Use AI to flag conflicting HVAC equipment identifiers

Keep equipment identity distinct from customer matching

A customer and property can be correct while the equipment reference is wrong. Gather an authorized sample of asset labels, service-note references and approved equipment registers. Preserve exact text, including leading zeros, alongside property ID, source date and record reference. Ask AI to flag conflicting or incomplete identifiers, not to produce a new master register.

Avoid putting access codes, unnecessary tenant details or full customer documents into the sample. A short evidence extract is usually sufficient to show why two labels conflict. Retain the original record outside the assistant.

Build an exception table instead of a repair script

Use columns for property, observed label, source, candidate matching record, conflict type, reviewer and next evidence needed. In a fictional example, the asset register lists RTU-03 while a recent note says RTU-08 at the same building. Similar descriptions do not prove they are the same unit. Mark the reference unresolved and request verification through the field or asset-data owner.

Treat OCR from a photographed label as a candidate transcription. A blurred serial digit must remain uncertain. Do not infer equipment age, capacity, refrigerant, compatibility or service requirements from a guessed model number. Those decisions need reliable records and qualified staff.

Close the exception with a documented decision

The responsible person checks the underlying evidence and authorizes any record correction. Record the old value, approved value, reason and supporting reference under existing data procedures. The assistant may prepare a proposed correction list but should have no permission to merge, rename or delete assets.

Test with a small known sample containing two genuinely different units with similar labels. Review whether the assistant preserves both records and leaves uncertain pairs open. Count reviewer-accepted resolutions and false matches separately; a high number of proposed matches is not evidence that the equipment inventory improved.

Review before use

• Preserve exact source labels and property IDs.
• Leave unreadable digits unresolved.
• Never merge similar-looking units automatically.
• Document the reviewer and evidence for every correction.

Source context and limits

Jobber documents typed custom fields associated with records, with account configuration and object values kept separate. This exception-review method does not assert that Jobber provides an equipment-matching engine.