How Kraken finds hidden bottlenecks across thousands of engineers | Nik Sudan

Operationalizing AI in engineering requires a clear separation between rapid, throwaway proof-of-concept experimentation and production-ready development. Scaling AI maturity depends on moving beyond vanity metrics like adoption rates or token consumption, instead utilizing a balanced framework of throughput, quality, and stability to measure actual business impact. Engineering leaders must act as translators, converting technical data into plain language to align non-technical stakeholders with engineering-led initiatives like tech debt reduction. By using tools like LinearB to aggregate data and identify hidden bottlenecks—such as time zone-induced review delays—teams shift from anecdote-driven to evidence-driven decision-making. This disciplined approach ensures that AI serves as a force multiplier for engineering output rather than an expensive, unverified experiment, ultimately fostering organizational trust and long-term technical health.