The Problem Is Structural, Not Accidental
Elena Verna's teardown of AI pricing published this week is worth reading in full, but the core finding is simple: you cannot compare AI vendors on cost-per-task because vendors have deliberately made that comparison impossible.
Grok prices its higher tier at "4x more than Plus" — but never discloses what Plus costs in real usage terms. Anthropic's Max plan promises "5x or 20x more usage" than Pro without ever anchoring what Pro delivers. Muse hides its pricing page entirely from logged-out visitors. These aren't oversights. They're the product.
Verna identifies four root causes: arcade-style credit currencies that obscure real spend, the psychological trick of large numbers feeling like large value, no agreed unit of consumption for general-purpose agents, and the fact that vendors themselves cannot reliably predict their per-interaction costs at scale.
What This Means When You're Negotiating
If you're evaluating AI vendors right now — for a sales tool, a support automation, a RevOps workflow — you're being asked to commit budget against numbers that don't have a common denominator. That's not a minor inconvenience. It's a material risk to your AI investment.
A few things you can actually do:
Demand a cost-per-task estimate in your currency, not theirs. Before any contract, run a defined test: a fixed number of representative tasks, with vendor support, and document actual cost. If a vendor won't participate in that exercise, that's your answer.
Strip out the credit layer. Ask explicitly: what does one credit translate to in compute or tokens? If they can't tell you, build in a true-up clause or a hard monthly spend cap.
Build in a 90-day exit. Pricing opacity is most dangerous when you're locked in. Until pricing in this space stabilizes, annual commitments without exit ramps are the wrong default.
The Window to Act Is Now
Verna's broader warning is worth taking seriously: this opacity works today because early adopters tolerate complexity. Mainstream buyers won't. The first vendor that publishes clear, comparable, predictable pricing will pull ahead structurally — and every operator who signs opaque contracts before that happens absorbs the transition cost.
If you're building AI into your own product or service pricing, the same logic applies in reverse. Predictable pricing is a feature. Your buyers are living through the same confusion you are on the vendor side. Being the vendor who makes costs legible is a real competitive position, not just a nice-to-have.