How do I turn agency plan review comments into a response matrix and resubmittal check with AI?

Short answer

Use Claude or ChatGPT, then Copilot in Excel or Gemini in Google Sheets: Claude or ChatGPT extracts every agency comment into a matrix with the reviewer's numbering kept, drafts response skeletons that say what changed and where, and before resubmittal lists comments with no response, a bare done, or a sheet reference that is not in the revised set. Plan review comments arrive as PDF letters, scanned markups, portal exports or a spreadsheet, a different format for every agency, and someone retypes them into a response matrix before the team can start. Have Claude or ChatGPT extract every comment into one table with the reviewer's numbering kept, assign each to a discipline, and draft a response skeleton that states what changed and where, so the letter never says just "done". Before resubmitting, have the AI compare the matrix against the revised sheet list and your own notes to list comments with no response, no sheet reference, or a reference to a sheet that did not change. What the response says, and whether it satisfies the reviewer, stays with the engineer. A person still checks: every response and sheet reference with the responsible engineer, a read-through of the revised sheets, and unclear comments with the reviewer by phone before resubmitting.

AI tie-in

Tool: Claude or ChatGPT, then Copilot in Excel or Gemini in Google Sheets

Claude or ChatGPT extracts every agency comment into a matrix with the reviewer's numbering kept, drafts response skeletons that say what changed and where, and before resubmittal lists comments with no response, a bare done, or a sheet reference that is not in the revised set.

Prompt 1

Here is a plan review comment letter [paste text or upload the PDF]. Extract every comment into a table: agency, review cycle, comment number exactly as written, discipline, comment text copied exactly, sheet referenced (or NOT SHOWN) and comment type if stated. Keep the reviewer's order. Write UNREADABLE for anything you cannot read. Do not merge, reword or skip comments, and do not suggest how to resolve them.

Prompt 2

Here is my completed response matrix [paste: comment number, comment, response, sheet and detail changed, status] and the list of sheets revised in this resubmittal [paste]. List every comment with no response, any response that only says done, corrected, fixed or similar, any response citing a sheet not in the revised list, and any comment from the previous cycle [paste] that has no closing response. Do not judge whether a response satisfies the reviewer.

Human check: Every response and sheet reference with the responsible engineer, a read-through of the revised sheets, and unclear comments with the reviewer by phone before resubmitting.

The problem

Turn agency plan review comments into a response matrix and resubmittal check with AI. This is a common, time-consuming task for a civil engineer.

What this workflow does

Plan review comments arrive as PDF letters, scanned markups, portal exports or a spreadsheet, a different format for every agency, and someone retypes them into a response matrix before the team can start. Have Claude or ChatGPT extract every comment into one table with the reviewer's numbering kept, assign each to a discipline, and draft a response skeleton that states what changed and where, so the letter never says just "done". Before resubmitting, have the AI compare the matrix against the revised sheet list and your own notes to list comments with no response, no sheet reference, or a reference to a sheet that did not change. What the response says, and whether it satisfies the reviewer, stays with the engineer.

How it works: AI drafts or analyzes. A person operates the tools, checks the work, and takes action.

Your checkpoint: The AI organizes the comments and checks the matrix for gaps. It does not decide whether a design change satisfies the reviewer or the applicable code. Every response, change and sheet reference is the responsible engineer's, and unclear comments are best settled by calling the reviewer.

Before you start

Access to the named tool; confirm license, version and cost with the vendor.
A small, non-confidential sample to try the workflow on first.
A qualified person to review the output before it is used.

Tools and access

Follow the steps

Collect every comment source for the review cycle: the comment letter, any markup PDFs and portal exports. Remove anything your company policy does not allow you to upload, and use a business AI account where required.
Have Claude or ChatGPT extract each comment into one table: agency, review cycle, reviewer's comment number exactly as given, discipline or department, the comment text, sheet or document it refers to, and type (correction, question, information only) when the agency marks it. Comments it cannot read are listed as UNREADABLE, not paraphrased.
Paste the table into Excel or Google Sheets and add your internal columns: responsible person or discipline, due date, status, the sheet and detail where the change was made, and an internal reviewer sign-off. Keep the reviewer's wording untouched in its own column.
For each comment, have the AI draft a response skeleton in the form the agency asks for: what was changed and where, or the answer to the question, with blanks for the sheet and detail. The responsible engineer writes the actual response.
Before resubmittal, give the AI the finished matrix and the list of revised sheets with their revision clouds or deltas. Have it list comments with no response, responses that say only done or corrected, responses that cite a sheet that is not in the revised list, and comments from the previous cycle that were never closed.
Build the response letter from the matrix with the original comment above each response, have the internal reviewer sign off, and keep the matrix for the next cycle so repeated comments are easy to spot.

Try a small example

Replace the bracketed fields with your own project details. No tested sample run is published for this workflow yet.

Sample input

Fill in the bracketed [fields] in the text below with your project data.

Starter prompt

Here is a plan review comment letter [paste text or upload the PDF]. Extract every comment into a table with these columns: agency, review cycle, comment number exactly as written, discipline or department, comment text copied exactly, sheet or document referenced (or NOT SHOWN), and comment type if the letter states one. Keep the reviewer's numbering and order. Write UNREADABLE for anything you cannot read. Do not merge, reword or skip comments, and do not suggest how to resolve them.

Expected output

Not published yet. Treat any output as a draft and check it against the source documents.

Check the result

Every value, citation or quantity traces back to a source document.
Missing information is marked as missing, not filled in.

Limits and recovery

The AI organizes the comments and checks the matrix for gaps. It does not decide whether a design change satisfies the reviewer or the applicable code. Every response, change and sheet reference is the responsible engineer's, and unclear comments are best settled by calling the reviewer.
Check your organization’s data policy before sharing project files with third-party tools.

Stop or switch when: The source cannot be verified, permissions are missing, or the result contains unsupported facts.

Frequently asked questions

How does AI help with this?

Claude or ChatGPT, then Copilot in Excel or Gemini in Google Sheets: Claude or ChatGPT extracts every agency comment into a matrix with the reviewer's numbering kept, drafts response skeletons that say what changed and where, and before resubmittal lists comments with no response, a bare done, or a sheet reference that is not in the revised set. A person still checks: Every response and sheet reference with the responsible engineer, a read-through of the revised sheets, and unclear comments with the reviewer by phone before resubmitting.