I used to think leadership was mostly about technical direction. I still think direction matters — a lot — but experience has corrected me. The hardest distributed system in any organization is the people system. Incentives, fear, pride, curiosity, burnout, trust. All of it runs in production, all the time, with no staging environment.
Leading engineering through a platform shift — especially Generative AI — is not a tooling problem dressed up as strategy. It is a learning problem at organizational scale.
What I have led
My leadership work has focused on three intertwined outcomes:
- Rebuilding teams that deliver — growing a mission-critical group from 1 to 12 engineers without losing reliability, and establishing the habits (review, observability, ownership) that survive turnover.
- Making governance a force multiplier — tiered architecture review, SLO standards, cost visibility for LLM workloads, and guardrails clear enough that teams move faster inside them, not slower around them.
- Transforming how engineers relate to AI — from threat or fad to amplified expertise, through training, guilds, pairing, hackathons, and production-grade patterns they can reuse on Monday morning.
The GenAI adoption lesson
When we rolled out enterprise Generative AI adoption, the breakthrough was not the first impressive demo. It was the hundredth engineer who could build a governed agent without asking for permission slips in triplicate.
We trained 200+ engineers on prompt engineering, RAG patterns, and agentic development. We stood up an AI guild so knowledge did not die in silos. We paired on production deployments until “AI project” stopped meaning “special exception” and started meaning “another system we operate responsibly.”
Leadership as architecture
In my view, good engineering leadership applies the same principles as good system design:
- Reduce ambiguity — people cannot execute what they cannot see. Standards, templates, and explicit decision rights lower friction.
- Design for feedback loops — metrics, retros, incident reviews, and visible cost data. Organizations learn the way people do: by observing outcomes, not by memo.
- Protect expertise — senior domain knowledge is more valuable in the AI era, not less. Leadership should make experts more effective, not obsolete.
- Match governance to maturity — lightweight paths for experiments; rigorous paths for member data and production agents. One-size-fits-all governance is just fear in a blazer.
A question for every leader
Are you asking your teams to adopt AI — or are you giving them a credible path to adopt it well?
That difference shows up in retention, delivery velocity, incident rates, and whether your best people stick around to build the second wave. The work below is how I have tried to answer that question in practice.