A while back, another CEO gave me one of those pieces of advice that doesn't land at first and then won't leave you alone.
He said: "Your job as a CEO is to make yourself redundant for your company. And you'll fail at it every single day."
His point was simple. Everything I personally do is, by definition, a bottleneck. The company depends on me to do it, which means it can't move faster than me on that thing. So my real job is to give it away: to empower other people, to build systems, to make sure the company doesn't need me for as much as possible.
But why is this doomed to fail? Because every time I succeed in offloading something, I create capacity in myself to take on something I never had time for before. The bottleneck just moves. I become more valuable, not less, because my output keeps growing. I will never actually finish making myself redundant. The list of things only I can do gets shorter on one end and longer on the other, forever.
I keep thinking about this advice because I think it's a healthy way to talk about what AI is doing to data teams right now.
The discomfort is real. So is the opportunity.
There's an uncomfortable feeling in the air for a lot of data people I talk to. The sense that the ground is shifting under you faster than you can keep up. I'm not going to pretend that feeling is wrong.
An inspiring Israeli president once said: "Optimists and pessimists die the same way. They just live differently". The change is real, and it's inevitable. The pessimists obsess about "will AI replace me?", but the optimists should focus on the more useful question, the same one that founder asked me: What will you do with the capacity you get back?
Because if you're a data person reading this, I'd bet two things are true about you right now.
One: you're underwater. You have too much on your plate. I meet teams who tell me about projects that span one or two years. Not because the projects are that hard, but because there are only so many hours in a sprint and only so many sprints in a quarter. What if those projects could ship in a quarter? What if your two-week sprint could be done in a day? What if the tedious manual work, the broken pipeline at 2am, the documentation nobody has time to write, the test coverage you keep meaning to get to, was just... handled?
Two: there's a version of your job you've never gotten to do. There are things you've wanted to build for the business but written off as impossible, not because they were technically hard but because you didn't have the people. You could build a beautiful finance data model. But you couldn't assign a financial analyst to every single employee making budget decisions. You could build great dashboards. But you couldn't sit next to every PM and help them interpret the numbers in their specific context. There was always a ceiling on how far your work could reach, and that ceiling was headcount.
That ceiling is coming down. And the backlog? That's shrinking too.
Human intelligence used to be the bottleneck for a huge category of data work. It's not anymore. At least not for the repetitive, well-scoped, executional parts. Where human intelligence is still very much the bottleneck is in designing intelligent systems: figuring out what to build, deciding how the pieces fit, encoding the business logic and the guardrails, choosing what's worth automating and what isn't.
That, to me, is the actual job description for data people going forward. Designing the intelligent systems that leverage data, and building the data platform that powers them. It's more interesting work. It's more ambitious work. It's more impactful work.
The same president also said: “We should use our imagination more than our memory”. This is what’s required of data people today. Imagine what can be done, forget that it was impossible until recently.
What we're building at Elementary
We're going all in on one idea: the experience of having extra data engineers on your team.
Not "an AI that helps you." A team. Agents that take ownership of tasks, do the work, and bring it back to you for review. You stop being the person doing every manual task and start being the person managing a team that does them for you. You go from individual contributor to a tech lead overnight, except your team never sleeps and never gets bored doing repetitive work.
The point isn't to remove you from the loop. It's to put you in a different position in the loop. You set the standards. You review the work. You decide what's worth building. The agents handle the parts of the job that were burning your hours without growing your skills.
If the last decade of data tools was about giving you better instruments to do the work yourself, this future is about giving you a team to do the work for you.
And it doesn't stop with data people
Here's something we didn't fully expect, but is now one of the most exciting parts of what's happening:
The biggest adoption of our agents isn't coming from data engineers. It's coming from business users.
People who have always had questions about the data, always had ideas for what they wanted to track, always had small changes they needed made but were blocked by not knowing SQL, not knowing to code, not knowing the infra. So they'd file a ticket. Wait two weeks. Get something close to what they asked for. File another ticket.
Those people can now complete end-to-end data tasks on their own. Not because they suddenly learned to code, but because they don't need to. They can describe what they want and an agent can do it within the standards and guardrails the data team has set.
This is the thing self-service data promised and never really delivered. Not because the idea was wrong, but because the interface was always still some version of "here's a tool, now learn it." The new interface is just: tell someone what you need.
For data teams, this changes the relationship completely. You stop being a service desk. You start being the people who design the system that lets the whole company work with data directly. You become more leveraged, not less.
So, back to the original advice
Make yourself redundant. You'll fail at it every day.
The version of your job that AI can do, that's the version you should want to give up. Not because your job is going away, but because the more of that you offload, the more space you have for the work that actually requires you. The strategic calls. The system design. The judgment about what's worth building. The relationships with the business.
That's the future we're building Elementary for. Not a smaller role for data people. A bigger one.
