AI isn’t just changing how engineers code—it’s transforming how we hire, lead, and build resilient systems. In this Q&A with Chime VP of Engineering, Michael Barrett, we dig into what’s shifting, what’s still unsolved, and how to balance “painting” and “plumbing” to create durable impact in the age of AI.
Q: How has your role changed in the last two years?
Two years ago, I didn’t run weekly AI Build Hours across engineering and partner teams. Now it’s a ritual: every session, we ship a real workflow improvement. It’s made AI how we work—not a side project.
Q: How is hiring evolving in the AI era?
We definitely don’t have this all figured out—and honestly, I don’t think anyone does yet. The pace at which AI is developing and the way new tools come up every day consistently make it feel like you're behind. But at Chime, we’re leaning in, trying new approaches, and learning as we go.
Filtering out fake applications feels like the easy part. To me, the trickier, more important piece is understanding how someone actually works with AI. That’s where we’ve been experimenting: putting candidates in real scenarios like debugging a messy repo with an AI assistant, adapting when constraints change, or showing before-and-after examples of how they’ve used AI in their past work.
We’re not just looking for the “right” answer—we’re looking for judgment, adaptability, and the discipline to verify AI output. What we’ve seen is that candidates who do well in these trials ramp faster and thrive on AI-native teams.
It’s still a work in progress, but that’s the point—we’re building and refining our playbook as AI evolves, and our willingness to experiment is helping us move forward.
Q: What does successful AI leadership look like—not just tool adoption, but true transformation?
I tend to think of it as something I’ve been calling the “Painting & Plumbing” problem:
Painting is the fun part: writing code, experimenting, creating new things. It’s expressive, visible, and rewarding.
Plumbing is the less immediately glamorous stuff: tests, docs, bug fixes, incidents. Nobody brags about it (though we should!), but when it breaks, it’s all anyone cares about, and do you want experts focused on it!
Right now, I think the bulk of what we hear about is AI turbocharging the painting. We hear about people shooting for 10x coding gains. But if we’re generating 10x more code, and we’re ignoring the plumbing? I worry we’ll risk generating 10x the bugs, troubleshooting, and incidents.
The opportunity here is to use AI for both. Imagine AI that writes tests as you code, helps engineers troubleshoot issues, provides helpful guidance before and after incidents, helps you figure out who is the expert on a topic you’re looking into, or suggests fixes from past bugs. That’s how we transform: not just making engineers faster painters but smarter plumbers, too.
To me, AI leadership means investing in both painting and plumbing, so engineers can do more of the creative work they love while our systems stay resilient.
Q: How do you model the right AI behavior?
I try to model the right AI behavior by showing my own work—prompts, checks, and when I don’t use AI.
We set clear guardrails with green/yellow/red tasks and require explainability before deployment.
We also train both senior and junior engineers to verify rigorously so speed and safety scale together.
Q: Fast forward 2 years. What skill or capability will separate great engineering leaders from average ones that barely matters today?
The differentiator will be systems literacy: the ability to design and lead socio-technical systems where humans, models, data, and policy co-evolve. At the end of the day, it’s all about systems, whether they are computer, human, or organizational.
Great leaders will see where those systems are breaking down, clear the friction, and provide clarity. AI is just another tool in that mix—one with big benefits but also added complexity—and the leaders who can integrate it responsibly will create durable advantage, not just ship features.
