GTM is trying to sell AI like SaaS. It isn’t.
If it looks like a duck,
acts like a duck,
and quacks like a duck…
it’s probably a duck.
AI looks like SaaS.
Is sold like SaaS.
But it doesn’t behave like SaaS at all.
In SaaS:
- cost is largely fixed
- margins are modeled upfront
- pricing holds as you scale
In AI:
- cost scales with usage
- usage expands during execution
- and much of that expansion isn’t even visible
So GTM is being asked to:
- price something that doesn’t have a stable cost
- sell outcomes that don’t have consistent delivery
- defend margins that can move underneath them
Meanwhile, engineering is measuring tokens —
but tokens don’t capture the real work:
- agents retry
- branch
- call tools
- coordinate across systems
→ tokens ≠ total compute
→ cost shows up after execution
So by the time anyone sees the number, it’s already been spent.
That’s the tension:
- GTM is accountable for revenue
- Finance is accountable for margin
- But neither controls how the cost behaves
This isn’t a messaging problem.
It’s not a positioning problem.
It’s a model problem.
AI behaves like a utility.
But it’s being sold like SaaS.
Until that changes:
GTM will keep getting pulled into conversations they can’t confidently close.
– Published on Tuesday, March 31, 2026