An agency used to charge for the distance between an idea and a shipped asset. That distance had a price because it had a cost: a copywriter, an art director, a producer, a week of revisions. The client paid for the making, and the making was visible, so the invoice explained itself.

The distance has closed. A brand team that wants forty versions of a campaign line will have them before lunch, in nine sizes, in three tones. The work is competent. It is also close to free at the margin, and the retainer that priced the making has nothing left to attach to.

What survives is the part that never sat on the invoice: the decision about what to make.

The cheapest part was the easiest to price

Production is the most visible layer of agency work, which made it the easiest thing to sell and now makes it the easiest to automate. Writing, retouching, resizing, versioning, localizing. Each is a task with a clear input and a clear output, and tasks like that are what the current models do well.

When the cost of an asset falls toward the cost of the tools, procurement notices. A client who once accepted sixty assets a month as a line item starts asking why sixty, and why monthly, and whether anyone would miss ten of them. The count stood in for effort, and effort just got cheaper.

The squeeze lands hardest in the middle. A shop whose whole proposition is volume, that it can produce more content than the client's in-house team, is now selling against a subscription. Publicis told investors in 2024 that it would spend a nine-figure sum over three years on AI across the group, and the message was blunt: the largest networks are buying the tools and running them behind one platform. A fifteen-person content studio does not get those terms.

The invoice cited the making. The value was always in the choosing.

Where the premium actually sits

Clients still pay for judgment, and it shows up in four places.

  • Selection comes first. A model will hand a brand two hundred competent assets; choosing the twenty that matter is the skill that separates a campaign from noise. Volume is free now, which makes taste the scarce input.
  • Risk is second, and the least glamorous. A line that reads well and breaks a rule in one market is worse than no line at all. Someone has to know which claim clears which regulator, and that knowledge does not come out of a text box.
  • Context is third. Which partner is about to run out of money. Which competitor is bluffing. Which story the category is ready to hear this month and which it will punish next. This is what people who have watched a market for a decade sell.
  • Accountability is the last, and the hardest to automate. When a campaign misses, the client wants a person attached to the call, someone who owns the miss and can fix it. A model version does not do that.

The middle is the dangerous seat

The worst position is the middle: too big to be a specialist, too small to own a platform, selling hours and assets by the dozen. That shop competes on price against software that never sleeps and never asks for a raise.

Producing faster is the trap, because it doubles down on the one input that is losing value. The escape is to sell the decision, price it on its own line, and let production be a cost of goods the client can see through.

Omnicom and Interpublic agreed to merge in December 2024, which says less about creative ambition than about where the pressure sits. When the work in the middle commoditizes, the groups combine to protect the parts that still carry margin: media, data, and the senior counsel clients keep buying.

What to sell now

An agency that still bills by the asset is selling the part its client can buy elsewhere for less. An agency that bills for the call, and can defend the call, sells something the client cannot yet generate.

The practical version has three moves.

  • Price the thinking apart from the making, so the client can see which one they are buying.
  • Charge for the judgment that carries risk: the strategy session, the legal read, the selection, the name on the failure.
  • Produce at cost, or let the client produce, and stop pretending the asset was ever the point.

Marketing teams already understand this. The ones still paying agencies in 2026 are paying for a call they trust, made by someone who will still be around when the call is wrong.

Production keeps getting cheaper. The question every agency now faces is which half of its invoice it can still defend. The ones that answer with the call, and reprice around it before their clients ask, are the ones that survive the decade.