Engineering Leadership · AI Adoption
I'm Leif Nelson, an engineering leader with two decades of experience building and running teams, from org-building at Amazon to hands-on digital delivery in the agency world. I think a lot about adopting AI in a way that lasts: sustainable, maintainable, and honest about what today's tools actually do, and I like talking it through with the people making those calls.
AI is an amplifier of engineering discipline, not a substitute for it.
Treat it as a replacement and you create messes you'll spend years unwinding. Treat it as leverage, with clear eyes on the costs, and you pull ahead. The difference is judgment, not tooling.
What I keep seeing
AI doesn't remove the need for engineering judgment; it raises the stakes on it. The teams shipping fastest still have strong engineers. They've just changed how those engineers spend their time.
Agentic tooling isn't effortless or cheap. Token spend, review burden, and confident-but-wrong output add up quietly. Adoption that ignores these costs looks like a bargain, until the rework lands.
AI shifts some parts of the craft and leaves others untouched. Teams that can't tell the difference automate the wrong things, and erode the practices that kept their software maintainable in the first place.
Background
I started in video and front-end development, then spent years in the Minneapolis agency world, growing from senior engineer to senior technical director. In 2017 I joined Amazon, where over nearly eight years I led teams at IMDb, built two 30-plus-person orgs in Amazon Advertising, and helped build Amazon Autos from the ground up.
Today I'm Senior Director of Technology at Bold Orange, a customer-experience agency, where I lead our digital delivery teams. On the side I build tools and write about rethinking the software development lifecycle for the agentic age, which is where my take on durable AI adoption comes from: not slideware, but real teams shipping real software.
In practice
Automate and orchestrate your software development lifecycle
A Claude Code plugin: skills, subagents, and hooks that automate common SDLC tasks and enable end-to-end orchestration, from requirements through QA and deployment.
Contact
I talk with engineering and technology leaders who are working through AI adoption, comparing notes on experimentation, process, and what actually holds up in practice. If that's on your mind, I'd be glad to trade perspectives.
Thanks, that's on its way. I read every note myself, and usually reply within a day or two.