Metrics That Survive Contact with 1,000+ Software Engineers: Measuring AI Across the SDLC

Everyone is rolling out AI across the SDLC — almost no one can prove it works. Adoption dashboards glow green, demos impress leadership, and yet nobody can say whether AI is actually saving engineering effort or quietly relocating it. This talk is a field report from a large GenAI adoption program, spanning 1,000+ engineers. I'll share the measurement framework we built to separate real impact from hype: which metrics survived contact with a live engineering organization, which collapsed into vanity numbers, and how we caught AI "wins" that were cost-shifting in disguise. Concrete cases include LLM-generated documentation, AI-accelerated end-to-end automation, and autonomous regression agents — each with the numbers behind it. You'll leave with a practical metric set for your own AI program: definitions, instrumentation pitfalls, and the traps we wish we'd known about earlier.
Format: SOLUTION SHOWCASE, TRANSFORMATIONAL TALK
Location: ROOM C
Speakers: Vladislav Mironov
Level: Intermediate
Audience: Registered participants