How do you measure an AI sales lift honestly when the market is also moving?
Most pilots get graded against last year, and last year always loses. This team refused to run that play.
A manufacturer with a territory-based outside sales force refused to grade its AI pilot against the prior year. It compared pilot reps against non-pilot reps selling the same products into the same conditions over the same period, subtracted what the market was doing anyway, and published the roughly five percent lift in sales per account that survived.
- Industry
- Territory-based manufacturing
- Who uses it
- Outside sales
- What it does
- AI sales lift measurement
- Measured
- Single pilot month, 2026
Their pilot month came in well ahead of the same month a year earlier, by enough that the deck would have written itself. But the market had also moved during that window, and a rising market makes everyone look like a genius.
So they took the number away from themselves. They compared the reps in the pilot against reps outside it, selling the same products into the same conditions over the same period, and subtracted what the market was doing anyway. What survived that subtraction was a lift of roughly five percent on sales per account.
That is the figure they publish internally, and it is the one we publish here.
The reps who used it daily did considerably better than the average. Their sales rose close to thirty percent. The strongest performer on the team was up nearly seventy percent in sales and almost doubled units moved.
The mechanics behind it are unglamorous, which is the point. A rep works through a queue of prioritized accounts. A dozen campaigns get activated instead of forgotten. A couple dozen product-and-account combinations that had never been paired get paired. Roughly a dozen accounts reorder products they had quietly stopped buying, which nobody had noticed because noticing would have required somebody to compare this quarter's order lines against last year's, account by account, on a Tuesday.
One rep got flagged that a campaign had passed an account by. He called. The account ordered.
Measurement window
Single pilot month, 2026. Lift measured against a matched non-pilot rep cohort over the same period rather than against the prior year, with market movement subtracted. Daily-user and top-performer figures cover the same month.
Questions this answers
1Why is year-over-year the wrong comparison for an AI sales pilot?
Because a rising market inflates the result. This company compared pilot reps to a matched cohort of non-pilot reps over the same period and subtracted market movement, which cut the headline substantially and left a defensible number.
2What lift is realistic from an AI sales assistant?
At this company, roughly five percent on sales per account across the full pilot population after controlling for the market, with the daily-user cohort up close to thirty percent.
