What's left for analysts when agents do everything?

The question isn't hypothetical anymore. Analysts are watching AI handle work that used to take hours in seconds. Queries they would have written, explorations they would have run, analysis they would have compiled — all automated.

So what's the job now?

If the anxiety feels real, that's because the stakes are real. To the uninitiated, AI looks like an analyst in a box, but we know the technology isn't perfect — agents still hallucinate, miss nuance, and need correction. But the speed at which they're improving is remarkable. The gap between "helpful tool" and "fundamental shift" is closing fast.

Here's what got lost: the nuance of analysts' work was always obfuscated by SQL. Code writing was seen as "the hard part," when in reality it was the price of admission.

The role isn't disappearing. It's finally becoming what it should have been all along.

Analysts can focus on the human parts — connecting dots that don't live in a database and driving action from insights — that's the future.

We curate exploration

Analysts have always been curators. We documented metrics, added dashboard notes about outages and launches, left breadcrumbs about data quality issues. The problem? No one prioritized it. Except us.

This left work to “clean things up” as back burner. Growing tech debt that every data team looks at with a side eye and heavy sigh. We know it’s important, but it never felt like high ROI work because the human in the middle was there to curate and translate the mess into beautiful dashboards.

Agents change that. The context you were already creating — how metrics are calculated, where data gets messy, what caveats matter — finally gets used. The agent reads your documentation and applies it. The curation work you were doing suddenly has leverage.

In practice: if you're embedded in product, you're curating the agent's understanding of product metrics. If you own finance reporting, you're ensuring the agent knows what "revenue" actually means in your business.

Each analyst manages context for their domain and oversees an agent trained on that context. The agent handles volume. You handle depth, exceptions, and questions that need human reasoning.

Instead of answering the same question five times, you teach the agent once. And unlike a dashboard note, the agent actually pays attention.

We tell stories, not just build dashboards

Here's the part that separates analysts from people who happen to know SQL: turning data into decisions.

AI might be great at analyzing data, but here's what it can't do: make people care. Research in Made to Stick found that we remember stories far better than we remember raw statistics. So, let AI crunch the numbers — but if you want anyone to actually act on your insights, you'd better tell a damn good story.

This was the why of the job. But it got deprioritized, squeezed out by the endless queue of ad hoc requests. Storytelling only surfaced during big fire drills — the board deck, the post-mortem, the "explain what happened" moment after something broke. The rest of the time? You were too busy pulling numbers.

When agents can handle the volume, storytelling stops being the thing you do when there's a crisis and becomes the thing you do every day. But that story only works if you understand the business deeply enough to know which narrative will land.

An agent can help you assemble that story. But it can't tell you if it's the right one. We're the ones who know the room we're walking into, the unstated concerns, the political context, the decision that hinges on how we frame what we found.

These were always the skills that mattered. We just buried them under layers of query-writing because that's what the job demanded. Now we can resurface them.

The agents work for you

I think the fear comes from feeling disconnected. We love this job because we love being in the weeds, solving problems, and seeing solutions come together. When agents handle more of that, it's uncomfortable.

But agents aren't taking the job away. They're taking the tedious parts away.

The parts where you're copying SQL from last month and swapping dates. The parts where you're rebuilding the same dashboard for the third team. The parts where you're answering "how many X happened last week?" for the hundredth time.

The analyst role isn't disappearing. It's shedding the parts that have become commoditized.

The job is shifting into something different. Better, I'd argue.