Forty hours of work became four. Most agencies that say this build the sentence as an achievement and then bill the same job to the client by the hour. The arithmetic is merciless: do the work ten times faster, sell it by the hour, and you earn a tenth.
The agency that sells hours gets poorer as it gets faster.
This should not be read as prophecy, because the mechanism is simple. An hourly rate was a fair measure in a world where labour was scarce. Adapting a campaign to sixty formats really did take days, a storyboard for a film really was a week, translation into twelve languages really was twelve separate people. The client paid for the hour because the hour was an honest indicator of the toil inside the work. Generative AI broke that indicator. The toil fell; the value did not. An agency that still measures both in the same unit is giving its own efficiency to the client as a gift.
Split agency work into three layers and what has changed becomes clear.
- Derivation. Multiplying an idea that already exists: format adaptations, language versions, size variations, twenty backgrounds for a product shot. This layer has gone to the machine. Anyone who says it has not either has not tried or is protecting an invoice.
- Exploration. Looking for an idea that does not exist yet: route experiments, tone research, animatics, first-draft copy. This layer has sped up with the machine. Clients used to be shown three routes because the budget did not stretch to a fourth. Now thirty can be tried and three shown.
- Decision. Saying which of the thirty is right for this brand, in this market, this year, and standing behind that call. This layer has not moved.
The trouble is that until now agencies earned most of their revenue from the first layer. Derivation work was invisible and dull and won nobody an award, but it was the backbone of the invoice. The creative director’s salary was often paid by the adaptation team’s hours. The machine took exactly that backbone.
The first reaction is usually to hide it. The agency uses AI, gets faster, and quotes the client the old timings. It works in the short run. In the long run it collapses for two reasons. The client has the same tools and can now see in its own team how long a job takes. And when hidden efficiency comes to light one day, the argument is no longer about price. It is about trust.
In Pricing Creativity, published in 2018, Blair Enns argued that creative work should be priced by value, not by the hour. When the book came out it read as well-meant advice: you would do well to. Today the same proposal has stopped being advice. For an agency that prices its input, the input is heading towards zero. Pricing the output is no longer one business-model option among several. It is the condition for staying on your feet.
Pricing the output is harder than it sounds, because it forces the agency into a conversation it is not used to. Selling hours, you tell the client how much you worked. Selling value, you have to know what this will earn them. The first needs a timesheet. The second needs you to understand the client’s business as well as the client does. This is the gap many agencies have to close before they get to AI: the thing they cannot price turns out to be the client’s business, more than their own.
The list of what an agency can do with generative AI today is long and grows every month. Produce the whole media set from one key visual. Build a watchable draft of a film before the shoot. Have a line written in twenty tones and pick the one closest to the brand’s voice. Generate thirty concept images overnight for a presentation. Try a voice-over without entering a studio. All of it is real, and all of it sits in the derivation and exploration layers.
The one thing not on the list is saying which of these should not be done. Throwing away twenty-nine of the thirty concepts. Telling a client “this is beautiful, but it isn’t yours”. In a world where the machine can produce without limit, the most expensive work has become taking away, and taking away has no hours.
In the period ahead we will see two kinds of agency. One will use the machine to cut cost, and its price will fall with it. The other will use the machine to try more, and will sell the client the accuracy of the choice, not the number of attempts. Both will use the same tools. The difference between them will be written on the first line of the invoice.