A fixed panel of pricing, RevOps and finance practitioners answers the same ten questions every month. The answers become diffusion indices, built the way the PMI is built, and every reading carries its n, its range and its composition. First reading 12 December 2026.
THE INDEX READS PAGES · THE PULSE ASKS PEOPLE
The Agent Pricing Index reads what vendors publish, one page at a time, with a verbatim quote and a read date. The Pulse asks the people who set and pay the prices what moved this month. Same panel discipline as the ISM's PMI: a fixed panel, the same direction-of-change questions every month, confidential answers, published aggregates.
Who answers. Two respondent types, kept separate in every cut. Sellers: pricing, RevOps, finance or founders at companies that sell AI software. Buyers: the people who approve AI software spend. A respondent answers for one company and names its category from the Index's list. Only aggregates publish; nothing publishes below the floor.
The one number. The AI Monetization Pulse runs 0 to 100, where 50 means no change from last month. Above 50, sellers are getting more for their units; below 50, less. It is the equal-weight average of five seller indices: list prices, realized price per unit, discount pressure (inverted), the shift of the billable unit toward usage and outcomes, and the 90-day price expectation. Buyers answer a parallel set about spend, published as the Buyer Pulse.
AS ASKED · THE WORDING NEVER CHANGES SILENTLY
| # | Question, as asked | Answer | Feeds |
|---|---|---|---|
| S1 | Which best describes you this month: we sell AI software; we buy it; both | one of three | Screener: seller or buyer |
| S2 | Your category (sellers) or the category you spend most on (buyers) | the Index's category list | Segment |
| S3 | Typical deal size | the Index's five bands | Segment |
| 1 | Compared with last month, your list prices are | higher, same, lower | Sellers: list-price index |
| 2 | Compared with last month, the price you actually realize per unit, net of discounts, is | higher, same, lower | Sellers: realized-price index |
| 3 | Compared with last month, pressure to discount is | more, same, less | Sellers: discount index, inverted |
| 4 | In the last 30 days, did your billable unit move | toward seats, no change, toward usage or outcomes | Sellers: unit-shift index |
| 5 | In the next 90 days you expect to | raise, hold, cut | Sellers: expectation index |
| 6 | Compared with last month, your gross margin on AI delivery is | higher, same, lower | Sellers: margin index, published beside the composite and not in it |
| 7 | Compared with this quarter, your AI software budget next quarter is | up, same, down | Buyers: budget index |
| 8 | Compared with last month, the price per unit you are being quoted is | higher, same, lower | Buyers: quoted-price index |
| 9 | In the last 30 days, did a vendor change the thing you are billed for | yes toward seats, no, yes toward usage or outcomes | Buyers: unit-shift index |
| 10 | The price of one unit you sell or buy, and what the unit is (optional) | a number and a word | A row for the Index |
| F | What the price does not show (optional, one line) | text | Anonymized quotes in the release |
Sellers answer S1 to S3, 1 to 6, 10 and F. Buyers answer S1 to S3, 7 to 10 and F. "Both" answers all. The screeners never change; a question's wording never changes without a note in the release and a break marked in the series.
THE ISM'S ARITHMETIC, STATED IN FULL, SO EVERY NUMBER CAN BE RECOMPUTED
A diffusion index for one question is the share answering the up direction plus half the share answering no change, times 100. Fifty means no change from last month. "Up" is always the direction in which sellers get more for their units: for discount pressure, less pressure is up; for the billable unit, toward usage and outcomes is up.
The composite is the equal-weight average of the five seller indices, computed over respondents who answered all five. The gross-margin index publishes beside it and is not in it, because margin is a cost signal rather than a price signal. The three buyer questions form the Buyer Pulse, published as three indices with no composite until the buyer panel passes 100.
Intervals. Every respondent contributes 1, 0.5 or 0 to an index, so an index is a mean with a standard error, and the release prints the value with its 90% interval. The composite's 90% range comes from a respondent-level bootstrap, which keeps the correlation between questions. The release prints the up, same and down shares for every index, so nothing hides inside the number.
The floor. The composite publishes at 50 complete seller responses or more; any index or segment cut publishes at 20 or more. Below the floor the release prints "not published" with the n. No reading is padded, back-filled or imputed.
Weighting and adjustment. None in year one. Composition by category and deal band publishes with every reading. When the seller panel passes 300, readings are post-stratified to the Index's category mix and the series carries a marked break. No seasonal adjustment until 24 readings exist.
Non-response. The response rate publishes every month. A panelist who misses three readings in a row is retired and told so. The roster count that publishes is active panelists, never invitations.
Labels. A published index is OBSERVED: a dated count of dated answers. A reading is never MODELED. Anything derived from a reading carries ANALYSIS and shows its arithmetic.
PUBLISHED HERE ON THE FIFTEENTH · THE CHART GROWS A POINT A MONTH
Reading one fields from 24 November to 5 December 2026 and publishes here on 12 December, with the n and the composition. Whatever clears the floor publishes; whatever does not is named. From reading two, the form opens on the first business day of the month and closes on the tenth, with one reminder on day six, and the reading publishes on the fifteenth.
Each reading publishes as this page, as JSON beside the Index aggregates on the API, as a release note to the list, and on the Company Page. Segment cuts go to Pricing Tracker subscribers inside the Tracker.
VISIBILITY AND ACCESS · NEVER THE QUESTIONS
| Product | Release sponsorship | Investor subscription | Custom question |
|---|---|---|---|
| Price | $2,500 per reading, billed quarterly | $25,000 a year | $5,000 per question per reading |
| Opens | Reading three, February 2027. Three seats. | Reading six, May 2027, with six months of history. | With sponsorship. One per reading at most. |
| What it buys | Logo on the release and this page, a 30-minute briefing before publication, the segment cuts. | The anonymized response-level file each month, 48 hours before publication, with the history. A quarterly call. | One question added to the month's form, wording approved by the operator, results published to everyone. |
| Never | The questions. The respondent file. The release date. | Respondent identities. Early release to anyone else. | A question that names a sponsor's product. |
Sponsorship buys visibility and access, never influence. The questions, the floors and the release date are the operator's, in writing on this page, and a sponsor who asks for the questions is shown this paragraph. Tracker subscribers get the segment cuts and the early read inside the Tracker from $299 a month.
Panelists are paid in the early read, never cash: the reading two days before it publishes, your own category and deal-band cut, and, for sellers, the Index benchmark unlocked. You sit as a seller or a buyer for one company. Verification is a LinkedIn profile plus a manual role check, once. Your answers are confidential and only aggregates publish.
Leave your work email and you get the questionnaire when reading one opens on 24 November.
The same panel discipline runs the Panel Verdict: practitioners who estimate alone, see each other, and revise, weighted by their record on questions with known answers.