I Stopped Being an Individual Contributor and Started Managing AI Agents

A Solutions Architect's first week with Codex

There's a moment in every technology shift where you stop reading about it and start living it. For me, that moment happened this week. I was lying in bed, verbally describing a macOS app I wanted to build, watching an AI agent spin up a git repo and start writing code while I pulled up Xcode on my laptop to prepare the environment.

I didn't write a single line of code to kick it off. I just spoke.

The Backlog Problem

If you work in pre-sales engineering or solutions architecture, you know the backlog. Not the product backlog. Your backlog. The personal one. The list of demos you've been meaning to build, dashboards you've wanted to refactor, proof-of-concept apps that would absolutely crush it in a customer meeting but never get prioritized because there are only so many hours in a day.

Mine had been growing for months. A real-time audio application that needed a rewrite. A public sector use case dashboard that was functional but rough around the edges. A mobile forensics demo app built on sample data. Each one valuable. Each one sitting in a queue of good intentions.

Then I got access to Codex.

From IC to Manager, Overnight

The shift was immediate and almost disorienting. I kicked off the first task, refactoring the real-time audio app, and while the agent worked through it, I had a thought I'd never had before in this context:

"Okay, it's working on that. What else can I start?"

I opened another thread. Refactor the public sector dashboard. Then another. And another. Within an hour, I had multiple agents running in parallel, each chewing through a project that would have taken me a focused afternoon. Or more realistically, a weekend I'd never actually dedicate.

I wasn't writing code anymore. I was delegating. Reviewing pull requests. Providing direction. Reprioritizing. I was managing a team of tireless, focused agents who didn't need coffee breaks or context-switching recovery time.

The mental model flipped entirely. I went from thinking like an individual contributor ("what's the one thing I can get done today?") to thinking like a manager ("what are the highest-impact things I can get moving simultaneously?").

Building in Bed

The macOS app was the moment it really clicked. I'd been curious about building a native Mac application but hadn't invested the time to explore the toolchain. That evening, I was in bed, phone in hand, and decided to just... try.

I created a git repository. I described what I wanted the app to do, out loud, conversationally. The agent started scaffolding. I got up, downloaded Xcode on my laptop to set up the IDE, and by the time I came back, there was already a foundation to work from.

No tutorial rabbit holes. No Stack Overflow deep dives. No "I'll figure out SwiftUI this weekend." Just a running start on something that had been an idle curiosity an hour earlier.

The Paradigm Shift

I've been in enterprise tech long enough to be skeptical of "paradigm shift" language. But I don't have a better term for what's happening here.

The constraint on my output is no longer my ability to write code. It's my ability to think clearly about what needs to be built and to articulate it well. The bottleneck moved from execution to vision and direction, which, honestly, is where a Solutions Architect's time should have been spent all along.

My backlog isn't gone. But for the first time, it's shrinking. And the projects coming off that list aren't half-finished prototypes. They're functional, reviewable, demonstrable.

This Is Just the Beginning

I'm writing this as a first entry. A timestamp. Because I think we're at the very start of something that fundamentally changes what it means to be a technical individual contributor, or rather, what it means to stop being one and become something new.

I'll keep documenting what I'm building, what's working, what breaks, and how my workflow continues to evolve. But if you're in a technical role and you haven't explored agentic coding tools yet, my honest advice is simple:

Start your backlog. Then hand it off.

Matt is a Solutions Architect working in the federal technology space. This is his first post in what he hopes will be a running series on AI-augmented technical work.