For decades, business intelligence has operated on a simple premise: someone anticipates the questions, builds the reports, and everyone else consumes what’s been pre-built. The analyst or developer acts as an intermediary. The business user learns to think within the boundaries of what’s already been constructed.
That model worked. It still works for a specific category of need. But something fundamental has shifted, and most organizations haven’t fully reckoned with what it means.
The Question You Didn't Know to Ask
The most valuable analytical moments in any organization rarely start with a known question. They start with curiosity. A CFO notices something unexpected in the numbers. A VP of Sales sees a pattern that doesn’t fit the narrative. A COO has a hunch about why costs are climbing.
In the traditional BI world, that curiosity hits a wall almost immediately. You can click a pre-built drill path. You can request a new report and wait. You can export to Excel and start wrestling with the data manually. Each of these options interrupts the cognitive flow at exactly the moment it matters most.
With MCP (Model Context Protocol) connections to live data sources and a conversational interface powered by an LLM, that wall disappears. Not because technology is flashier, but because the interaction model is fundamentally different. The person asking the question holds the analytical initiative, not the person who built the dashboard.
Walking Through Data
Here’s what this looks like in practice.
You sit down and say: “Show me revenue by product line.”
The system queries the relevant source, returns the breakdown. You see something interesting in the numbers and follow the thread.
“Break that down by customer segment for the top three product lines.”
Now you’re looking at a more granular picture. One segment catches your eye.
“What’s the trend on that middle segment over the past eight quarters?”
A pattern emerges. You want to understand what’s driving it.
“Pull in the customer acquisition costs for that same period.”
And then the real question forms:
“Is there a correlation with marketing spend by channel?”
Five questions. Three different data sources. Zero pre-built reports. And the entire thread took less time than it would have taken to submit a request to your BI team.
This is what I mean by walking through data. Each step builds on the last. The context carries forward. You’re thinking, not operating software. The conversation is fluid in a way that clicking through dashboards has never been, because the interface adapts to your inquiry rather than constraining it.
Nobody anticipated this specific sequence of questions when the data environment was designed. Nobody needed to. The AI navigated across sources, maintained context, and followed the analytical thread wherever it led.
Critically, MCP aligns with modern governance standards. It runs as a local or managed service under enterprise control, allowing security teams to enforce auditability, encryption, and lineage tracking. It integrates with Metadata catalogs and access policies rather than bypassing them through opaque vendor connectors.
Connectors integrate. MCP orchestrates. The former accelerates deployment. The latter future-proofs architecture. For executives steering AI adoption at scale, that distinction determines whether today’s proof of concept becomes tomorrow’s constraint, or tomorrow’s foundation for sustainable, model-agnostic intelligence.
Who Holds the Initiative?
This isn’t just a convenience upgrade. It represents a shift in who drives analysis.
In the traditional model, the BI team or data engineers decide what questions are answerable. They build semantic layers, design data models, and create visualizations. The business user operates within those boundaries. If your question doesn’t fit what’s been built, you wait, or you never ask it at all.
In the conversational model, the business user leads. They follow their own thread of inquiry across whatever sources are connected. The AI serves as a navigator that understands both the question being asked and where the relevant data lives. The analytical initiative moves from IT to the business, not because IT failed, but because technology no longer requires that intermediary step.
This is a bigger deal than it might sound. How many strategic insights have been lost over the years because the question didn’t fit an existing report? How many threads of inquiry died because following them meant a two-week turnaround on a new dashboard?
The Bifurcation: What BI Still Owns and What It Doesn't
Now, here’s where I expect some pushback, and I welcome it.
Traditional BI tools are not going away. They still own operational reporting, and they should. The scheduled report that hits inboxes at 6 AM, the pixel-perfect financial statement for regulatory submission, the factory floor dashboard refreshing every 15 minutes, the embedded analytics inside transactional applications: these are production workloads. They need to be reliable, auditable, and consistent. Enterprise BI vendors have spent decades engineering exactly that, and those capabilities remain valuable.
But strategic and exploratory analysis? That’s a different matter entirely.
When a CFO is evaluating whether to enter a new market, they don’t need a pre-built dashboard. When a COO is diagnosing why fulfillment costs are rising faster than volume, they don’t need to wait for someone to build a new report. They need to think through the data, pulling from multiple systems, often in ways nobody anticipated when the BI environment was designed.
Traditional BI still owns operational reporting. But it no longer owns strategic conversation.
That’s the bifurcation. And once you see it, you start to question the massive investments organizations continue to make in BI tool infrastructure for use cases that a conversational interface handles more naturally, more quickly, and with fewer intermediaries.
Implications for the Future
I’m not arguing that organizations should rip out their BI platforms tomorrow. Operational reporting is real, and it needs real infrastructure.
What I am arguing is that the reflex to route every analytical question through the traditional BI stack is becoming harder to justify. Exploratory and strategic analysis, which genuinely influences business decisions, now has a more suitable place. Companies that quickly adapt to this division will progress faster than those still relying on traditional dashboard development.
The data still needs to be well-architected. Governance still matters. Garbage-in still produces garbage-out. But the idea that you need a heavyweight BI tool to stand between a business leader and the answers they’re looking for? That wall is coming down.
