If you're thinking about what's next in AI — I'd focus on solving the friction between data and decisions.
Most organizations have more data than ever. The gap isn't collection. It's what happens after. Data sits in dashboards nobody opens, reports nobody reads, and pipelines nobody trusts — while the decisions that matter get made on gut instinct, in meetings, without a clear source of truth.
The technical work is only half the job. The other half is making sure the right data reaches the right person at the right moment — and that they trust it enough to act.
The gap isn't collection. It's what happens after.
Here are three things I keep coming back to — three areas where I think the biggest leverage is in the next few years.
1. Signal analytics — data that finds you
Instead of building reports people have to remember to look at — build the trigger that tells them when to act, and sends it directly to where they work.
The CS team doesn't need a churn dashboard. They need a nudge at 9am: "This customer hasn't logged in for 7 days — worth a check-in." The marketing team doesn't need a weekly traffic report. They need an alert when organic sessions drop 20% week-over-week so they can investigate before it compounds.
The difference isn't technical. The data is already there. The signal already exists. What's missing is the last mile — routing it to the right person, automatically, at the moment they can still do something about it.
This is proactive analytics. And it's one of the highest-leverage things a data team can build — because it removes the assumption that people know what to look for, and when.
2. Analytics as a product — your analyst, 24/7
Think of BI not as a reporting layer but as a product — a structure people visit regularly, built around the decisions they need to make, with AI embedded to surface what matters.
Most dashboards are built around what data is available, not around what decisions need to be made. They require the user to interpret, compare, and draw conclusions — every single time they look. That's the bottleneck. Not the data. The interpretation.
The next version of BI is a structure: your core KPIs organized consistently, rule-based insights that flag what's healthy and what's not, and an LLM on top that reads the bottom line for you. Like having a senior analyst available 24/7 — working through the same framework, surfacing what matters, before you think to ask.
Less time interpreting numbers. More time acting on them.
This isn't "ask your data." Ask-your-data still puts the burden on the user to know what to ask, and when. The model I'm describing flips that — the system tells you what's going on before you know to look.
I built this for GTM funnel analytics. One place, every metric that matters, written insights on what's working and what's not, AI-powered channel intelligence. The goal was zero interpretation overhead — you open it and you know the bottom line.
3. Governance — the unsexy one that makes everything else work
Before your LLM can tell you the bottom line — everyone in your organization needs to agree on what "active user" means. Governance isn't bureaucracy. It's the foundation.
This is the one nobody wants to talk about. It's slower, less exciting, harder to show in a demo. But it's the reason most AI analytics projects underdeliver.
Your LLM needs clean, consistent definitions to work correctly. If "churn" means one thing to CS, another to Finance, and a third thing to the CEO — your AI analyst is analyzing a different reality than the one your leadership is operating in. The output might be technically correct and practically useless.
Company-wide KPIs. Tracked metric definitions. Clear ownership of what each number means and who's responsible for it. A shared language between AI, BI, and the people making decisions.
This is also what makes the first two pillars actually work. Signal analytics only drives action if everyone agrees on what the signal means. Analytics as a product only builds trust if the numbers are consistent across every view, every team, every meeting.
You can't have an AI analyst working 24/7 if it's analyzing a different version of reality than your CEO.
Where to start
If I were advising an organization on where to focus right now, I'd start with one question: what decisions does your team make every day — and is data part of how they make them?
Map it out by role. What does the CS manager decide on Monday morning? What does the VP of Sales need before the pipeline review? What does the CMO look at before shifting budget?
Then build backwards from those decisions — not forwards from the data you happen to have.
That's the work. And it's not primarily a technical problem. It's a product problem, an alignment problem, and a trust problem. The technical side is just the execution.
The organizations that figure this out — that embed data into how they actually operate, not just how they report — will have a genuine competitive advantage in the next few years. Not because they have more data. Because they act on it faster.