Every data team I've talked to has the same problem: they spend more time building dashboards than answering questions.
A stakeholder asks: "Which accounts are most at risk of churning?"
The data team says: "We can build a dashboard for that. It'll be ready in two weeks."
Two weeks later, the dashboard is live. It answers the original question. And then the stakeholder asks a slightly different question, and the cycle starts again.
Why dashboards are the wrong abstraction
Dashboards are answers to questions you predicted in advance. They're good for monitoring known metrics. They're terrible for exploration, ad-hoc analysis, or any question that wasn't anticipated when the dashboard was built.
The world doesn't ask predictable questions. Business moves faster than dashboard backlogs.
The alternative: queryable data
What if instead of building a dashboard, you made your database directly queryable in natural language?
Not "give everyone raw SQL access" — that's a different kind of problem. But a controlled layer where someone can ask:
"Show me all customers who upgraded in the last 30 days but haven't used the new feature yet."
And get an answer from live data, in seconds, without a ticket.
This is what MCP (Model Context Protocol) enables when it's connected to your database. The AI constructs the query, validates it, and returns the result — without the stakeholder needing to know SQL, and without the data team building yet another dashboard.
It's not about replacing analysts
Data analysts are still valuable — for modeling, for data quality, for building the infrastructure that makes this possible. But routing every ad-hoc business question through a two-week dashboard build is a waste of everyone's time.
If you're curious what this looks like in practice, I wrote about it here: Kill the data request ticket.
The dashboard era isn't over. But for ad-hoc questions? There's a better way.
Top comments (0)