Reading a Deployment Across Eight Seats

• Source: Report framework library

Encodes a Revenue AI Report framework so the agent applies the published method instead of improvising one.

1. 1.Copy the SKILL.md text below, or download the raw file.
2. 2.Create a folder named exactly reading-a-deployment-across-eight-seats in your agent's skills directory.
3. 3.Save the file inside that folder as SKILL.md.
4. 4.Ask the agent one of the trigger requests below.
5. 5.Check the output against what you already know before it leaves your desk.

• Run the eight-seat read on this AI SDR platform before we sign.
• Who owns this deployment across the revenue org, and which seat is exposed?
• Build the board scorecard for AI across sales, marketing, RevOps, CS, and finance.
• Set kill criteria per function before we launch this agent.

• Score this one initiative out of five and tell me whether to renew.
• Our AI reporting looks great and revenue is flat. Diagnose the pattern.
• Write a job description for a RevOps analyst.

---
name: reading-a-deployment-across-eight-seats
description: Read one AI deployment from eight functions so the conclusion cannot be disproved by a seat that was never asked, and report it on the thirty-two cell Eight-Seat Read scorecard.
---

# Reading A Deployment Across Eight Seats

Read one AI deployment from eight functions so the conclusion cannot be disproved by a seat that was never asked, and report it on the thirty-two cell Eight-Seat Read scorecard.

## When to use this skill

- A tool is about to be bought and only the sponsoring seat has been consulted.
- A live deployment is up for renewal and the case rests on one function's numbers.
- A board needs one defensible operating view of AI across the revenue organization.
- A win is being published internally and the exposed seats have not been named.
- Kill criteria are needed before launch and nobody has said who can trigger them.

## Inputs to collect

- One named deployment, not a category. Source: the program tracker.
- Per seat, the metric that seat owns today with its current value. Source: CRM for sales and customer success, marketing automation for marketing, finance system for revenue finance, enablement records for ramp.
- The stack diagram, the data model, and the workflow blueprint for the deployment. Source: RevOps, before the tool is bought.
- Access control and data boundary: what the agent can read, what it can write, who sees the output, and what happens to the transcript. Source: security and RevOps.
- Contract terms including usage-based inference charges and renewal date. Source: vendor invoice and procurement.
- The attribution rule that separates partner lift from direct revenue. Source: partnerships and RevOps.
- The named owner for each seat, with title. Source: the org chart.
- A quarter or more of production data before reversal and quality metrics are read. Source: the tool's logs and the systems owner.

## Process

1. **Fix the subject.** Take one proposed or live AI deployment, not a category (https://www.therevenueaireport.com/frameworks/eight-seats).
2. **Fill each seat.** For each of the eight seats, write what it owns, the one decision it has to make, and the failure mode it is exposed to. Use the published definitions below and change nothing about their order.
   - **Sales.** Owns pipeline created, conversion, and quota attainment. Decision: know what the AI SDR actually did before you sign the contract. Failure mode: activity volume counted as pipeline, with meetings booked reported without accepted, qualified, or closed rates behind it.
   - **Marketing.** Owns demand, attribution, and the brand surface an AI buyer reads first. Decision: see how attribution and demand hold up against an agentic buyer. Failure mode: output volume rises, unique reach does not.
   - **RevOps and GTM engineering.** Owns the data model, systems of record, workflow, and access control. Decision: get the stack diagram, the data model, and the workflow blueprint before the tool is bought. Failure mode: the pilot works in a clean sandbox and breaks on live CRM data quality nobody audited.
   - **Enablement.** Owns ramp time, coaching, and the skill stack reps are held to. Decision: decide what the new rep is trained on when the tool does the first draft. Failure mode: tool rollout treated as enablement, with adoption measured by logins rather than behavior change.
   - **Customer success.** Owns retention, net revenue retention, and expansion. Decision: decide which signals an agent may act on alone and which require a human. Failure mode: automated outreach reaches accounts that were already unhappy, and churn risk is discovered after the renewal.
   - **Partnerships and business development.** Owns ecosystem sourced and influenced revenue. Decision: decide whether the ecosystem motion produces lift you can separate from direct. Failure mode: influenced revenue claimed twice, with no attribution rule separating partner lift from direct.
   - **Exec and founders.** Owns the AI position the board is told and the budget behind it. Decision: hold a position on AI you can defend to a board without hedging. Failure mode: the narrative ships before the evidence, and the reversal, when it comes, is public.
   - **Revenue finance.** Owns spend, unit economics, and what the investment returned. Decision: decide what the spend bought, in numbers that survive an audit. Failure mode: cost per seat is tracked, cost per outcome is not, and usage-based inference cost is discovered at renewal.
3. **Set kill criteria per seat before launch.** For each seat write the number that, if unmet by a stated date, ends the deployment.
4. **Name the owner for every seat.** An empty seat is the risk, and it is where reversals start.
5. **Build the Eight-Seat Read scorecard.** For each seat report one dollar metric, one adoption metric, one quality metric, and one reversal metric, producing a thirty-two cell operating view (https://www.therevenueaireport.com/frameworks/eight-seat-read).
6. **Re-read at the review date from the seat that is worst off,** not the seat that sponsored it.

## Decision rules

- Treat any seat you cannot fill in as an unowned risk, not an omission. That is the published purpose of forcing the read to be complete before it is published (https://www.therevenueaireport.com/frameworks/eight-seats).
- Read from eight seats because a deployment crosses functions. It is bought by one seat, configured by another, staffed by a third, and paid for by a fourth, so a single-seat read produces a conclusion the next seat can disprove (https://www.therevenueaireport.com/frameworks/eight-seats).
- Use the four fixed metric classes and no others: dollar covering pipeline, ARR, and cost per outcome; adoption covering active seats, usage frequency, and coverage; quality covering accuracy, hallucination rate, and escalation rate; reversal covering tools shut off, spend clawed back, and seats reallocated (https://www.therevenueaireport.com/frameworks/eight-seat-read).
- Do not run the Eight-Seat Read on teams under fifty employees where a single seat covers three or more of the eight functions (https://www.therevenueaireport.com/frameworks/eight-seat-read).
- Do not run it during the first ninety days of a deployment. The reversal and quality metrics need at least a quarter of production data (https://www.therevenueaireport.com/frameworks/eight-seat-read).
- Update quarterly, with monthly delta reads on the reversal and adoption metrics (https://www.therevenueaireport.com/frameworks/eight-seat-read).
- Keep individual company Reads private and publish only anonymized medians, matching the published methodology (https://www.therevenueaireport.com/data/eight-seat-read).
- Give the RevOps seat the stack diagram, data model, and workflow blueprint before purchase, not after. That sequencing is the published decision for that seat and it is the one that prevents the clean-sandbox failure (https://www.therevenueaireport.com/frameworks/eight-seats).
- Give the partnerships seat a written attribution rule before counting influenced revenue, because the published failure mode is influenced revenue claimed twice.
- Expect the partnerships seat to have the thinnest evidence. Report research on GTM job postings found six of the eight seats are losing definition rather than headcount, marketing leadership shows the sharpest title-level movement, RevOps and GTM engineering is the only seat pointing up, and partnerships carries no series with a stated sample size (https://www.therevenueaireport.com/research/eight-seats-baseline).
- Run the labor-market baseline before attributing any seat change to AI. Marketing postings were already 18 points below their February 2020 level in January 2025, before the 2025 wave, and Sales held through all of 2025 and only broke in the first half of 2026 (Indeed Hiring Lab postings index via FRED, indexed to February 2020 = 100, seasonally adjusted, United States, https://www.therevenueaireport.com/research/eight-seats-baseline).
- Do not average two sources that measure different universes. Bain's Aura tool puts marketing postings at minus 30 percent and sales at minus 27 percent year over year for Q1 2026 with no stated sample size, while Indeed's index says minus 6.0 and minus 5.0 points over nineteen months. Where they disagree, report both and average neither (https://www.therevenueaireport.com/research/eight-seats-baseline).
- Weight the customer success seat's quality gate against the rollback base rate. Three-quarters of enterprises have rolled back a customer-facing AI agent at least once, n = 2,527 (https://www.therevenueaireport.com/research/eight-seats-baseline).
- The Eight-Seat Read dataset publishes a schema and no observations. There are no benchmark medians yet, so set each seat's threshold internally, write it into the scorecard, and label it as internally set rather than benchmarked (https://www.therevenueaireport.com/data/eight-seat-read).

## Output requirements

Deliver two artifacts. First, the seat read, one row per seat, in the published order.

| Seat | Owns | The one decision | Failure mode exposure | Named owner | Kill criteria, metric, threshold, date |
|---|---|---|---|---|---|
| 01 Sales | | | | | |
| 02 Marketing | | | | | |
| 03 RevOps and GTM engineering | | | | | |
| 04 Enablement | | | | | |
| 05 Customer success | | | | | |
| 06 Partnerships and BD | | | | | |
| 07 Exec and founders | | | | | |
| 08 Revenue finance | | | | | |

Second, the Eight-Seat Read scorecard using the published column definitions, one row per seat per metric class.

| quarter | seat | metric_class | median | delta_qoq | panel_n |
|---|---|---|---|---|---|
| | | dollar, adoption, quality, or reversal | | | |

Also deliver: the list of empty seats named as unowned risks, the seat that is worst off, and the review date with the seat the re-read will be run from.

## Verification loop

Validate the read before it is used in a purchase, renewal, or board decision.

1. Count the filled seats. All eight must have an owner, a decision, and a failure mode. Any blank is recorded as an unowned risk in the output, not left empty.
2. Confirm every kill criterion has a metric, a threshold, a date, and one person who can trigger it alone.
3. Confirm the scorecard has all four metric classes for every seat. A seat missing its reversal metric produces a read that cannot show what was shut off, which is the class the framework says most scorecards omit.
4. Confirm the deployment has at least ninety days of production data before quality and reversal cells are filled, and that the organization is above fifty employees with distinct seats.
5. Fix every gap and repeat checks 1 through 4 across all eight rows, because naming a new owner usually changes that seat's kill criteria as well.

Only proceed to present the read when all eight seats are filled or explicitly marked as unowned risks, and all four checks pass on the same version of the document. If the deployment is under ninety days old, stop and deliver the seat read alone, marking the scorecard as not yet readable rather than filling quality and reversal cells with early data.

## Quality checks

- Exactly one named deployment, not a category.
- All eight seats present, in the published order.
- Every kill criterion names a metric, a threshold, a date, and a single trigger owner.
- All four metric classes present for every seat in the scorecard.
- Empty seats are listed as unowned risks with a name to be assigned.
- Every external number carries its publisher, sample size, and field date.
- Internally set thresholds are labelled as internally set.

## Limitations

- The Eight-Seat Read is an operating scorecard. It shows exposure and movement; it does not decide whether the deployment is strategically correct.
- The published dataset is header-only today. No observations have been published, so there is no median to benchmark a company's cells against (https://www.therevenueaireport.com/data/eight-seat-read).
- Seat-level labor evidence is uneven. Partnerships carries no series with a stated sample size, so conclusions about that seat are weaker than the others (https://www.therevenueaireport.com/research/eight-seats-baseline).
- No function-level layoff breakdown is published, so any sales-only or marketing-only layoff total used in a seat read is an estimate (https://www.therevenueaireport.com/research/eight-seats-baseline).
- The framework does not fit organizations under fifty employees where one person covers three or more seats.

## Example input

A proposed AI SDR platform for a 600-person B2B company. Sponsored by the VP of Sales. Contract includes usage-based inference charges. No stack diagram has been requested. Partnerships was not consulted. Illustrative and synthetic, provided to show output shape.

## Example output

Seats filled: six of eight. Sales, marketing, RevOps, enablement, customer success, and revenue finance have named owners and kill criteria.

Unowned risks: partnerships and exec. Partnerships has no attribution rule separating partner-influenced from direct revenue, which is the published failure mode for that seat. Exec has no written AI position, so the narrative would ship before the evidence.

Worst-off seat: revenue finance. Cost per seat is in the contract and cost per outcome is not defined anywhere, and usage-based inference charges are not modelled, which is the failure mode the framework names for that seat.

Kill criteria examples: sales, meetings accepted rate below the current human baseline by day 90, triggered by the VP of Sales. Revenue finance, cost per accepted meeting above the current manual baseline for two consecutive months, triggered by the Director of Revenue Finance.

Scorecard status: not yet readable. The deployment has zero days of production data, and quality and reversal cells need at least a quarter (https://www.therevenueaireport.com/frameworks/eight-seat-read).

Recommendation: do not sign until RevOps has the stack diagram, the data model, and the workflow blueprint, and until partnerships and exec have named owners. Flagging for human review by the CRO.

## Rules of conduct

- Write for a Director, VP, or operator. Short sentences. Explain uncommon terms.
- Separate facts from assumptions. Never hide uncertainty.
- Do not invent numbers, benchmarks, quotes, or customer names.
- Do not send messages, change CRM records, or publish anything unless the user explicitly asks.
- Flag when a decision needs human review.

## Evidence

- https://www.therevenueaireport.com/frameworks/eight-seats
- https://www.therevenueaireport.com/frameworks/eight-seat-read
- https://www.therevenueaireport.com/data/eight-seat-read
- https://www.therevenueaireport.com/research/eight-seats-baseline
- https://www.therevenueaireport.com/research/rollback
- https://www.therevenueaireport.com/frameworks/reversal-ledger
- https://www.therevenueaireport.com/frameworks/proof
- https://www.therevenueaireport.com/frameworks/optimization-theater
- https://www.therevenueaireport.com/data/reversal-ledger

## Cite this framework

Kvarfordt, Jonathan. "The Eight Seats." The Revenue AI Report. https://www.therevenueaireport.com/frameworks/eight-seats

What does the Reading a Deployment Across Eight Seats skill do?

Read one AI deployment from eight functions so the conclusion cannot be disproved by a seat that was never asked, and report it on the thirty-two cell Eight-Seat Read scorecard.

Where does the Reading a Deployment Across Eight Seats skill come from?

Report framework library. It was written by The Revenue AI Report against a 12 criterion quality rubric and graded in an independent scoring pass.

Why was the Reading a Deployment Across Eight Seats skill chosen for this library?

Encodes a Revenue AI Report framework so the agent applies the published method instead of improvising one.

When should the Reading a Deployment Across Eight Seats skill not be used?

Do not use it for: Score this one initiative out of five and tell me whether to renew. Or: Our AI reporting looks great and revenue is flat. Diagnose the pattern. Or: Write a job description for a RevOps analyst.

How do I install the Reading a Deployment Across Eight Seats SKILL.md file?

Download the file, create a folder named exactly reading-a-deployment-across-eight-seats inside your agent's skills directory, and save the file inside it as SKILL.md. The agent loads it when a request matches the description.