Descriptions:
Mike Spitz, an engineering leader at PFF — a sports data company serving NFL and NCAA teams with 100 million annual page views and 9 million fantasy drafts — presents a detailed case study of running a two-engineer AI-native team against a conventional ten-engineer team over roughly two months, from January to March.
The results are striking: the two-person team deployed five times per day while the ten-person team averaged one deployment every five days, a 25x difference in deploy frequency. On output quality, PFF blended ticket count with code complexity estimates and found approximately 10x higher output from the smaller team. Customer satisfaction scores moved from a previous average of 7–7.5 out of 10 to 8.6, suggesting the speed increase did not come at a quality cost. Spitz is candid about the most obvious confound — smaller teams are inherently faster — while noting the two engineers still had to coordinate daily with the larger team, partially controlling for that variable.
The process changes that enabled this are as significant as the numbers. Scrum ceremonies were dropped: no standups, no sprint refinement, no retrospectives in the traditional sense. Replaced with every-other-day huddles involving engineers, product, and design. Ticket creation, status updates, and dependency flagging are automated. The case study closes with a talent observation: the engineers who thrived were curious self-starters comfortable with ambiguity, while those who needed prescriptive specs struggled — a shift Spitz frames as a permanent change to what the engineering role requires.
📺 Source: AI Engineer · Published May 15, 2026
🏷️ Format: Workflow Case Study
