In board meetings, I increasingly hear some version of the following statement:
“Our sales organization is highly AI-enabled.”
My response: “How do you know? ”
Is it because the team is using an AI assistant to write emails? Because an agent is entering information into the CRM? Because the company has purchased several copilots, automation platforms, buyer intent discovery tools, or forecasting applications?
Those activities may demonstrate adoption. They do not necessarily demonstrate impact.
AI is now a meaningful line item in nearly every operating plan. Yet relatively few management teams can confidently explain whether that spending is increasing revenue, lowering customer-acquisition costs, improving retention, or generating an acceptable financial return.
Sales gives us a useful starting point because sales productivity is among the most measurable outputs in an organization. It’s hard to claim that Mary is our top engineer by 6%. It is easier to claim that Julie is our top AE by 7%.
The primary metric I recommend is, no surprise, Productivity Per Rep, or PPR: the amount of new revenue generated by the average account executive over a defined period, such as a quarter. If an AI investment is making the sales organization more productive, PPR should improve.
PPR alone, however, is not sufficient. An average can be distorted by a small number of exceptional performers, so we should also measure the percentage of account executives achieving quota. If the benefit of AI is broad-based, productivity should improve across the team rather than being concentrated among a few top sellers.
Finally, higher short-term productivity cannot come at the expense of customer quality. A company can temporarily increase revenue by selling to poor-fit customers, overpromising capabilities, or failing to set appropriate implementation expectations. The result may appear positive initially, only to surface later through churn, contraction, or poor customer economics.
For that reason, the scorecard should include three outputs:
Productivity Per Rep as the primary measure, quota attainment as the distribution check, and the Leading Indicator of Retention by acquisition cohort and account executive as the quality check.
Healthy AI investments should improve productivity broadly while preserving—or ideally strengthening—customer fit, retention, and lifetime value.
To further dissect how AI should improve Productivity Per Rep, we can use the Revenue Velocity Formula:
Productivity Per Rep = Active Opportunities × Close Rate × Average Contract Value ÷ Sales Cycle
For example, an account executive managing 20 active opportunities, with a 30% close rate, a $100,000 average contract value, and a two-quarter sales cycle would produce approximately $300,000 of new revenue per quarter.
The formula gives management teams a practical way to evaluate AI initiatives. Rather than beginning with a list of available tools, they can begin with the operating constraint they are trying to improve.
Our initial hypothesis is that the four variables do not offer equal potential.
Average contract value is probably the least attractive target. Pricing is primarily a function of product value, positioning, packaging, and market structure. There is little reason to believe that adding AI to the sales workflow will, by itself, allow a company to charge significantly more. Instead, pushing on a higher ACV could unintentionally push the company into a market segment where they lack Product-Market and Go-to-Market Fit.
AI can improve seller preparation, provide real-time coaching, recommend next steps, generate customized demonstrations, diagnose skill gaps, and strengthen deal strategy. These capabilities should help sellers win more often and execute more efficiently. We are seeing meaningful gains in these inputs with the use of AI.
The constraint is that both metrics remain dependent on the buyer. Even if AI prepares every follow-up, updates the CRM immediately, creates a customized demo, and generates a mutual action plan, the buyer still needs to evaluate alternatives, coordinate stakeholders, secure budget, and obtain executive approval on their side. The seller’s workflow may become dramatically faster while the customer’s buying process remains largely unchanged.
Unlike buyer approval processes, this variable is completely under the company’s control. AI can reduce the administrative burden associated with account research, prospect identification, meeting preparation, CRM maintenance, forecasting, follow-up, proposal creation, and RFP responses. When those efficiencies are real, a seller can manage more qualified opportunities without compromising execution quality.
The most important underlying KPI may therefore be selling time: the number of hours each week that a seller spends interacting directly with prospects and customers.
Historically, even strong sales organizations have struggled to generate 15 hours of weekly selling time, while many remain below 10. Our working hypothesis is that an elite, AI-enabled organization could eventually reach 20 to 30 hours—not by asking sellers to work twice as much, but by removing a substantial portion of the non-selling work surrounding each customer interaction.
That is one of the clearest ways to understand AI’s immediate value in sales. The opportunity is not simply to replace sellers. It is to give good sellers more time with buyers.
The metrics also provide the guardrails. If sellers take on more opportunities without achieving the expected efficiency gains, the warning signs should appear through declining close rates, longer sales cycles, slower stage conversion, or weaker retention. AI should increase capacity without degrading quality.
The initial operating playbook is straightforward.
Build the measurement system before scaling the investment. Track Productivity Per Rep, quota attainment, retention by cohort, and each component of the Revenue Velocity Formula. Where possible, add selling time as a leading indicator (I’d love to hear the community's experiences in doing this).
Organize existing and proposed AI initiatives according to the operating metric each is expected to influence. The objective is not perfect attribution. It is a practical comparison of expected business impact, implementation effort, operating cost, and execution risk.
Test before scaling. Establish a pilot group, benchmark its performance, and measure whether the targeted revenue-velocity metrics improve. Expand the initiative only when the improvement is repeatable, broad-based, and accompanied by healthy customer outcomes.
This framework also informs how we think about the AI market as investors. “AI-enabled” is no longer a sufficient product claim. We want to understand which customer outcomes change, how quickly customers can measure that change, whether the improvement survives outside a pilot, and whether the vendor can deliver it with sustainable gross margins and retention.
The next generation of durable AI companies will not merely help customers deploy more AI. They will help customers produce measurably better business results.
We view this framework as an initial model rather than a finished answer. Over the coming quarters, we plan to test, refine, and expand it with the collective experience of the Stage 2 LP community, our founders, and their executive teams. That combination of operating expertise and real-world portfolio data gives our community an opportunity to develop a more rigorous standard for evaluating AI productivity.
The organizations that win will not necessarily be those using the most AI. They will be the ones that connect AI investments to operational metrics, improve those metrics systematically, and convert the resulting gains into stronger unit economics, customer outcomes, and long-term enterprise value.