AI for electrical testing companies, tried first on reports you already signed.

A closed job already has its answer. On a project you finished, the sheets your technicians filled in sit beside the report your engineer signed. A finished study holds the sheets on one side and, on the other, the input data the engineer ended up trusting. Those pairs are all a proof of concept needs, and nothing new has to be tested to get them.

Independence is why the paper stays yours. NETA describes its accredited companies as third-party firms that must be divested of competing service or manufacturing interests. The University of Maryland's facility standards want a testing agency whose principal revenue is testing, free of common ownership or control with manufacturers, contractors or consulting firms. On request, Portland State's 2013 specification wants a statement of no conflict of interest and examples of your test data forms. A buyer pays for an unbiased opinion, so this test assembles paper and offers no opinion of its own.

Testing is limited to qualified people. OSHA's electrical standard, at 29 CFR 1910.334(c), says only qualified persons may perform testing work on electric circuits or equipment. NETA's certified technicians hold Level 2, 3 or 4, and NETA treats the certification as valid only while the technician works for an accredited company. Nothing here moves that line. The AI reads copies of what those people already produced.

Few binders have one author. The University of Arizona's design manual has a third party test medium voltage cable, transformers and switchgear, leaves the rest to the electrical contractor, and tells the engineering firm to rewrite NETA's text for the project instead of simply citing it. So one acceptance binder can mix your sheets with the installer's, in formats nobody chose together, and a missing page may be someone else's.

The standards run on a calendar. NETA revises each ANSI/NETA standard on a four-year cycle: ATS in 2025, ETT in 2026, MTS in 2027 and ECS in 2028. Specifications stay frozen at whatever they cited when written. Portland State's, for one, names ATS-2013 for test voltages and ETT-2000 for technicians. A package has to say which edition it followed, and a sample of closed jobs shows how often your files answer that.

Hospitals add a retention clock. NFPA 99-2021, the Health Care Facilities Code, added site acceptance testing for electrical components serving Category 1 and 2 spaces, such as operating and trauma rooms and inpatient bedrooms. Consulting-Specifying Engineer's summary says the records are kept at least five years, go to the authority having jurisdiction on request, and that no third party is required. It points to NETA ATS and Chapter 31 of NFPA 70B as references.

A study can't begin without the walk. That's what makes the field sheets the other fair first job. IEEE 1584.1-2022 covers how a study's scope and deliverables get specified, and IEEE has a project open to revise 1584 itself. Choose this one first if your study backlog is worse than your report backlog.

Where this stops. Nothing in the proof of concept reaches test equipment, a study model, a live report or a live site. It doesn't judge a result, calculate incident energy, set a protective device or sign anything. A qualified person makes every technical call and signs every report. The documents above are summarized as background, and none of this is legal or engineering advice.