Outcome-priced AI: how we scope without timesheets

“Outcome-based” is easy to say and easy to fake. Here’s how we make it concrete.

Start with a sentence both sides can measure

A good outcome is not “stand up an AI Centre of Excellence.” It’s closer to:

By date D, capability C is live in production for cohort X, with policy gate P enforced, and adoption metric M above threshold T.

If you can’t write that sentence, you’re not ready to price the outcome — you’re still discovering.

Separate discovery from delivery

We sometimes run a short, fixed discovery to write the outcome. That’s fine. What we avoid is an open-ended “advisory retainer” that never graduates into a shippable gate.

Discovery ends when:

1. The production surface is named
2. Non-goals are written down
3. The outcome sentence is agreed

Then delivery begins — on BaseGrid or in your environment — with the commercial tied to that gate.

What belongs in the commercial

Include	Exclude
Outcome definition	Unbounded “innovation workshops”
Explicit non-goals	Guaranteed ROI claims without a metric
Acceptance / outcome gate	Body-shop rate cards as the primary frame
Support window after cutover	Surprise change orders for basics

Proof before pitch

We operate our own systems — Qwikyo, Pivra, MoltBillboard — because we refuse to sell a model we haven’t lived with. When we talk about outcome gates, we mean the same bar we apply to ourselves.

More on how we work: About · Case studies · Contact.