AI Adoption GuideConstructionHandover
Digital Twin Asset Data Population
Agentic system maps as-built BIM attributes and O&M data into the client digital twin platform for FM use. (e.g., Autodesk Tandem)
Construction processBidAwardPlanMobilizeBuildInspectHandoverClose
By Don, DoneThat’s AI coach · updated
What the twin is allowed to hold at handover
A mapped asset is a record the handover lead can defend. Every filled attribute points to an as-built BIM object or an O&M submission the lead still owns. Blank attributes stay blank. The twin is not a scratchpad for values remembered from site, copied from a design issue, or guessed from a catalogue.
The client may load that payload into a digital twin platform in the Autodesk Tandem class. The same mapping rule holds if they named another twin. The platform stores the record. It does not prove a value is true. The cite does.
As-built packages, O&M PDFs, and transmittals usually sit in a project system in the Procore class. Mapping those files into the twin does not replace them. It points at the revision the handover lead still holds.
Do not treat a populated twin row as FM-accepted. Facilities still have to confirm that the record matches what they will operate. Until that sign-off, the quality outcome is mapped with cites, not ready for the helpdesk.
Use the as-built model, not the design model
Copy attributes only from the as-built model the contract named for handover. Design BIM is a coordination model. Clash-checked geometry from design stage is not as-built evidence. If the live file still carries design-phase parameters, stop. Name the as-built source: file, revision, and discipline model.
BIM design clash detection belongs earlier in the job. It does not prove that a pump, damper, or panel was installed as drawn. Using a clash-clean design object as the twin source is the first failure mode. The twin looks complete, while serial, location, and maintainable identity are still design fiction.
When the as-built object exists, map only attributes that are actually on that object. Type mark, system, level, and unique ID are usable when those parameters are populated in the as-built file. If the as-built parameter is empty, the twin field stays empty. Do not backfill from the design model because the objects look the same.
If the as-built model has no object for an installed asset, do not create a twin asset to fill the gap. Empty stays empty. Raise the missing object as a BIM close-out item. Inventing an asset so the twin looks populated is a quality failure, even if FM later finds the kit in the plant room.
Map O&M only from submissions that match the asset
O&M fields (manufacturer, model, duty, media, spares, maintenance task) come from O&M submissions, not from BIM guesswork. Map them after the pack is searchable. O and M manual auto-assembly gathers those manuals. This page only maps what that pack, plus any structured O&M schedule, actually contains.
Match an O&M document to a BIM object with evidence both sides share: tag, type mark, system, or a unique ID written on the drawing and in the manual. If the match is ambiguous, leave the O&M fields blank on the twin record and flag the collision for the handover lead. Do not pick the most likely PDF.
Warranty dates and coverage belong on the twin only when they come from a cited warranty document or register line. warranty register extraction is the source for those fields. Do not copy a warranty period from a generic manufacturer brochure if the subcontract submission is silent.
The second failure mode is inventing a serial number. Serials come from a nameplate photo, a commissioning sheet, or an as-built parameter that already holds the number. If none of those exist, the serial field stays empty. A generated or typical serial makes the twin look complete and poisons spare parts, recall notices, and warranty claims.
The third failure mode is declaring the twin FM-ready while O&M is missing. A mapped geometry record with empty manufacturer, empty media, and empty maintenance task is not a maintainable asset. Close the O&M gap, or hand FM a mapped record that still shows those fields blank. Do not fill them from a similar unit on another floor.
Put a cite on every filled field
Each populated twin attribute carries a cite the handover lead can open. For BIM-sourced fields, cite the as-built file name, revision, object unique ID, and parameter name. For O&M-sourced fields, cite the document number or filename, revision, and the page or schedule row. If two sources disagree, do not average them. Keep the field empty, or keep the conflict visible, until the lead chooses.
Cites belong on the mapped record, not in a side spreadsheet the twin will drift away from. When FM asks where a duty came from, the answer is the cited O&M revision, not a mapping log nobody kept.
Run a completeness pass on the handover pack before treating mapping as finished. handover package completeness check tells you whether as-built models and O&M submissions are even present. Mapping a thin pack produces a thin twin. That is honest. Padding it is not.
One rooftop air-handling unit, mapped without invention
Take a rooftop air-handling unit that exists in the as-built mechanical model. The object has a type mark, a system name, a level, and a unique ID. The serial-number parameter is empty. The O&M folder has a manufacturer PDF for that type mark, with model, filter sizes, and a recommended service interval. There is no nameplate photo and no commissioning sheet with a serial.
Map type mark, system, level, and unique ID from the as-built object, each cited to that object's ID and the as-built file revision. Map manufacturer, model, filter sizes, and service interval from the O&M PDF, cited to filename, revision, and page. Leave serial number empty. Do not copy a serial from an identical unit on the adjacent roof. Do not copy a serial from the design model's type catalogue.
If the design model still shows a different type mark for the same grid location, ignore the design object. The as-built object is the source. Note the mismatch for BIM close-out. Do not reconcile by writing both type marks into the twin.
The handover lead then confirms three things: the BIM cites open the right as-built object, the O&M cites open the right manual revision, and every blank field is blank on purpose. Only after that confirmation does the mapped record leave the mapping queue. It still is not FM-accepted until the client signs that they will operate from it.
Handover confirms the map; FM still has to sign
The handover lead owns mapping quality. Review a sample of records, including blanks and conflicts, not only the rows that look complete. Confirm that no design-BIM attributes were used as as-built, that no serial or asset was invented, and that missing O&M was not treated as FM-ready.
Send FM the mapped extract with cites visible. Ask them to accept or reject records. Do not assume acceptance because the twin platform ingested a file. A rejected record goes back to the source: missing as-built parameter, missing O&M page, or a wrong match. Fix the source, remap, and cite again. Do not edit the twin by hand in a way that drops the cite.
If the client twin platform in the Autodesk Tandem class requires a value for a field you cannot cite, leave it empty in the payload or use the platform's explicit empty value, and record the exception. Do not satisfy a required-field prompt with a placeholder serial, a dummy identifier, or TBC dressed up as data.
When mapping is done, the deliverable is a set of mapped assets, each with cites to as-built BIM and O&M submissions the handover lead still owns. That is the quality bar. FM use starts after they sign, not when the import job reports success.
Is this worth automating for you?
Whether this pays back depends on how much time it takes your team today. Most teams estimate that from memory, and the estimate is usually wrong in one direction or the other. This one is rated high effort to implement, so the baseline matters more than usual.
DoneThat reconstructs where the time actually went, with no timers to forget, so you can measure the baseline before committing to a project and check the gain afterward.
Measure the baseline first