What an AI Plan Reviewer Actually Looks At

When people hear "AI plan review," they often picture something like autocomplete for drawings — a tool that checks spelling on notes or flags missing title block fields. That's not what we do. The questions worth asking are: what does a rigorous civil plan review actually cover, and how does AI address those areas systematically?

CivilPlanAI uses four specialized reviewers, each focused on a distinct domain. Here's what each one looks at and why it matters.

Grading and drainage

Grading and drainage errors are among the most consequential in civil engineering — and among the easiest to miss when you're deep in a large plan set. This reviewer checks slope calculations, flow direction, inlet sizing, pipe inverts, and whether the grading plan is internally consistent. It looks for conditions that could cause ponding, erosion, or downstream drainage conflicts, and flags locations where the design may not perform as intended.

This is the reviewer that catches the 0.2% slope on a parking lot that should be 2%, or the pipe invert that doesn't match between the plan and the profile sheet.

Code and standards compliance

Civil plan sets need to comply with a layered set of requirements — municipal standards, state highway design manuals, ADA guidelines, local drainage criteria, and more. This reviewer checks the plan set against commonly accepted publicly available standards and flags potential non-conformances.

Firms can also provide jurisdiction-specific standards and local requirements, which the reviewer incorporates into its analysis. The output isn't a legal opinion — it's a structured list of items that warrant the engineer's attention before submittal.

Drawing quality

A plan set that's technically correct but hard to read creates problems downstream — for contractors, reviewers, and inspectors. This reviewer looks at annotation clarity, dimensioning completeness, label accuracy, scale consistency, and whether the drawings communicate what they need to communicate without ambiguity.

It's the reviewer that catches a note referencing a detail that doesn't exist, a label pointing to the wrong feature, or a dimension that contradicts what's shown graphically.

Cross-sheet coordination

This is the reviewer that catches the class of errors that's hardest for humans to find: inconsistencies between sheets. When the grading plan gets revised but the drainage plan doesn't get updated to match. When the utility plan shows a manhole at a different elevation than the profile sheet. When a callout on one sheet references a standard detail that was superseded on another.

Cross-sheet coordination errors are so common because they require holding the entire plan set in mind simultaneously — something that's genuinely difficult for a human reviewer working under time pressure, and something that AI is well-suited to do systematically.

What the output looks like

Each reviewer produces findings that are pinned to the specific sheet and location where the issue was identified. Every finding includes a plain-English description of the issue, a code or standard reference where applicable, and a confidence level. The engineer of record reviews each finding and decides — accept, edit, or reject — before anything is incorporated into the work product.

The goal isn't to replace the engineer's judgment. It's to make sure that by the time the engineer sits down to review, the systematic work is already done.