Example · Field photos Field photo checks
An AI agent that checks each set of site photos against a checklist, writes notes on what it saw, and passes anything it is unsure of to a person.
What it was for
- Problem
- A contractor that scopes work and checks feasibility from photos sent in from site, before the designers take a project on. Someone in the office checked every set against a checklist: is the right equipment in frame, is the label readable, is anything missing. Sets waited in a queue until a reviewer was free, and a missing photo meant chasing someone for it.
- How we worked through it
- One of us wrote down how an experienced reviewer works through a set, and built an agent that does the same. It zooms in where it cannot read a detail, completes the checklist with notes, and says how sure it is.
- Outcome
- Reviews start as soon as the photos arrive. Most finish at high confidence and go straight to the next team. Anything the agent is unsure of goes to a person, with its notes attached.
Where a person stays involved
Any review the agent rates medium or low goes to a person. The team that picks the job up still does its own work on every project, and the agent’s notes travel with the job so they can see what it looked at and why.
The same pattern in an engineering firm
- Site visit photos checked against the inspection checklist before the report is written
- Field reports checked for missing photos, readings or signatures
- Incoming submittals checked for completeness before an engineer opens them
Next example: single-line diagram generator →All examples
This tool was built by one of us as an employee, for that employer’s own work, before Datum existed. The tool itself stays with that employer. The approach is ours to carry: a vision model doing the first-pass review of site photos at volume is something we can set up again for any firm with the same kind of queue. The example is drawn, the figures are rounded, and nothing shown comes from a live system.
Have a review step that sits in a queue? Tell us about it.
