Talk to an Expert

Staged change

The Observatory, and how to use it each week

5 min on the light depth

The frame

The Observatory is the log of verified pricing moves in AI software. A move is a pricing change a vendor made, evidenced as a transition: the dated event plus what the vendor’s terms said before it. A change on a pricing page is not yet a move, because a page that reads differently this month may be a repricing, a rewrite or a correction, and only the before-state says which. Under each move a pricing architect writes the read. It names the mechanic under the vendor’s label, the unit, the grant or the package that changed, and states what changed and what did not. The vendor’s motive is left unstated, so the conclusion stays yours.

Every move is filed under the decision that moved, licensing, packaging or pricing, the three decisions a price is built from. Most are announced in pricing language, so the filing is the first read you take from an entry. The weekly email, the Observatory Weekly Register, and the Slack digest list the moves the week they are verified, and both are silent in a week with nothing verified. What the log cannot see is what buyers signed; that half is held apart as pricing intel.

Each week a vendor you buy from or compete with changes a price, and someone asks you what it means. The log gives you two jobs. Suppliers, for cost: if your product runs on a supplier’s model, their unit is your cost, and a unit passed through to your customers was chosen upstream. Competitors, for condition: take the four-part read to each move; the log supplies their shape and their condition, your won and lost deals supply yours, and the gap is yours to name. How to read the AI pricing observatory argues that a move is a transition, that the read names the mechanic and leaves motive out, and that the weekly habit is two jobs.

Which verified move this week changed a supplier's unit your cost runs on, and which showed a competitor's condition you can check in your own deals?

Three things to say in the room this week

  • A change on a pricing page is not yet a move; I want the dated event and the terms before it.
  • Most moves are announced in pricing language, so which decision the log filed it under is our first read.
  • Our supplier's unit is our cost. When their rate card moves, which part of our architecture responds?

Core reading

Go deeper: the heavy depth

Read the core first.

  • How to read the AI pricing observatory. What a verified move is, who writes the read and what it leaves out, the decision pages and vendor arcs, and the two weekly jobs.

The shelf, read with the four-part frame.

  • Agentforce pricing. Several constructs on one product: which were units for the meter, which were wrappers against it.
  • The DeepSeek price increase. A flat token rate card becoming a time-of-day grid, so the cost model acquires a clock.
  • The Windsurf pricing change. A retreat from credits to usage quotas, with a price increase in the same announcement.
  • The GitHub Copilot pricing change. A seat plan moving to token-denominated credits, one end of a spectrum of cost-curve risk.
  • When the meter catches the spike. The revenue spike after that move, and why a meter the customer controls is managed down.
  • OpenAI’s plan sharing and marketplace. A subscriber’s included usage spendable inside partner products, and what that path does to a vendor’s budget.
  • OpenAI vs Anthropic pricing. Two rate cards read as cost per completed task, and why the posted token rate rarely sets margin.
  • Outcome-based vs consumption-based AI pricing. Two vendors reported as one retreat, one metering attempts and the other metering successes.

Where this leads. The Observatory page is a running read on vendor pricing moves in AI software, grouped by vendor and by decision. Its pricing page holds the moves where the rulebook moved: what the units cost, on what rate card, under what rules. The alerts page puts the weekly summary in a Slack channel you choose, silent in a week with no verified move. Pricing intel holds what the Observatory cannot see, what buyers signed, apart from the log and returned as the aggregate.

Next module: Continuous monetization as the operating model.

The shelf

Where this leads

Which depth did you take?