor permanently, and you'll never get that number back without a fight.

Give the freed capacity back as more testing and learning, at the same fee. This feels generous. It's the option most owners drift into, and I'll come back to why it's usually the worst of the four.

Redesign the service around the value that still requires human judgment. This is the real answer. It's also the only one that requires you to know what your fee is made of.

Three of those four are decisions you make by default, in the sense that not deciding picks one for you. The fourth takes actual work.

The Four-Layer Rate Card

Most agency fees are one undifferentiated number. That number was fine when everything inside it moved at roughly the same speed. AI broke that assumption, because it accelerated one component enormously and left the other three roughly where they were.

So take the fee apart. Four layers.

Layer One: Human Judgment

Strategy, taste, positioning, client context, and deciding what's worth making in the first place.

AI didn't touch this. It arguably made it scarcer, because when producing a hundred options costs nothing, choosing the right one is the entire job, and humans still do the choosing. This layer is where your pricing power lives and it's the layer most rate cards never name out loud.

If you can't articulate what your firm knows that a competent stranger with the same tools doesn't, you have a positioning problem that no pricing model fixes. That's the same test as deciding which work should never leave your building.

Layer Two: Machine Throughput

Research, drafting, versioning, production. The steps where AI genuinely multiplies what a small team can put out.

This is the layer that collapsed. It's also the only one that did.

Price it by output and capacity now, not by the hours it used to take. And price it knowing it's the layer with the least defensible margin over time, because your competitor is buying the same subscriptions. Throughput built on somebody else's platform is rented, not owned, which is its own kind of exposure.

Layer Three: QA and Risk Control

Brand review, factual verification, approvals, compliance, and deciding what should never reach the client or the customer.

Here's what almost nobody prices correctly. This layer got bigger, not smaller. Sixty variants need more review than six, not less. The volume that made layer two cheap made layer three more expensive, and if you netted those two against each other in your head and called it a wash, you netted wrong.

Somebody is absorbing that review load right now. It's humans doing it, at human cost. Find out who, and find out whether anyone's billing for it.

Layer Four: Commercial Outcome

What the added speed and testing actually improved. Conversion, revenue, cost of acquisition, cycle time, learning velocity.

This is the layer that justifies a fee going up rather than down, and it's the only one your client can independently verify. It's also the layer you can't invoice for unless you were measuring it before the AI showed up. If you weren't, start now, and accept that this layer prices at zero for a couple of quarters until you have something to show.

The Option That Feels Generous and Costs the Most

Back to option three, because it deserves its own section.

Handing the client ten times more testing for the same fee feels like the honorable move. You're not gouging them and you're not devaluing yourself. It reads as partnership.

Look at what it does to your numbers.

You removed labor cost from delivery, which helped. Then you added it right back, because more testing means more setup, more review, more reporting, more client meetings about results. Layer three grew. Layer two's savings got spent on layer three's expansion. Revenue didn't move at all.

That's the evaporation case, and it's the one that looks most like good service while it's happening.

Meanwhile you've reset the client's expectation of what your fee buys, permanently and without discussing it. Next year's scope conversation now starts from the bigger deliverable. You gave away the increase before you ever asked for it.

Testing more is often right. Doing it silently, at the old fee, with no conversation about scope, is how a good decision becomes an expensive one. That's a profitability problem you can't cut overhead to solve.

Where This Shows Up on Your Books, and Where It Doesn't

I'm a CPA, so let me be clear about which part of this I can help with and which part I can't.

A fee you set too low shows up as a thinner gross profit line, this month, before any of it reaches a tax return. It's a delivery-margin problem and it lives on the right denominator, not on your top line. Billings can hold perfectly steady while the money you keep quietly erodes underneath them.

And no deduction closes a pricing gap. There's no election, no entity change, and no strategy in my toolkit that recovers a fee you set wrong twelve months ago. Tax planning multiplies what you keep. It can't manufacture what you never charged.

That's not me talking myself out of work. It's the order of operations. Get the pricing right, then let's make sure you keep more of what it produces.

Who This Is, and Isn't, For

This is for owners of marketing and PR agencies who've already put AI into delivery and haven't touched the rate card since.

It's not for you if you're still evaluating tools. Nothing here is decidable yet, and you'd be pricing a change you haven't made.

It's also not for you if you bill purely on outcomes already, with no labor component anywhere in the model. You solved this before it was a problem, and the four layers are describing something you've been doing for years.

And I'll say the uncomfortable part plainly. If taking your fee apart reveals that layer one is thin, that most of what you sell is throughput anyone can now buy, the four-layer exercise didn't create that problem. It found it. Better this quarter than during a renewal.

The One Question for Your Next Pricing Meeting

Put this on the agenda, out loud, and make somebody answer it.

If AI cut our delivery hours in half tomorrow, what part of our fee would still be obviously worth paying?

Whatever survives that question is layer one and layer four. That's your actual business. Everything else is throughput and review, and both are now priced by a market that moves faster than your proposal template.

This is the same question I opened with, asked from the other end. The first one is diagnostic: where did the speed go? This one is prescriptive: what would you still be paid for if it all went away tomorrow? Answer both and you've done the work.

If the room goes quiet, that's information. Sit in it for a minute before you rush to fill it.

The mistake is thinking the agency's value was the hours that disappeared. If AI removes twenty hours of low-value production and gives the client far more useful testing, the answer isn't automatically a cheaper retainer. The answer is a better-designed service.

Frequently Asked Questions

Should we lower our prices because AI made us faster?

No, not as a default. Clients don't buy speed, they buy results, and cutting the fee resets your price anchor permanently in exchange for goodwill you likely already had. Change what the fee is attached to instead: price the judgment and the outcome, price throughput by capacity, and bill the review load that grew.

How do I know which layer AI actually changed?

Take one recent project and split the delivery time four ways: deciding what to make, making it, checking it, and proving it worked. Do that for the same project shape a year ago. In almost every case only the second bucket shrank, and the third one grew. That comparison takes an afternoon and it settles the argument with evidence rather than opinion.

Isn't giving clients more testing for the same fee good for retention?

It can be, if you name it as an increase in scope and value out loud. Done silently it costs you twice: your review and reporting load grows with no revenue attached, and the client's baseline expectation resets upward before you've had the scope conversation. Deliver the extra testing. Just don't deliver it invisibly.

What if my clients push back on paying for AI-assisted work?

That objection is almost always about the pricing unit, not the tool. If the fee is quoted as hours, AI-assisted work looks like fewer hours and the client does that math out loud. If the fee is quoted against a defined outcome and a defined scope of judgment and review, how the work got done stops being the negotiation.

Does this apply to a small agency, or only to larger shops?

It applies harder to a small shop. A ten-person agency has no bloat to absorb a margin slip, so a fee set for the old cost structure shows up on the bottom line within a quarter or two. Larger firms can hide it longer, which isn't the same as it costing them less.

Where does this fit against everything else AI is doing to agency finances?

Pricing is one question out of several, and it's not the first one. What the tools cost you comes before what you charge, and what the tools can reach and spend comes alongside both. This piece covers the rate card only.

The Bottom Line

The market will catch up on the vocabulary. That takes care of itself.

Your rate card won't. It changes when somebody sits down and changes it, and until then it's still quoting a cost structure you no longer have.

Take the fee apart into four layers. Find out which one AI collapsed. Price the other three like they're the business, because now they are.

If you'd like a second set of eyes on what your agency is actually keeping out of what it bills, let's talk.

Book a Free Tax Analysis

Craig S. Cody is a CPA, Certified Tax Coach, and retired NYPD Lieutenant. His firm works with more than 70 marketing and advertising agency owners every month, helping them keep more of what they make through proactive tax planning.

This article is general education, not advice for your specific situation. Confirm your own facts with your advisor before acting.