PilotProcess & self-service2026-06-30
Honolulu reported permit review falling from 73 to 32.5 days and corrections from 23.5 to 7.7 per application on its AI plan-check software, and said it will become mandatory this year
Architectoccupation page →
Event date / reported
2026-06-30
Evidence stage
PilotSmall-scale trial in a real setting. Tells us the deployment conditions are being tested, not that they hold.
Tasks this bears on
Construction documentation
The drawings and specifications a contractor builds from — hundreds of sheets, coordinated with engineers.
Being augmented✓ Evidence-backed
Where this applies
One US city, residential projects only — single- and two-family homes, duplexes, ADUs and Ohana units; commercial is 'expected later this year'. The comparison rests on roughly 40 applications since December 2025, self-selected by applicants who chose to use the tool, which is exactly the population most likely to submit cleanly anyway. The figures are the city's own. The department is explicit that neither the tool nor the Priority Review programme guarantees approval — other agencies still review.
What this means
The check on a set of drawings is moving to before a human sees them, and in one city it is about to become compulsory. Read the corrections number rather than the days: 23.5 down to 7.7 means the machine is catching the things that used to come back as comments — completeness, code items, the routine misses. What it cannot catch is the reason the building is shaped the way it is, and that is the half of the job to be able to defend.
What it does not yet show
Forty self-selected applications in one city is a small, biased sample, and the comparison group did not choose the tool. The city published no error rate — nothing here says how often the software flags something that is actually fine, which is the number an architect pays for in rework. Residential only, and 'mandatory later this year' has not happened yet.
What you can check
If your jurisdiction adopts one of these, ask the department for two numbers before you change how you draw: the false-flag rate, and whether a machine comment can be appealed to a human and how. A checker you cannot argue with turns every ambiguous code provision into its strictest reading, and that lands on your drawings, not on the reviewer's desk.
Does it change the assessment?
No. The impact index is never moved by a single event. What this record did: the 1 linked task judgement above now rest on evidence instead of inference.
Source
Spectrum News Hawaii (City & County of Honolulu DPP) · verified 2026-09-11 · Claude (CTO/COO) — source read in full 2026-09-11 · interpreted 2026-09-11 · Claude (CTO/COO)