Introduction
Kraken moves fast. The team prizes a lean, startup pace, and AI has put that speed in everyone's hands, including non-engineers. But shipping AI-powered ideas from pilot to production, and proving the spend pays off, is where most teams stall.
Nick Sudan
Measuring whether AI is actually working
Adoption dashboards and seat counts look great, but they don't tell you whether the work got better. Kraken balances throughput, quality, and stability instead, and watches cost per contribution, or AI spend divided by impact. LinearB provides the high-level signals that anchor the picture, such as merge frequency, PR size, rework rate, and cycle time.
Finding the real bottleneck
Review time is the biggest constraint on cycle time at Kraken, and at most companies. To find out why, the team pairs the LinearB MCP server with their git provider data and lets AI move between the two. LinearB surfaces the trend and the relevant merge requests, repositories, and teams; the git layer fills in the detail.
The results
The fix, expanding the pool of code owners so a review never waits on a single time zone, came straight from the data. The bigger shift is cultural. Leaders now work from evidence instead of anecdote, and engineers and non-technical stakeholders share one language for the health of the org.
