Helonic is an AI construction drawing analysis platform for architects and contractors reading the Honolulu and Seattle AI permit pilot results. Honolulu and Seattle are the first two US cities to publish numbers from an AI permit pilot, and both tested the same tool, CivCheck. This Helonic breakdown reads the city reports directly: what each one measured, how big the samples were, and what the results mean for the team preparing a permit set.
Honolulu's Department of Planning and Permitting (DPP) reported that residential permits screened with CivCheck came back with 7.7 corrections on average, against 23.5 for comparable permits that skipped it. Seattle's Department of Construction and Inspections (SDCI) ran a proof of value with the same tool and published accuracy figures, a model of how many review cycles it could remove, and a list of the corrections applicants miss most often.
Both are city-side programs. The city picked the tool, configured the checks against its own rules, and asks applicants to run their plans through it before intake. A design team reviewing its own set before it reaches a city portal is doing a different job, and that earlier step is where Helonic works.
Honolulu's AI permit pilot reported 7.7 corrections per CivCheck-screened residential permit versus 23.5 without it, from 19 and 17 permits. Seattle found CivCheck matched SDCI reviewers on 87% of in-scope intake issues and 91.7% of plan review checks. Both are city-configured tools. Helonic reviews the full drawing set for the applicant before submittal, across architecture, structure, and MEP.
Honolulu started with a five-month CivCheck pilot on building permit code review, noted in the city's FY2024 Citizen-Centric Report, then launched the tool for applicants in December 2025. In its July 2026 Priority Review release, DPP published the comparison below. It covers eligible residential work only: single-family, two-family, duplex, ADU, and ohana projects, including new construction, additions, and alterations.
| Honolulu DPP metric | Without CivCheck (17 permits) | With CivCheck (19 permits) |
|---|---|---|
| Average corrections per permit | 23.5 | 7.7 |
| Average plan review cycles | 3.4 | 1.4 |
| Average days to permit decision | 73 | 32.5 |
The sample is small, and DPP says so. A DPP spokesperson told the Honolulu Star-Advertiser that 19 completed applications is "a relatively small sample" and that the department will keep reporting as the dataset grows. The comparison group is 17 permits with enough data to compare.
The program has since moved from optional to required. According to Hawai'i Public Radio, residential applicants had to start using CivCheck in September 2026. The same report notes that building permit applications rose 34% after the HNL Build portal launched and the backlog appeared to grow. Fewer corrections per permit and a shorter queue are two separate measurements.
Mayor Rick Blangiardi put the goal plainly in his 2026 State of the City: "Plans cannot go through eight, nine, or ten review cycles." He also asked design professionals to hold up their end, which is the part of the problem a city tool can only partly reach.
Seattle's test ran from March to October 2025 and focused on middle housing: townhouses, single-family homes, and ADUs, which make up most new residential permits in the city. CivCheck ran real applications, SDCI staff reviewed the same applications, and the city compared the two. Innovation & Performance published the findings in June 2026, with the detail in the CivCheck evaluation report.
For intake, CivCheck was configured with 34 completeness checks and run on 29 applications SDCI staff had already screened. Staff had logged 90 rejection reasons on those applications, and 35 of them fell outside the checks in the pilot. Of the 55 that were in scope, CivCheck and staff both flagged 48, which is the 87% figure. CivCheck missed 7.
For plan review, SDCI reviewers saw CivCheck's pass or fail result on 67 configured zoning and ordinance/structural checks across 57 applications, 1,262 checks in all. CivCheck matched the reviewer's final decision 91.7% of the time. It passed an item the reviewer sent back on 6.4% of checks, and failed an item the reviewer approved on 1.9%.
The calendar and cycle figures in the report are scenarios. Scenario A uses the checks as configured in the pilot. Scenario B assumes CivCheck automates every intake check and removes one full plan review cycle. Both assume applicants fix everything CivCheck flags before they submit.
| Seattle SDCI metric | Actual, no pre-screen | Scenario A (pilot checks) | Scenario B (projected) |
|---|---|---|---|
| Median calendar days to pass intake | 54.5 | 29 | 8 |
| Average intake review cycles | 2.55 | 1.75 | 1 |
| Median calendar days to pass plan review | 161 | 146 | 117 |
| Median correction cycles in plan review | 3.72 | 3.39 | 2.72 |
Seattle's Q2 2026 AI usage report to City Council lists the CivCheck pilot as completed with a decision pending. SDCI had $750,000 appropriated for an AI permitting tool and had not picked one, and it was running a separate pilot with Archistar.
Seattle's most useful finding is about coverage. An applicant saves the most time when pre-screening removes an entire review cycle, and that only happens if the tool flags every correction a reviewer would raise. If it catches some of them, the applicant can still go through the same number of cycles.
Completeness checks are easier to cover than compliance checks. Completeness is a short list of city document standards. Compliance means testing the design against thousands of code requirements, some of which Seattle calls conceptually complex and hard to automate. The report also notes that plan review averages six review teams working in parallel across three or more cycles, so shortening one review only helps if it was the one holding up the cycle.
In both cities the applicant runs CivCheck, but the city decides what it checks. Honolulu's version covers residential work, with commercial projects expected later. Seattle's pilot covered 34 intake checks and 67 plan review checks for middle housing.
An applicant-side review starts from the drawings instead of the city's list. Helonic reads every sheet of the 2D PDF set and compares the sheets with each other: the code analysis against the life-safety plan, the door schedule against the floor plans, ducts against the framing. Each finding carries a severity and a sheet location, across ten categories that include code compliance, coordination, missing information, fire safety, and accessibility.
The tool-by-tool version, with who buys each one, is in city AI plan review vs applicant AI plan check. For how building departments are using AI on their side of the counter, see AI plan check for AHJs.
Helonic runs before the set leaves the office, which is also before a city pre-screen would see it. The two stack. A design team can clear code and coordination conflicts in Helonic, then run the city's required pre-screen on a set that already agrees with itself.
On a 448-page Phoenix hospital addition, a general contractor ran Helonic the week before the plan-check set and confirmed 632 issues. The design team moved ducts off the beams and added medical-gas penetrations before issue. Plan check still commented on finishes. It did not send back the duct clashes.
The code side of that review is described on Helonic's code compliance checking. For the category as a whole, start with what AI plan review is and the longer AI plan review guide.
Practitioner insight
“When a city adds an AI pre-screen, the first thing architects ask us is whether they still need their own QA. Most decide they do. The city tool checks the city's list. The comments that cost a whole cycle are usually the ones where the architectural sheets and the consultants' sheets disagree, and nobody's checklist has that one on it.”
Source: Conversations with architects and permit expediters preparing residential and commercial permit sets in jurisdictions adding AI pre-screening, synthesized from Helonic customer interviews, Q3 to Q4 2026.
Milind is the co-founder and CEO of Helonic, where he leads product and go-to-market for AI-powered construction drawing analysis. He works closely with general contractors, project managers, estimators, and owners to understand how drawing quality drives project outcomes - and where AI can reduce RFIs, change orders, and rework. Milind has interviewed hundreds of construction professionals across project delivery roles, from preconstruction estimators at ENR top-400 contractors to facilities directors at institutional owners, and uses those conversations to shape both product direction and the way Helonic talks about the work.
How this page was researched: Reviewed against Honolulu DPP's July 2026 CivCheck Priority Review release, Seattle Innovation & Performance's 2026 CivCheck evaluation report, and Seattle IT's Q2 2026 AI usage report to City Council.
Last reviewed by Milind Sagaram · October 2026
More on getting a permit set through plan check in fewer rounds.
Which tools cities buy, which tools design teams buy, and what each one checks.
The discipline-by-discipline check to run before a permit set goes to the city.
Response letters, clouds and deltas, and the reasons sets go to a third round.
How building departments are using AI on the review side of permitting.