Industry field guide
Can plumbing companies license job records or first-person training footage?
An anonymized service ticket and a first-person camera recording teach different things.
- Sector
- Construction
- 2022 NAICS
- 238220
- Practice pack
- Example and answer key (JSON)
- Updated
What this guide tests
Does the packet establish P1's final visible result? Identify the usable evidence and the gap.
See the specimen and fields · Work through the fictional records
Task and source context
An anonymized service ticket and a first-person camera recording teach different things. Scale AI publicly lists plumbing among field services of interest for structured, de-identified operating data. That is a path for companies with established records and authority to share them. micro1 separately advertises paid capture of approved trade recordings and shows plumbing among its examples. Meta’s first-person video research illustrates what models can learn from hands, objects and actions. A contributor should verify the actual project and permissions before recording. Scale AI data partnership · Ego-Exo4D research · micro1 trade recording
For a safe first-person pilot, use an isolated training fixture in your shop. Show two or three trap assemblies and one staged failure such as an incorrectly seated washer. Record a clean start state, the part selection, the sequence of hand movements, the verification step and the result. Tag fixture type, parts, action start and end, error type and outcome. An evaluation set could ask a model to identify the next step or detect the failure before a leak test.
For a records pilot, start with job category, symptoms, diagnosis, work completed and outcome. Recreate a ticket example with fake addresses and identifiers. Keep customer names, home imagery, phone numbers, invoice identifiers, payment information and privileged property access out of the sample. Get approval from the company and the people appearing in any recording before sharing it.
First step. Tell Fiund whether you have permissioned historical records or a team that can produce a controlled demonstration. Record to a buyer specification once you have an agreed project. Describe collection capability.
Industry classifications covered
- 238220 · Plumbing, Heating, and Air-Conditioning Contractors
Build a pilot sample
Make the service-ticket and first-person samples separate deliverables. In a staged trap task, record the part selection, seat check and result as three marked intervals. The ticket should preserve the approved fix and evidence of completion without using a customer job as the test set.
Fields the reviewer needs
- Fixture and attempt
- Keep verification attached to the same training example.
- Part identifier
- Record what was visibly selected.
- Action and result interval
- Separate the handling sequence from its observed endpoint.
- Participant and location authority
- Connect the recording to the permissions for the practice setup.
Use How construction and trade teams can contribute data for AI to turn the specimen into a brief with a defined task, source and review owner.
Fictional exercise
Match a fixture demonstration to its own verification
A closed training exercise uses an isolated, unpowered practice fixture and inert marked props. It evaluates the recording and its evidence, without instructing a repair or establishing leak performance.
| Record ID | Source | What it records |
|---|---|---|
| TASK-P1 | Original practice instruction | Attempt P1 names fixture F1 and marked practice ring A. |
| VIEW-P1 | Source-view record | The view shows ring A selected for F1, but the final contact region is obscured. |
| CHECK-P2 | Independent review for another attempt | A reviewer confirms a visible result on F2 during attempt P2. |
| NOTE-P1 | Operator note | The operator says P1 was intended to be complete, with no matching final check. |
Question for the model. Does the packet establish P1's final visible result? Identify the usable evidence and the gap.
Show the reference answer
VIEW-P1 supports the selected ring and fixture identity but not the hidden final region. CHECK-P2 belongs to F2 and P2, so it cannot verify F1's result. NOTE-P1 states intent only. A matching final view or authorized independent check for P1 is missing. The answer should preserve that visibility gap and make no functional or leak-performance finding.
How to check the result
Ask for the answer, the record IDs that support it and any fact the evidence leaves unresolved. Have a reviewer compare the response against the reference before deciding whether the sample is ready to expand.
Errors to flag in this case
- Treating selection of the requested part as proof of the final state.
- Borrowing another fixture's review to close an obscured result.
Change the inputs and run it again
Add a matching clear endpoint for P1 while hiding the selected ring's identifier. The remaining uncertainty should move to identity rather than disappear.
Add a second attempt with a similar-looking component or a different closeout state. Have a skilled reviewer identify the relevant interval and explain the label.
Track correct answers, answers using the wrong source and appropriate abstentions separately. Hold the answer key apart from the input and record which file or version would resolve each disputed case.
Check permission and scope
Use approved props and site procedures, with no customer premises or live service. Exclude addresses, invoices, private access, real work orders and unsupported safety conclusions.
Work with a controlled training setup or authorized records. Keep customer premises, identities and unsafe live equipment out of practice capture.
Describe an authorized archive through Fiund’s existing-data assessment, or describe a new capture capability through the collection brief. The contributor agreement covers the terms for participation.
Sources and further reading
Resources for this workflow
Includes worksheet
How construction and trade teams can contribute data for AI
Compare a documented service ticket with an authorized trade demonstration.
Includes worksheet
First-person video capture checklist for an AI data pilot
Plan the view, visible task outcome and source permissions for a practice fixture.
Company and program reviews
Use these reviews to compare sourcing, capture or evaluation approaches. Confirm the applicable program, accepted sources and terms with the provider.
Provider cited in this guide
Scale AI review: data partnerships, Outlier and buyer services
Its company partnership lists plumbing under field services and describes structured operational records.
Provider cited in this guide
micro1 review for experts, trade footage and company data
Its public trade capture materials show plumbing as an example. Confirm the actual project and review terms.
Related industry examples
Industry guide · NAICS 238220
HVAC diagnostics data for evaluating AI fault review
A technician may see two alarms that look alike while the system’s operating conditions point to different causes.
Industry guide · NAICS 238210
Electrical contractor data for AI starts with a controlled training setup
Electrical service companies know how technicians turn a messy symptom into a verified repair.
Industry guide · NAICS 238290
Elevator installation handoff documents for AI review
A work packet can contain a supplier report, installer acknowledgment and independent acceptance record.