Some of the engineering management skills I spent years developing are becoming less valuable. I’m mostly happy about it.
I used to take a lot of pride in technical intuition. For example:
-
Guessing how long an implementation would take after a short discussion with the product team by knowing all the “gotchas” in an old codebase
-
Knowing when a “small rewrite” was actually a six-month migration that would shut down new feature development
Those skills still matter. But they are becoming less scarce - and not informed by intuition alone. What surprises me is how much more I enjoy the new parts of engineering management that AI makes possible. Experience gives me a good hypothesis. AI lets me test it.
AI has changed how technically engaged I can be, and it’s been very exciting to see my GitHub contribution history get greener and greener. Before AI, coding myself as the leader of a large team was typically an impediment. Now, I can pick up the things the team has no time for. For example, I cloned my last team’s monorepo onto a machine that had none of the necessary binaries configured for a particular UI and API service. I asked an agent to get the servers running locally with test data. In the past, that could easily have consumed half a day of archaeology. This time, I used that same amount of time to implement a small UI feature, test it, record a working demo, and get my feature merged and deployed.
On the people management side, AI has also had a huge impact.
-
Recently, I needed to identify engineers who might be a strong fit for a team move. Previously, the information gathering part of this task would have depended heavily on which people and projects I happened to remember. Instead, I could bring together standard sources: recent work, project roadmaps, pull request descriptions, ticket assignments, and past discussions on engineers’ strengths and goals.
AI helped me start with a more complete picture:
-
Who had relevant technical skills
-
Who enjoyed multidisciplinary and customer-facing work
-
Whose career goals might align with the opportunity
-
How each move would affect the receiving and sending teams’ roadmaps
-
Two of the three people that AI identified weren’t on my initial mental shortlist.
-
My daily briefing pulled together information from more than ten sources: meetings, project plans, calendars, messages, documents, tickets, and code activity. It improved my “manager SLA” - the lag between something needing me and me acting on it:
-
Which decisions were unresolved
-
Which commitment might have slipped
-
Which meeting needed a real agenda
-
Which person needed context or support
-
Two techniques made using AI indispensable for these tasks.
The first was connecting AI to real systems rather than limiting it to a chat window. I used internal tools and systems like GitHub, Jira, Snowflake, Datadog, Google Workspace, and Slack that were hooked up to MCP servers / CLIs to enable my agent to investigate and act across the same sources where work happens.
The second was building a trusted internal knowledge layer in a local Markdown wiki that consolidated project, org, and people context from verified sources, with links back to the evidence. The useful parts aren’t just the answers. It’s that I could see which answers are thin and need more work on my part, and I could pull others into my thinking without re-explaining the context.
The most important parts of engineering management haven’t changed. AI gave me a better picture of which engineers fit my proposed team move, but the final decision rested on my knowledge of who was looking for a change, how overloaded sending teams would be, and which organizational swim lanes made sense in the medium to long term.
I don’t know exactly what engineering management will look like five years from now. I’m fairly sure it will reward fewer gut decisions, better use of evidence, broader technical participation, and the same old responsibility for people and outcomes.
I’m excited to keep learning what comes next - even when it makes something I used to be good at less important. If you want the tooling details or AI bloopers anecdotes - mostly confident, wrong inferences about what’s important or connected - I’m happy to share.
Note: I discussed these ideas with my colleagues at VideoAmp over the last few months. This blog post would not have been possible without them.