Migrating configuration data off a legacy ITSM tool means mapping hundreds of fields to ServiceNow’s multi-layered CMDB schema by hand. It’s slow, it’s error-prone, and it depends on the kind of SME expertise that’s always in short supply during a migration.
The team built an AI-powered field-mapping accelerator to take that manual work off the table. It reads a legacy application’s fields, automatically maps them to the correct ServiceNow CMDB tables and attributes, and scores its own confidence on every single mapping before a person ever reviews it.
How it works
A CMDB Mapping Assistant, built in ServiceNow AI Agent Studio, takes a legacy field catalog and works it through an import job: analyzing the source catalog, suggesting a target CMDB class for each legacy object, then mapping individual fields to their ServiceNow equivalents. Every suggested class comes with a confidence score and, where the match is weak, a plain note like “low-confidence guess, review before using.” Every field mapping gets its own confidence tier, high, medium, or low, along with alternate suggestions the agent considered and didn’t pick.
Nothing merges into the live CMDB without that scoring attached. A field mapped at high confidence, like Operational Status to operational_status, moves through automatically. A field the agent is less sure about gets flagged for review with its reasoning attached, so a person is deciding on genuinely ambiguous cases instead of re-checking work the agent already got right.
Why it matters
The alternative to this is a spreadsheet and an SME’s calendar. Mapping hundreds of fields by hand for one legacy platform is tedious enough; doing it again for every additional source system a company happens to be consolidating is where migration timelines actually blow up. Automating that first pass is what turns weeks of manual mapping into something closer to minutes, freeing up scarce SME time for the mappings that genuinely need a human judgment call instead of the ones that don’t.
Because the logic isn’t hardcoded to one platform, the same accelerator extends across every source system a migration needs to touch, without rebuilding the mapping logic each time. That’s also what makes the output trustworthy at scale: a confidence score on every field means data quality and CMDB consistency improve as a byproduct of the process, not as a separate cleanup phase bolted on afterward.
What the pilot showed
Tested across four enterprise ITSM platforms, Cherwell, Ivanti, BMC, and OpenText, the accelerator processed 909 fields with a 100 percent auto mapping rate, meaning every field received an automated mapping suggestion rather than being left blank for a person to fill in from scratch. Confidence varied by platform, as expected given how differently each tool structures its data: high-confidence mappings needing no review ranged from 19 percent on Cherwell’s field set up to 37 percent on OpenText’s, averaging around 29 percent across all four, with the rest routed to medium or low confidence for a reviewer to confirm.
In one completed import job, a field-level result included the full reasoning trail: a suggested target field, its confidence score, and every alternate the agent considered along with their own confidence levels, giving a reviewer the “why” behind a suggestion rather than just the suggestion itself.
AI-Powered CMDB Field Mapping Accelerator is one way to bring automated, confidence-scored mapping to a CMDB migration. If yours could use the same solution, reach out to KeenStack.
Built by
Debabrata Sarkar
Arun Kailash
Arjun Bangera