How to measure a CRO in the AI age
The median CRO lasts twenty-three months. Next year’s plan will ask every revenue leader to show AI lift, and most scorecards cannot — they were built to measure people, and the engine now has agents in it. Here are the Three Measures applied to the revenue organization, and the eight numbers a board should ask for.
The median chief revenue officer now lasts about twenty-three months in the seat, by Carta’s count. Two years of AI pilots and AI spend are behind every revenue organization — company spend on AI tokens has risen roughly thirteen-fold since the start of 2025, by Ramp’s index — and next year’s plan is the one in which the CEO and the CFO will ask the revenue leader to show what it bought. Most cannot, and not because nothing happened. They cannot because the scorecard they carry was built to measure people, and the engine now has agents in it.
Why the usual scorecard fails
The usual scorecard has two halves and both mislead. The first half is usage: seats activated, tools adopted, prompts per rep. Usage is an input, and Gartner reported this spring that around nine in ten executives see no measurable productivity impact from their AI investment — which is what usage without redesign looks like from the top. The second half is the board’s standing metrics: attainment, forecast accuracy, ramp time, revenue per dollar of sales compensation. Those are the right outcomes, but read alone, in the year the work is being redesigned, they dip before they rise — the same J-curve that cancels good enterprise AI programs cancels good revenue leaders.
What the board needs is not more metrics. It is the Three Measures — inputs, outcomes, organization — applied to the revenue organization itself, read together, on one page.
Measure one — Inputs, weekly
The question: are sellers and agents doing more of the work that produces revenue, and is it costing what we said?
Four inputs, all from the systems, none from a survey. The share of each seller’s week spent with buyers rather than on research, CRM hygiene and internal prep — the number AI should move first and most. The share of pipeline sourced or advanced by agents, kept separate from human-sourced pipeline so the two can be compared on conversion. Spend on AI per account executive against the limit written in the mandate. And adoption at week twelve among the sellers the tools were bought for, not the launch-week number.
Measure two — Outcomes, monthly and quarterly
The question: is the revenue engine measurably more productive, in numbers Finance will sign?
The outcome that matters most is productivity per seller: ARR per account executive, and — this is the change — ARR per selling agent, human or machine, because a revenue organization is now a headcount of both. Around it sit the standing board metrics, each with a before-and-after: quota attainment across the distribution (AI should lift the middle, not the top), forecast accuracy, ramp time for a new seller, revenue per dollar of sales compensation, and payback on customer acquisition. SaaStr’s public benchmark that half a million dollars of ARR per employee is the new two hundred thousand is the direction of travel, and the board will use it whether or not the CRO does.
Two columns, never merged: what the telemetry says was saved, and what the controller will book. The gap between them is the CRO’s realization gap, and reporting the first as if it were the second is how revenue leaders lose the CFO in month nine.
Measure three — Organization, quarterly
The question: has the work itself changed, or has a tool been added to the old work?
This is the measure that predicts whether the outcomes will arrive, and it is scored with yes-or-no questions. Have the ratios been redesigned — SDR to AE, SE to AE, manager to rep, RevOps to AE — on an explicit assumption about what share of the work agents now do? Has a role description changed? Has a step been removed from the sales process rather than accelerated? Is every agent in the revenue organization on the inventory, under a mandate, with an evidence feed someone reviews? Do reps sell while RevOps and GTM engineers build — or has the organization drifted into a hundred private AI experiments that no one can verify? The leaders I talked with this summer watched productivity fall when tinkering surged; the organization score is where that shows up before the bookings do.
The eight numbers a board should ask for
- 01OutcomeARR per seller, human and agent, this year against last
- 02InputShare of each seller’s week spent with buyers, measured from calendars and systems
- 03InputPipeline sourced or advanced by agents, and its conversion against human-sourced pipeline
- 04OutcomeForecast accuracy, with the forecast signed by Finance
- 05OutcomeRamp time for a new seller, before and after AI onboarding and coaching
- 06OutcomeRevenue per dollar of sales compensation, quarter over quarter
- 07InputAI spend per account executive against the mandate’s limit
- 08OrganizationThe organization score: ratios redesigned, roles changed, steps removed, every agent inventoried and mandated
How to read them together
The dip. Hold.
Sellers are with buyers more, agents are sourcing pipeline, the ratios have been redesigned — and productivity per seller is worse this quarter because the process is being rebuilt. This is what a good year looks like at month four. Set the date for the next reading.
Adoption without redesign. The tinkering trap.
Tools are being used inside the old process and reps are building their own. Productivity will stay flat until a step is removed, a ratio changes, and building moves to RevOps. This is a management decision, not a purchase.
Stop, or start over.
If the sellers the tools were bought for have stopped using them by week twelve, no outcome number will save the story. Find out whether the workflow, the mandate or the tool was wrong, and restart on purpose.
Scale. Set the ratios for the year after.
Productivity per seller is rising with Finance’s signature on it. Widen the agents’ mandates, set next year’s ratios on the evidence, and move the number into the plan before the CFO does it for you.
Before next year’s plan: five moves
Baseline this quarter. ARR per seller, buyer-facing share, attainment distribution, forecast variance, ramp time — captured now, so the story next year has a before.
Separate the columns. Telemetry-verified and Finance-verified value never share a cell in the board deck. Report the gap as a number and own closing it.
Write the ratios yourself. Decide what share of SDR, SE and RevOps work agents will carry next year and model the ratios on it. If the CFO builds this model first, it will be built without you.
Put every agent on the inventory. Prospecting, enrichment, call notes, forecasting: each one with an owner, a limit, and an evidence feed. It takes a week, and it is the difference between a forecast someone can stand behind and one nobody can.
Agree the Finance rule now. One conversation with the controller, this quarter, on what evidence turns a telemetry number into a booked one. It prevents the month-nine argument that ends tenures.
The AI ROI Scorecard
The three measures with baselines, deltas, the two value columns, the organization score and the computed reading — one sheet per workflow, or one for the revenue organization as a whole.
Download the workbook »The Revenue Diagnostic
Thirty questions across six pillars, including the one no other diagnostic scores: the AI agents already inside your revenue engine.
Run it »Questions this piece answers
How should a board measure a CRO in the AI age?
On three measures read together: inputs (buyer-facing time, agent-sourced pipeline, AI spend against the mandate, adoption at week twelve), outcomes (ARR per seller human and agent, attainment, forecast accuracy, ramp, revenue per dollar of comp, all Finance-signed), and organization (ratios redesigned, roles changed, agents inventoried). One number alone cancels good leaders in the dip.
Why does AI adoption not show up in bookings per account executive?
Because adoption is an input and bookings are an outcome, and between them sits redesign. Tools used inside the old process, and reps building their own apps, raise usage and lower productivity; bookings per AE move only after a step is removed, a ratio changes and building moves to RevOps.
What should a CRO do before next year’s planning cycle?
Baseline productivity per seller and buyer-facing time this quarter, separate telemetry-verified from Finance-verified value, model next year’s SDR, SE and RevOps ratios on an explicit assumption about agent work, put every agent on the inventory with a mandate, and agree the Finance rule with the controller before the plan is written.
This is the fourth piece in the Operator Edition of Field Notes on the Agentic Enterprise. Previous: The Three Measures. Start of the series: The Agentic Economy Runs on Authority.
Field Notes on the Agentic Enterprise
The six essays, the eight questions underneath them, and the Operator Edition tools — in order.
Open »The revenue engine now has agents in it
Every engagement includes the inventory, the mandates and the three measures for the agents in your own revenue organization.
How I work »New essays and operator tools, as they publish.
A few times a month, by email. No list is sold or shared; one click to leave.