An Engineering Manager Agent
Explorations in Building an EM Agent.
Running a product engineering organization means juggling a lot: opex management, partnering with product and customer success, tracking capex projects, addressing production issues, building strategy, growing engineering leaders, and shaping the culture of the org. The list doesn’t stop, and the decisions you make across all of it matter.
Good decisions need good data. And in my experience, getting that data from the right sources, in the right shape, at the right time is harder than it should be. Jira, GitLab, Slack, Notion, Aha! Every tool has a piece of the picture, and none of them talk to each other the way you need them to.
That challenge got more interesting recently. We’ve been experimenting with a unified engineering model, making developer assignments more fluid across teams so people can move where they’re needed most. It’s a better way to work, but it makes the metrics side of things trickier. Engineers are shifting across backlogs, working in new codebases with new languages, all while we’re experimenting with rapidly evolving agentic development tooling. On top of that, we recently moved to weekly cycles with a mix of kanban and scrum across teams.
I’m a metrics guy. I like to measure the work we do. Not to micromanage, but to understand it and to market the good work my teams are doing. And the traditional tools just weren’t giving me what I needed without a lot of manual stitching.
So I started building my own. Python scripts that pull data from our various platforms, combine it, and do the analysis and reporting my way. And I’m building all of it with AI, because when you can get a custom solution spun up in a fraction of the time, why fight with off-the-shelf dashboards that almost do what you want?
The next step is where it gets really interesting: turning these individual tools into MCPs and building a coordinator agent that orchestrates them. An AI-powered engineering management assistant that can pull the threads together on its own.
Early days, but I’ll share what I learn as it takes shape.