A CRO at a 58,000-person industrial automation company checks the recurring revenue page of the board pre-read three days before it goes out, with any AI chat her company allows.
A tidy draft page from revenue operations, finance's revenue bridge, customer success's export, and revenue operations' own notes on how each figure was built. Every number comes from a real report. They don't all count the same way.
Here the AI checks a board page where every number has a source, but not the same one. The prompt asks which customers each figure counts.
Use the AI tool your company approves: ChatGPT, Claude, Gemini or Copilot.
Copy the prompt and the four fictional inputs from the appendix into one message.
Work out net revenue retention from finance's bridge alone. Then decide what the board sees first.
I'm Chief Revenue Officer at an industrial automation company that sells equipment, service contracts and software subscriptions. Revenue operations has drafted the recurring revenue page of our board pre-read. It goes out Thursday, February 4, and the board meets Thursday, February 11. Today is Monday, February 1. Below are the draft page, finance's revenue bridge, the customer success export and revenue operations' notes on how the page was built.
Please give me:
Every figure on the page must come from one source and one definition, and say which customers it counts. Check the arithmetic. Don't add facts that aren't in the inputs; if you assume something, say so.
Exactly as it came back, not edited. Read it the way Helena did: check each point against the source it cites.
Five of the seven lines on the draft page don't hold as written. Most of the problems come from mixing two sources and two definitions. The draft adds a one-off $40M project, counts the Arden subsidiary twice, leaves out a customer that has already churned, and divides finance's revenue by customer success's site count. Corrected on finance's basis: run rate $820M (up 9%), net revenue retention 103%, gross revenue retention 92%, 1,240 customers, expansion $78M, average $0.66M.
Assumptions I've made (please confirm):
| Draft line | Verdict | Input that shows it |
|---|---|---|
| Recurring revenue $860M, up 15% | Doesn't hold | [2] ends at $820M. [4] adds the $40M Brightwater project, which finance calls "a one-off installation, invoiced on completion" (860 = 820 + 40). found |
| Net revenue retention 113% | Doesn't hold | [4] uses customer success's $96M expansion, which includes Arden's $18M that [2] counts as new business. It also drops Kestrel's $18M from churn, even though the contract ended Dec 31 [3], and adds the $40M Brightwater project. found |
| Gross retention 97% | Doesn't hold as labelled | [4]: it's a customer count, (1,150 - 38)/1,150 = 96.7%, shown next to revenue measures as if it were one. Gross revenue retention is 92%. It's also unclear whether the 38 includes Kestrel, which revenue operations says it left out. found |
| Customers 1,240, up 8% | Holds (legal entities) | [2]: 1,150 - 38 + 128 = 1,240. 1,240 / 1,150 = 1.078, so up 7.8%, about 8%. It depends on A1 and A3. found |
| Expansion $96M | Doesn't hold | [3]: it includes Arden Foods' new subsidiary ($18M). [2] counts that subsidiary as a new customer, so it's counted twice. 96 - 18 = 78, which matches [2]. found |
| New customers 128, adding $49M | Holds (legal entities) | [2]. Note it includes Arden's subsidiary at $18M. |
| Average per customer $0.62M | Doesn't hold | [4]: $860M / 1,380. That's revenue including Brightwater divided by customer success's site count (1,380 sites, [3]), not finance's 1,240 legal entities. found |
All figures use finance's bridge [2]: annual run rate, with customers counted as legal entities.
| Measure | Counts | Which customers |
|---|---|---|
| Run rate (ARR) | Revenue: annual recurring run rate at year end, excluding one-off projects | All 1,240 legal-entity customers at year end |
| Net revenue retention | Revenue | Only the 1,150 customers at the start of the year. Includes their expansion, downgrades and churn. Excludes new customers. |
| Gross revenue retention | Revenue | The same 1,150 start-of-year customers. Downgrades and churn only, no expansion. Can't exceed 100%. |
| Customer retention | Customers | The 1,150 start-of-year legal entities, and how many were kept (1,112) |
| Customers | Customers | Legal entities, as finance counts them. Not customer success's sites (1,380). |
| Expansion | Revenue | Existing legal entities only. A new subsidiary with its own contract counts as a new customer, not expansion. |
| New customers | Both (count and $) | Legal entities that signed in FY26, including the Arden subsidiary |
| Average per customer | Revenue ÷ customers | $820M ÷ 1,240 legal entities |
Recurring revenue, FY26. Source: finance revenue bridge; customers counted as legal entities.
*Assumes finance's churn includes Kestrel; finance to confirm by February 3.
Claude found every planted problem, recalculated each figure and named what finance must confirm. Helena's edits are about what the board sees first, how the fix lands with two teams, and one call it left to her.
Use it whenever a page for the board or the executive team pulls figures from more than one team: retention, customer counts, pipeline, headcount, cost.
I'm [your role] at a [industry] company. [Who] has drafted [the page] of our [board pre-read or report]. It goes out [date], and [the meeting] is [date]. Today is [date]. Below are [the draft page, the source reports, and notes on how the page was built].
Please give me: 1. Each line on the draft page: holds, doesn't hold, or unverified, with the input that shows it. 2. The corrected figures, with the calculation for each. 3. The definitions [the readers] should see: for each measure, whether it counts [revenue or customers], and which [customers] it includes. 4. What must be confirmed before [date], and by whom. 5. The revised page, no longer than the draft.
Every figure on the page must come from one source and one definition, and say which [customers] it counts. Check the arithmetic. Don't add facts that aren't in the inputs; if you assume something, say so.
Copy the prompt, then all four inputs, into one message in a new chat. Everything here is fictional and safe to paste.
I'm Chief Revenue Officer at an industrial automation company that sells equipment, service contracts and software subscriptions. Revenue operations has drafted the recurring revenue page of our board pre-read. It goes out Thursday, February 4, and the board meets Thursday, February 11. Today is Monday, February 1. Below are the draft page, finance's revenue bridge, the customer success export and revenue operations' notes on how the page was built. Please give me: 1. Each line on the draft page: holds, doesn't hold, or unverified, with the input that shows it. 2. The corrected figures, with the calculation for each. 3. The definitions the board should see: for each measure, whether it counts revenue or customers, and which customers it includes. 4. What must be confirmed before Thursday, and by whom. 5. The revised page, no longer than the draft. Every figure on the page must come from one source and one definition, and say which customers it counts. Check the arithmetic. Don't add facts that aren't in the inputs; if you assume something, say so. [1] Draft board page: recurring revenue, FY26 (revenue operations) - Recurring revenue: $860M, up 15% on FY25. - Net revenue retention: 113%. - Gross retention: 97%. - Customers: 1,240, up 8%. - Expansion from existing customers: $96M. - New customers: 128, adding $49M. - Average recurring revenue per customer: $0.62M. [2] Finance: FY26 recurring revenue bridge (annual run rate, $M) | Line | $M | Customers | |---|---|---| | Start of FY26 | 750 | 1,150 | | Expansion | +78 | | | Contraction (downgrades) | -15 | | | Churn | -42 | -38 | | New customers | +49 | +128 | | End of FY26 | 820 | 1,240 | Finance counts a customer as a legal entity. New customers include Arden Foods' new subsidiary ($18M), which signed its own contract. [3] Customer success export and notes - Expansion from existing customers: $96M. This includes Arden Foods' new subsidiary ($18M), which customer success counts as expansion of the Arden account. - Active customers: 1,380. Customer success counts each customer site separately. - Kestrel Grid ($18M a year) gave notice in October. Its contract ended December 31. We hope to win them back with a new deal in the second quarter. [4] Revenue operations: how the page was built - Expansion uses the customer success figure. New customers use the finance figure. - Kestrel is left out of churn and retention because customer success says talks are still open. - Recurring revenue includes the $40M Brightwater upgrade project, which the region booked as recurring. Finance's note: "Brightwater is a one-off installation, invoiced on completion." - Net revenue retention: (750 + 96 - 15 - 24 + 40) / 750. Gross retention: customers kept, (1,150 - 38) / 1,150. Average per customer: $860M / 1,380.I'm Chief Revenue Officer at an industrial automation company that sells equipment, service contracts and software subscriptions. Revenue operations has drafted the recurring revenue page of our board pre-read. It goes out Thursday, February 4, and the board meets Thursday, February 11. Today is Monday, February 1. Below are the draft page, finance's revenue bridge, the customer success export and revenue operations' notes on how the page was built.
Please give me:
Every figure on the page must come from one source and one definition, and say which customers it counts. Check the arithmetic. Don't add facts that aren't in the inputs; if you assume something, say so.
| Line | $M | Customers |
|---|---|---|
| Start of FY26 | 750 | 1,150 |
| Expansion | +78 | |
| Contraction (downgrades) | -15 | |
| Churn | -42 | -38 |
| New customers | +49 | +128 |
| End of FY26 | 820 | 1,240 |
Finance counts a customer as a legal entity. New customers include Arden Foods' new subsidiary ($18M), which signed its own contract.