Enterprise IT
From hours of reporting to a 12-second answer
A global IT management platform could see everything about its customers’ estates but could not easily ask it anything. We made the whole estate answerable in plain English, from inside the AI tools teams already use.
- Sector
- Enterprise IT · IT asset management
- Engagement
- AI readiness → conversational data layer → MCP integration
- Built with
- Databricks Genie · Model Context Protocol (MCP) · governed access controls
12s
Answer time, down from 3 to 4 hours of manual reporting
Plain English
Query the whole estate from the AI tools teams already use
The challenge
A global IT management platform sat on an enormous amount of asset and visibility data: every device, licence, contract and configuration across its customers’ estates. The data was rich. Getting answers out of it was not.
The only way in was dashboards, saved reports and exports. Answering a question like “where are we exposed”, “which assets are out of compliance” or “where is spend leaking” meant finding the right report, knowing how to read it, or asking a specialist and waiting. Worse, the data was walled off from the AI assistants their customers’ teams increasingly run their day in, so the most valuable source of truth about the estate was the one place those teams couldn’t simply ask.
The bottleneck was never the data. It was access to it.
What we did
We started with the questions, not the technology
Before building anything, we mapped the questions customers actually ask of their IT estate, working from real usage rather than assumptions, and turned them into a taxonomy of high-value, answerable questions. That decided what was worth building and, just as importantly, what to leave out. Budget went to the answers customers needed most, not to a broad capability nobody would use.
We built a conversational layer over the data
Using Databricks Genie, we turned plain-language questions into governed queries against the asset data. A user can ask in their own words and get an accurate, structured answer back, without writing a query, learning the schema, or knowing where the data lives. The complexity stays under the hood; the user just asks.
We opened it to the tools teams already use
Rather than ship yet another dashboard behind yet another login, we exposed the capability through the Model Context Protocol (MCP), the open standard that lets AI assistants connect to external data and tools. That means the estate becomes answerable from inside the AI assistants customers already work in. The answer comes to where the team already is, not the other way round.
We built governance in from the first commit
Every answer carries its provenance, traceable back to the source data it came from. Access controls mean each user only ever sees what they’re entitled to. The whole path is auditable end to end. The result is something teams can act on with confidence and stand behind in front of an auditor or regulator, not a clever answer with no way to check it.
The outcome
- Customers query their entire IT estate in plain English, straight from their own AI tools, with no dashboards, no exports, and no waiting on a specialist.
- Answers that took 3 to 4 hours of manual reporting now come back in 12 seconds.
Why it worked
Because we started from the questions and not the technology, and wired governance in from the start, the result was something customers trusted enough to use every day, not a demo that impressed once and was quietly forgotten. The data was always there. We made it answerable.
Have data you can see but can’t easily ask?
Start with a fixed-scope AI Opportunity & Risk Audit.