Most agencies price a new AI service too low, and they do it for a reason that feels like prudence. They look at what the platform costs them, add a comfortable markup, and call it a price. That is not pricing. That is marking up a cost and hoping.
The fear underneath is understandable. It is a new offer, you have no case studies yet, and a low number feels safe. But a low number is not safe. It anchors the client to the wrong figure, teaches them to see the service as cheap, and hands back the margin that funds your delivery.
The anchoring mistake
The error is choosing the wrong anchor. Your platform cost is a floor. It tells you what you cannot go below. It tells you nothing about what the service is worth to the client, and the client’s number is almost always the larger one.
Anchor to cost and you build a price from the bottom, nervously. Anchor to client value and you set a price from the top, then check that it clears the floor. Same inputs, opposite direction, and the second direction is where the margin lives.
The client never sees your cost, which is what makes cost-anchoring a self-inflicted wound. They weigh your price against what the service is worth to them, not against what you pay to run it. Price against their value and the number explains itself. Price against your cost and you have quietly capped your own margin to protect a figure the client was never going to see.
A model you can run
Here is a retail price built from three inputs. Every figure below is an example so the arithmetic is concrete. Replace each one with your client’s real numbers and your own target.
Illustrative inputs: swap in your own
Cost sets the floor. Your margin target lifts it to a retail number you would be comfortable defending:
retail floor
= delivery cost × margin target
= $300 × 3
= $900 / month (example)
Now check that price against value, which is the step the fearful pricer skips. At $900, look at what the client still keeps every month:
what the client keeps
= client value − your retail
= $2,000 − $900
= $1,100 / month (example)
On these example numbers, $900 is not aggressive. It is timid. The client keeps more than they pay you, which makes it an easy yes, and you have left room you did not need to leave. The nervous version of this agency prices at cost plus a small markup, lands near $400, and never once looks at the $2,000. That is the difference a good anchor makes.
“Cost tells you what you cannot charge less than. Value tells you what you can charge. Price from the second number and defend it with the first.”
Why the structure works in your favor
The way partner pricing is built makes this yours to capture. You pay a flat platform fee plus usage at a wholesale discount. You set the retail number your client pays. The gap between what they pay you and what delivering the service costs you is your margin, and it is yours to keep, every month the service runs.
That is why the recurring part matters more than the first invoice. You are not selling a project. You are pricing a line of recurring revenue, and a number set out of fear compounds against you for as long as the client stays. Set it from value, hold it, and let the margin do the work it is there to do.