There's a quiet anxiety running through a lot of data teams right now. Every week brings a new demo of an AI that writes SQL, generates pipelines, or documents a codebase in seconds. For an engineer whose craft is writing SQL and building pipelines, the unspoken question is obvious: am I about to be automated away?
As a lead, you don't get to ignore that question. How you answer it — in words and, more importantly, in what you actually do — shapes whether your team leans into this shift or freezes up. Here's how I approach it.
Name the shift honestly
Pretending nothing is changing insults smart people. So does doom-mongering. The honest framing is this: AI is changing what data engineers do, not whether they're needed. The repetitive middle of the job — the boilerplate pipeline, the first draft of documentation, the obvious transformation — is increasingly something a machine can start. The parts that were always the real work — judgement, architecture, correctness, understanding the business — matter more, not less.
I say this out loud, early, and often. Clarity is a leadership deliverable.
Move the team up the value chain
If AI can generate the first draft, the engineer's job shifts toward reviewing, deciding, and owning outcomes. That's a promotion in disguise, but only if you actively help people make the move. In practice that means:
- Rewarding judgement over output. The engineer who catches that an AI-generated pipeline has a subtle correctness bug is doing the most valuable work on the team. Celebrate that, loudly.
- Raising the altitude. Push people from "build this pipeline" toward "own this domain" — its quality, its architecture, its roadmap. Agents can draft; humans own.
- Teaching the new tools as tools. An engineer who's fluent with agents and copilots ships more and thinks bigger. This is a skill to invest in, not a threat to defend against.
Keep the guardrails human
Here's the part I'm firm on: as we introduce AI into our own platform — agents that fix data quality, generate docs, scaffold pipelines — the human stays in the loop on anything that matters. Not because the AI is dumb, but because accountability doesn't automate. When something touches production, a named person reviewed it. That principle protects the business and, not incidentally, keeps the team's expertise central and valued.
Protect morale by protecting meaning
People don't burn out from hard work; they burn out from work that feels pointless or precarious. The fastest way to demoralise a data team in 2026 is to make them feel like caretakers of a system that's about to replace them. The antidote is to connect their work to outcomes they can see, and to be visibly investing in them — their skills, their scope, their growth — not just the tooling.
The lead's real job hasn't changed
Strip away the AI and the fundamentals are exactly what they always were: give people clarity, grow them deliberately, and point their energy at work that matters. The tools are new and genuinely exciting. The job of leading humans through change is as old as work itself — and it's still done with honesty, mentorship, and a bias for shipping things that last.