A conversation about CMDB accuracy, visibility and where AI like Claude can help.
If there's one question I'd like to ask you today, it's this: when something breaks, how confident are you in the answer to "what else does this affect?"
If the honest answer is "not very," you're in good company. A lot of IT leaders tell me some version of the same thing: they don't fully trust their CMDB. And that matters, because the CMDB is supposed to be how you see your IT environment, what you own, how it's connected and who's responsible for it. When the picture is off, change decisions, incident response, audits and every AI initiative you're planning inherit the error.
The health score isn't the whole story
Your CMDB Health dashboard is a good place to start. It scores completeness, correctness and compliance, and it shows you where the problem records are. But those scores tell you whether records follow your rules. They don't tell you whether a record matches reality. A server can pass every check and have been decommissioned last spring.
Catching that takes judgment: comparing sources, noticing what doesn't fit and understanding why. That's where a model like Claude can help.
What Claude can do for your CMDB
Think of it as a skeptic that checks the record against everything else you know. In practice, it can:
- Reconcile your sources. Compare the CMDB with your cloud, endpoint, network and asset data, and explain each difference, not just flag it.
- Catch what rules miss. Spot things like a database named for production but classified as a lab device, or an owner group that no longer exists.
- Check your service maps. Use incident and change history to suggest dependencies that are missing, for your team to confirm.
- Help you prioritize. Group thousands of failures by likely root cause, so you fix the source instead of chasing symptoms.
To be clear about what it doesn't do: Claude recommends, and your people decide. Your discovery tools and change processes stay the source of truth, and nothing gets written to the CMDB without the same approvals you'd require of anyone else. I'd also be careful of anyone promising a self-healing CMDB.
Where I'd start with a customer
I wouldn't try to boil the ocean. With a customer, I'd pick one business-critical service, ideally one with a recent major incident behind it, and run a focused validation:
- Pull the CMDB records for that service, along with the related discovery, cloud and incident data.
- Use Claude to reconcile them and list the discrepancies, with evidence.
- Have the service owner review the findings.
- Look together at what was fixed and what it would take to scale.
The customer gets a concrete, reviewable result and a clear view of what's driving the gaps, without a big commitment up front. Our ServiceNow specialists can help scope it.