Most analytics teams won't survive the AI shift. Here's what the ones that do are doing now.

There's a line going around that I keep coming back to: AI can't fix bad data. It scales it.

I've spent nearly 20 years building analytics functions inside regulated, data-sensitive industries like healthcare, insurance, banking, the kinds of places where "I trust this number" can't be a hope. It has to be the default, because someone is about to make a real decision on it. And from that vantage point, watching the current rush to bolt AI onto everything, I've become convinced of something uncomfortable:

The AI shift isn't going to reward the teams with the most tools. It's going to reward the teams with the cleanest foundations... and quietly punish everyone who skipped that step.

Here's what I mean, and what the teams pulling ahead are actually doing about it.

1. Governance before acceleration

Most organizations are approaching AI as an accelerator. Point it at the data, go faster, ship more. The problem is that acceleration is neutral. It speeds up whatever you already have. If your metric definitions are contested, if "revenue" means one thing to Finance and another to Product, if half your dashboards quietly disagree with the other half AI doesn't resolve that. It industrializes it. Now you're generating wrong answers at machine speed, with a confident natural-language wrapper that makes them harder to question, not easier.

The teams getting this right did the boring work first. They locked down the single source of truth. They agreed on what the numbers mean before they let anything generate more of them. It's unglamorous, it doesn't demo well, and it's the entire difference between AI that compounds good decisions and AI that compounds bad ones.

Governance used to be the thing you did to stay out of trouble. In the AI era it's the thing that determines whether acceleration helps you or hurts you. That's a real reframe, and the teams that have made it are the ones I'd bet on. It does both now.

2. Analysts become decision partners, not report builders

If your team's value is producing reports, that's an exposed position. If your team's value is owning decisions, it's a durable one.

The distinction matters more than it sounds. A report-builder answers the question they were handed. A decision partner questions the question, reframes it, notices the thing nobody asked about, sits in the room where the call actually gets made and takes responsibility for the recommendation. AI is very good at the first job and, so far, no good at all at the second. It doesn't know your business's politics, its history, or what leadership is actually optimizing for this quarter. AI lacks real-world context.

The teams that will thrive are the ones repositioning their people up the value chain now deliberately, while there's still time to do it on purpose rather than in a panic. That's a leadership job, not a tooling job. It means changing what you hire for, what you reward, and how you talk about the team's worth to the rest of the organization.

3. Literacy is the safety layer

Here's the tension AI creates: it makes powerful analysis available to people who've never been trained to know when analysis is wrong.

This is exactly how you get a confidently stated, completely wrong number in front of an executive, because the person who generated it had no framework for what to trust and what to challenge. In a regulated environment, that's not an embarrassment. That's an incident.

So the teams doing this well are treating data literacy as a safety layer, not a nice-to-have. They're teaching the whole organization not just how to use AI-assisted insight, but how to interrogate it: where the data came from, what it can't tell you, which questions it's answering and which it's only pretending to. I've trained hundreds of people across the functions I've led, and the lesson holds: adoption without literacy doesn't democratize insight. It democratizes the ability to be confidently wrong at scale.

Literacy is what lets you open the door safely. Without it, you either lock the tools away from the people who'd benefit most, or you hand out power with no seatbelts. Neither is the future anyone wants.

Where this goes in the next 18 months

My read: we're about to watch a split open up.

One group of teams will spend the next year-and-a-half in a quiet crisis moving faster than ever, trusted less than ever, generating more output and less confidence, wondering why more AI didn't make them more valuable. The other group will look almost boring by comparison. Governed foundations. Analysts who own outcomes. An organization that knows how to question its own numbers. And they'll pull away, because AI amplified something worth amplifying.

The gap between those two groups is being decided right now, and it's not being decided by who bought which platform. It's being decided by leaders who did the unglamorous work of getting trustworthy before they got fast.

I know which group I want to be building. The window to choose is narrower than it looks.

What are you seeing inside your own organization — is AI making your data more trusted, or less? I'd genuinely like to know if your experience matches mine.