Helonic is an AI construction drawing analysis platform for teams asking what issues Helonic catches on construction drawings.
The ten categories, with examples a reviewer would recognize on a set
Helonic catches ten categories of construction drawing issues: coordination, code compliance, missing information, structural, MEP, fire safety, accessibility, constructability, dimensions, and QA/QC. Each finding includes a high, medium, or low severity rating and exact page coordinates. Reviewers accept or reject findings before any RFI is created.
Helonic detects ten issue categories on 2D PDF construction drawings. The list below is the published scope, not a marketing synonym soup. If a finding does not fit one of these, it does not get forced into a nearby bucket.
| Category | What a finding looks like |
|---|---|
| Coordination | Duct routed through a beam, a door swing into a fixture, a riser that misses the shaft on the architectural plan. |
| Code compliance | Egress width that does not match occupant load, a dead-end corridor over the allowed length, a missing rated assembly at a shaft. |
| Missing information | A callout with no matching detail, a door tag with no schedule row, a note that says 'see structural' on a sheet that never shows the member. |
| Structural | A beam size that changes between the plan and the schedule, a foundation note that conflicts with the typical detail, an opening that was never framed. |
| MEP | A panel that lacks working clearance, a pipe slope that dies into a beam, a diffuser that lands on a light on the RCP. |
| Fire safety | A sprinkler omission in a rated corridor, a pull station that misses travel distance, a damper missing at a rated penetration. |
| Accessibility | A restroom clear-floor space that fails, a ramp slope over the allowed ratio, a door that cannot swing without blocking the accessible route. |
| Constructability | A sequence that cannot be built as drawn, an access panel with no path, a detail that assumes a trade is already finished. |
| Dimensions | A string that does not add to the overall, a grid that drifts between disciplines, a height that conflicts between section and elevation. |
| QA/QC | An unclouded revision, a spec section that names a different material than the finish schedule, a sheet index that omits an issued drawing. |
How those findings get produced, from upload to accept/reject, is on how Helonic reviews drawings. The product definition is on what is Helonic.
Helonic rates each finding high, medium, or low. High is the pile that stops work, fails plan check, or creates a safety problem. Medium is a real conflict that can wait a day. Low is a documentation miss that still needs a decision.
Severity is a triage hint, not a stamp. A reviewer can promote or demote a finding after looking at the sheet. The coordinates stay with the finding either way, which is what makes AI for construction drawings usable in a real queue.
A coordination finding cites two (or more) sheets that disagree about the same space. The duct is on M-201. The beam is on S-301. Both are drawn as if the other trade is not there.
Helonic does that from 2D PDFs. A BIM clash test can still run in parallel if you have a model. Most field conflicts still start on sheets that never made it into that model, which is why how to find clashes in 2D drawings remains a live skill. Clash detection is the Helonic surface for this category.
Helonic flags code-compliance and accessibility findings when the drawing set shows a condition that conflicts with published requirements such as IBC, ADA, NEC, and NFPA. A dead-end corridor, a restroom clear-floor miss, and a panel without working clearance are typical.
Code findings still need a licensed reviewer. Editions, local amendments, and occupancy choices change the answer. AI building code review and code compliance cover that path. How to review construction drawings is still the place to put those findings in a comment log.
Helonic flags candidates. A reviewer still decides which findings become RFIs. Intent that lives only in a conversation, a verbal field decision, or a model that was never issued will not show up on the PDF pass.
False positives happen. That is why accept and reject exist. Across 1,000,000+ pages analyzed, the useful pattern is the same: coordination and missing information make up most of what teams act on, and code or accessibility flags are the ones that used to wait until plan check.
Practitioner insight
“Coordination and missing info still make up most of what we act on. The code and accessibility flags are the ones that used to wait until plan check, and those are the expensive surprises.”
Conversations with QA/QC managers at design firms reviewing IFC sets before issue.
Manas is the co-founder and CTO of Helonic, where he leads engineering and AI research for construction drawing analysis. He works directly with structural, MEP, civil, and fire protection engineers to translate the way they review drawings into AI systems that flag the issues that actually matter in the field. Before Helonic, he built machine learning pipelines for technical document understanding and has spent the last several years interviewing licensed design engineers and discipline leads to ground product decisions in real practice rather than industry assumptions.
How this page was researched: Category list matches Helonic's published 10-category detection scope (coordination, code compliance, missing information, structural, MEP, fire safety, accessibility, constructability, dimensions, QA/QC). Examples drawn from patterns observed across Helonic's 1,000,000+ analyzed pages through September 2026.
Last reviewed by Manas Gandhi · September 1, 2026
Related reading on Helonic and drawing review.
The product definition and what Helonic does not replace.
Upload, full-set read, triage, and RFI push.
The manual method for the coordination category.
How code findings get checked against IBC and referenced standards.
Where category detection sits in a review process.
The product surface for 2D coordination findings.