If AI Raises Rep Capacity, Should Quotas Rise With It?

Por Brian Tr · Atualizado em 11 de outubro de 2026

The logic arrives in the budget meeting, usually from finance. Reps now have AI to research accounts and draft emails. They'll have more selling time. So next year's quota should go up, and the plan should assume the same headcount produces more.

It's a reasonable instinct, and eventually it will be right. Our view is that the order of operations is where companies get hurt. A quota raised on gains that haven't been measured is a pay cut with a delay on it. Reps feel it long before the tools deliver anything, and the ones who leave first are rarely the ones you'd choose.

The position we'd defend: raise your capacity assumptions only after you've measured a change in ramp time and conversion in your own team, and when you do raise them, share a visible part of the upside with the people producing it. The rest of this piece explains why, what the evidence does and doesn't support, and how to sequence the decision.

Key Facts

• Salesforce's State of Sales report, based on a survey of 4,050 sales professionals fielded in August and September 2025, says the average seller spends 40% of their time selling. Sellers expect agents to cut prospect research time by 34% once fully implemented. That's an expectation, not a measured result, and Salesforce sells the tools.
• The same Salesforce survey reports that 54% of sellers say they've used agents, and nearly 9 in 10 plan to by 2027. Adoption isn't the same as proven capacity.
• In an NBER working paper on 5,179 customer support agents, an AI assistant raised issues resolved per hour by 14% on average and by 34% for novice and low-skilled workers, with minimal impact on experienced workers. It's a support setting, not a sales one.
• In a field experiment with more than 700 consultants, AI improved performance on tasks inside its capability, but on a task outside it, performance fell by an average of 19 percentage points, according to MIT Sloan's October 2023 summary.
• Research from the University of Houston's Bauer College finds that when salespeople miss their quota by a small margin, their probability of quitting increases, and repeated near misses push them further toward leaving.
• Ordonez, Schweitzer, Galinsky and Bazerman's 2009 paper, summarized in strategy+business, warns that stretch goals which are out of reach can demoralize employees and push them toward narrow or unethical behavior.

What a quota actually is

A quota is a promise about capacity. When you assign one, you're telling a rep: given your territory, your tools and your time, this is a number a competent person can reach. Compensation plans are built on that promise. So are forecasts and hiring plans.

That's why the AI argument is more dangerous than it sounds. If you raise the quota, you've changed the promise. If the capacity didn't actually arrive, you've broken it, and reps notice immediately because their paycheck is the instrument that measures it.

The finance version of the argument has a hidden assumption: that time saved on research and drafting turns into more closed revenue at the same rate. Sometimes it does. But a rep with two free hours a week still needs buyers who respond, accounts that are ready, and deals that don't stall in legal. Selling time is one input. It isn't the only constraint on bookings.

What the evidence supports, and what it doesn't

Be honest about how thin the sales-specific evidence is. Here's how we'd read what exists.

Selling time is a real, measured bottleneck. Salesforce's survey puts the average seller at 40% of their time spent selling. That's the headroom the finance argument points to. It's also a self-reported figure, from a vendor, and it tells you where time goes, not how much extra revenue recovered time produces.

The AI gains studied so far are uneven. The best-known field evidence on generative AI at work comes from customer support, not sales. The NBER study found a 14% average productivity gain, but 34% for novices and almost nothing for experienced workers. If that pattern carries into sales, and it might not, the gains concentrate in your newest hires, which is exactly the part of the team a quota increase hits least gently.

AI can also make people worse on the wrong task. The consultant experiment found that on a task outside the AI's capability, performance dropped by an average of 19 percentage points. The researchers' point is that the boundary is jagged: tasks that look similarly hard can fall on opposite sides of it. A rep using AI on an account where it misleads them can lose ground.

Missing quota has real costs for retention. The Bauer research on near misses is the most useful finding for this decision. Reps who come close to quota repeatedly, which describes an aggressive but not impossible target, are more likely to quit. That is the failure mode of a quota raised a notch too far. You don't get a team that sprints harder. You get a team that watches a good year become a disappointing one.

What nobody has shown, as far as we can find, is a controlled study of a sales team whose quotas were raised on projected AI gains. So we aren't claiming a specific attrition number. Our claim rests on mechanism and the retention research above: quotas are a promise, unproven gains can't back it, and the people who leave first are the ones with options.

Illustrative math: when the plan outruns the gains

This example is illustrative, not drawn from a real company.

Say a team has ten reps with a $1.0M quota each. Finance assumes AI will lift capacity by 20% and sets next year's quota at $1.2M per rep. The tools help, but the measured gain turns out to be 8%. Realistic capacity is about $1.08M per rep.

Against a $1.2M quota, a rep who performs at full realistic capacity attains 90%. Every rep working at that level misses by the same margin. In compensation terms, a typical plan would pay them less than the year before, for doing better work than the year before.

Now run it the other way. Hold quotas at $1.0M, measure through the first half, then raise targets only to what the data supports. Reps who beat the old number feel the gain. The company still captures the improvement, a quarter or two later.

The cost of waiting is a few months of slower upside. The cost of getting it wrong is a team full of near-misses. For most companies that is not a close call.

A decision table: what to change, and when

Two rules sit under the table.

1. Measure before you commit the number. The signal you want is a change in ramp time or conversion in your own pipeline, compared against a team not using the tool. Activity metrics, like emails sent or calls logged, will rise with almost any tool and don't prove revenue.
2. Raise in steps, and tie each step to evidence. A quota that moves in two tranches, with the second conditional on the first half's data, keeps the promise intact.

Share the upside, or expect the gain to go to someone else

There's a second half to the decision that gets skipped. Suppose you do measure a real gain. Who keeps it?

If every dollar of AI-driven improvement goes straight into a higher quota, the rep sees a harder target and the same pay rate. They've done the work of adopting the tool and received nothing visible for it. That teaches your best people that learning new tools is a trap.

Sharing doesn't have to mean giving back all of it. It can look like this:

• Split the measured gain. Raise the quota by part of the measured improvement, not all of it, and let the difference show up as attainment above 100%.
• Add accelerators. Pay a higher rate on bookings past the old quota, so the gain is shared by default rather than negotiated.
• Fund the people who adapt first. Early adopters often have the best data on what works. Treat them as designers of the new process, not as proof that everyone else should keep up.

How pay should change when AI handles more of the pipeline is a related design problem, and we cover it in our look at sales compensation when AI prospects. The principle is the same in both: don't ask the team to absorb the risk of a change you haven't proven.

The CFO objection

The strongest objection comes from finance, and it deserves a straight answer. The CFO says: "We're paying for these tools. If the team isn't expected to produce more, why did we buy them?"

Fair question, and the answer isn't to refuse to plan around AI. Build the board plan with two scenarios: a base case at current capacity and an upside case that adds measured gains, with the trigger written down. Give the reps the base case. Report the upside case to the board as the target the company is working toward.

That framing keeps finance honest too. A plan that assumes unmeasured productivity is a forecast risk, and our piece on forecasting discipline covers why CROs are better off with a number they can defend. And the case for AI in sales operations is usually made on research and admin time, as we cover in why sales ops is the highest-ROI AI use case, not on instant bookings.

What to do next

Treat this as a two-quarter measurement project, not a reorganization.

1. Freeze the quota assumption until you have data. Tell the team that this year's targets won't move on AI projections, and say what would change that.
2. Set up a real comparison. Give the tool to half of a comparable group of reps, or stagger rollout by team, and track ramp time, conversion by stage and attainment against a team without it.
3. Segment by tenure. The research suggests gains are larger for less experienced workers. Look at new hires separately from veterans.
4. Write the trigger. Decide now what result, over how many quarters, will cause you to rebase capacity. Putting it in writing keeps the debate from restarting every planning cycle.
5. Design the share. Decide how much of any measured gain goes to reps through accelerators or a smaller quota increase, before you announce anything.
6. Watch near misses. Track how many reps finish within 10% below quota. A growing cluster is your early warning, and it shows up well before resignations. Our guide on manager metrics for team quota is a starting point for what to track, and what a sales manager is for when AI reviews every call covers the coaching that should follow.
7. Connect it to retention. A quota change is a staffing decision as much as a planning one. See our piece on sales retention and new hiring for the cost of getting that wrong, and sales org design for how quota, coverage and headcount fit together.