We are careful with multipliers. Everyone selling to engineering teams has one, and they rarely say what was measured. So here is ours, with the shape of it, from the first teams running sleds end to end: intake from the business, a band before the work, a seat that fits, a receipt after.
Where it was about 3x
Throughput of requests landed per week, against the same team's previous quarter, roughly tripled. Most of that was not agents working faster. It was work that used to wait now not waiting: the request arrived with its done line, the band meant the yes came in minutes, routing put it on a seat with headroom the same morning. The build was the same build. The gap around it collapsed.
Where it was closer to 10x
Bugs from support. The path from a customer's report to a merged fix and a reply to the customer used to be measured in days, most of it queueing: support to product to engineering and back, each hop a re-explanation. With intake filing from the ticket, Sentry attaching the evidence, worth answered by one tap and the receipt going back to the desk, the median went from days to an hour or two. Nobody got faster. The hops disappeared.
The other 10x was people who had never filed anything. Designers, customer success, the founder. The requests had been in their heads because the tracker was not theirs. When the way to file was the Claude tab they already had open, they filed, and some of the best work of the quarter came from rows that would not have existed.
Where it was 1x
Deep, uncertain engineering work: the migration nobody had scoped, the performance problem with no clear cause. A band on that work is wide, and sleds says so. It does not make hard problems easy; it keeps them from hiding among easy ones, and it tells the business the honest width.
What we take from it
The multiplier is on the seam, not on the engineer. Teams that bought a faster agent got a faster build. Teams that connected the people who want things to the people who make them got a faster everything else, and the everything else was most of the time.