Foundation-First AI
Everyone’s asking what AI will do for them. The better question: what will it do to what’s already broken?
The organizations that struggle with AI usually know they have a data problem. The dangerous ones are the organizations that don’t.
The organizations that struggle with AI usually know they have a data problem. The dangerous ones are the organizations that don’t. When a system is visibly broken, people work around it — informal knowledge, the right phone call, the record everyone skips. Those workarounds function well enough that nobody fixes what’s underneath. Then AI arrives and reads the official record. The one everyone quietly stopped trusting years ago. That’s not an AI failure. It’s an audit you didn’t know you needed.
. : Find / the records nobody trusted
Engineers at this organization were burning hours per incident not diagnosing problems, but locating them. Hunting across systems for answers that should have been one lookup away. The CMDB had the answers, technically. But it had drifted. Stale in places, inconsistent in others, quietly distrusted by everyone who used it. Not broken enough to flag. Just broken enough to route around. Nobody had fixed it because the workarounds worked. Until AI arrived and the workarounds disappeared.
. : Unite / the real decision is not technical
Before any tool was selected, the harder conversation had to happen. Who owns the record, who attests its accuracy, and who is accountable when it’s wrong. That conversation had been deferred for years because the workarounds made it optional. We made it unavoidable. We brought IT Service Management, Operations, and the portfolio organization together not for a data cleanup, but to establish decision rights. The data was a symptom. Ownership was the diagnosis.
. : Shape / making the record worth reading
We rebuilt the service taxonomy, reconnected services to their owners, and established consistent relationship standards across the environment. The test wasn’t technical. It was human. Could an engineer who had never touched this system read the CMDB and immediately know what a service does, what it depends on, and who picks up the phone when it fails? When the answer was consistently yes, the workaround economy was over. The record had stopped lying.
. : Execute / the multiplier pointed in the right direction
With the foundation in place, we integrated Microsoft Copilot into the triage workflow. Surfacing ownership, dependencies, and resolution context at the moment an incident opened. No hunting. No calling the one person who knows. Within three months MTTR dropped 65%, driven not by the AI being exceptional, but by it finally reading the truth. The multiplier was always there. It just needed something worth multiplying.
Before you ask what AI will do for you take the audit. The organizations that win with AI aren’t the ones who moved fastest. They’re the ones who knew what they were multiplying before they hit go.
Let’s find out what yours is multiplying.