The Construction Phase Documentation Problem

For electrical engineers who serve as engineer of record (EOR) on construction projects, the construction phase generates a relentless stream of documentation tasks. Requests for Information (RFIs) arrive from the electrical contractor asking questions about conduit routing conflicts, substitution requests, and code interpretation. Submittals — shop drawings, equipment cut sheets, and installation details — require review against the project specifications and NEC. Commissioning activities generate test reports that must be logged, verified, and incorporated into the project closeout package.

Each of these tasks follows a predictable structure: review a document, compare it against a standard or specification, draft a professional response, log the item, and track it to resolution. This is exactly the type of work where AI assistance delivers immediate productivity gains — without requiring the engineer to delegate technical judgment.

This article covers how to integrate AI into each phase of construction documentation: RFI intake and response, submittal review, commissioning test report generation, and open-item tracking. The workflows are directly applicable using ChatGPT, Claude, or similar AI assistants alongside your existing Excel or SharePoint-based project management systems.

AI-Assisted RFI Drafting: Structure, Language, and Code References

A well-written RFI response has a specific structure: it acknowledges the contractor's question, cites the applicable specification section and NEC article, states the engineer's ruling clearly, and notes any conditions or approvals required (such as AHJ acceptance for a code substitution). The formal language and code citation requirements make RFI drafting time-consuming — a 5-minute engineering decision can take 20 minutes to document properly.

AI compresses this documentation time dramatically. The workflow is:

  1. Log the RFI with its number, date received, spec reference, and contractor's question in your RFI tracker.
  2. Paste the RFI question into your AI assistant with a structured prompt.
  3. Review the AI draft, verify the code and spec citations, and approve or modify.
  4. Issue the response and update the tracker.

Example prompt for an RFI response: "Draft an engineering RFI response for a construction project. RFI #047, dated June 15, 2026. Contractor question: 'Can 2-inch EMT conduit be substituted for 2-inch GRC (rigid metal conduit) in the mechanical room where conduits run along the ceiling in a non-corrosive, indoor environment?' Reference: Spec Section 260533 (Raceways and Boxes for Electrical Systems), NEC 2023. Provide a professional response citing applicable NEC articles and spec section, stating whether the substitution is permitted, and noting any conditions."

An effective AI response: "RFI Response — #047. Per NEC 2023 Article 358 (Electrical Metallic Tubing — EMT), EMT is permitted in dry indoor locations that are not subject to physical damage. Article 344 (Rigid Metal Conduit — RMC) is specified in Section 260533-2.1; however, Section 260533-1.5 permits equivalent or superior materials where approved by the Engineer. Based on the non-corrosive, indoor, non-physical-damage environment described, EMT conduit is an acceptable substitution for GRC in the ceiling of the mechanical room, provided: (1) all fittings are listed for EMT per NEC 358.6; (2) the installation complies with NEC 358.10 through 358.60; and (3) this substitution is subject to AHJ acceptance at inspection."

The engineer verifies the NEC articles and spec section, modifies as needed, and issues the response. The time from question to formal response drops from 20–30 minutes to 5–10 minutes per RFI.

Submittal Review Automation with AI

Electrical submittals — switchgear shop drawings, panelboard cut sheets, lighting fixture schedules, generator data sheets, conduit and cable specifications — must be reviewed against project specifications and NEC equipment ratings. The review requires checking a matrix of items: voltage rating, interrupting capacity, enclosure type, listing (UL or equivalent), and any project-specific options.

AI accelerates the extraction and compliance checking steps. The process:

  1. Export or paste the submittal text (most equipment cut sheets are available as PDFs that can be pasted or uploaded).
  2. Prompt AI to extract key parameters and compare against spec requirements.
  3. Use the AI output to populate your submittal compliance checklist.

Prompt for submittal review: "Review this switchgear submittal. Spec requirement: 480/277V, 3Ø, 4W, 65 kAIC, NEMA 1 enclosure, UL 1558 listed, main breaker 1200A, 3P. Submitted equipment: [paste cut sheet text]. Compare submitted values against spec requirements and generate a compliance table with columns: Item, Spec Requirement, Submitted Value, Complies (Y/N/Check), Notes."

A critical AI output might flag: "Short-circuit rating: Spec requires 65 kAIC at 480V. Submitted equipment rated 42 kAIC at 480V. NON-COMPLIANT — requires correction or engineer's acceptance of lesser rating with documented justification." This type of flag — comparing a single number against a spec requirement — is exactly where AI adds value and where human reviewers sometimes let errors slip through on large submittals.

The compliance checklist output from AI maps directly into your firm's submittal review log format. For recurring equipment types (panelboards, transformers, switchgear), save the prompt template and the compliance table format so the next reviewer can process the submittal with consistent criteria.

Commissioning Test Report Generation

Electrical commissioning generates a significant volume of test documentation: breaker primary injection test results, insulation resistance (megger) test results, ATS transfer time tests, UPS battery capacity tests, and generator load tests. Each test result must be logged against an acceptance criterion and a standard (NEC, NFPA 70, NFPA 110, IEEE 946, project specifications).

AI is highly effective at two steps in commissioning documentation: drafting the test plan before commissioning begins, and generating narrative summaries of test results for the commissioning report.

For test plan generation: "Generate a commissioning test plan checklist for an automatic transfer switch (ATS). Include: visual inspection items, utility-to-emergency transfer time test (acceptance criterion: ≤ 10 seconds per NFPA 110), emergency-to-utility retransfer test, voltage sensing test (transfer on voltage dip to <75% rated), generator cranking voltage test, load test (rated load for 2 hours per NFPA 110 8.4.2), and fail-to-transfer test. Format as a table with Test, Method, Acceptance Criterion, Standard Reference, Measured Value (blank), and Pass/Fail (blank) columns."

After logging test results, draft the commissioning narrative: "Summarize ATS commissioning test results. ATS-1 (600A, 480V, 3Ø): utility-to-emergency transfer time = 8.4 seconds (PASS, criterion ≤ 10 s per NFPA 110); voltage sensing test at 72% voltage = transfer initiated (PASS, criterion <75%); load test at rated load for 2.5 hours = PASS. Write a 2-paragraph commissioning report narrative for this ATS in professional engineering language, referencing NFPA 110 and project specification Section 263213."

Tracking Open Items and Generating Status Reports

Construction projects generate dozens to hundreds of open RFIs and submittals simultaneously. AI can automate the status report that every project team needs weekly: how many items are open, which are overdue, and what action is required from whom.

With an Excel-based RFI/submittal log (columns: ID, Type, Date Received, Spec Ref, Description, Assigned To, Required Response Date, Status, Response Date), prompt AI to analyze the data: "Here is my project RFI/submittal log [paste table]. Identify: (1) all RFIs and submittals overdue by more than 5 business days; (2) items due in the next 5 business days; (3) items with a status of 'Revise and Resubmit' that have been open more than 14 days. Generate a project manager's status summary with action items for each category."

For projects using Procore, Autodesk Construction Cloud, or similar platforms, most tools support CSV export of RFI and submittal logs. This creates a repeatable workflow: export the log, paste into AI, generate the weekly status summary. For teams that have integrated AI API access with their project management platform, this can be fully automated — the AI reads the platform data directly and generates the status report on a scheduled basis.

Prompt Templates for Construction Phase Documentation

Building a prompt library for construction phase AI documentation is one of the highest-return investments a project team can make. These templates, refined over several projects, produce consistent, professional output without starting from scratch each time:

  • RFI response: "Draft an RFI response for RFI #[number]. Contractor question: [paste question]. Spec section: [section number and title]. NEC 2023 article: [article]. Engineer's ruling: [state yes/no and conditions]. Format: professional engineering response with code citations."
  • Submittal compliance table: "Compare this submittal to the spec requirements listed below. Generate a compliance table: Item, Spec Requirement, Submitted Value, Complies (Y/N/Check), Notes. Spec: [paste spec section]. Submittal: [paste cut sheet data]."
  • NCR (Non-Conformance Report) draft: "Draft an NCR for the following field observation: [describe the issue — conduit installed without required support spacing, breaker installed in wrong position, grounding conductor missing from panel]. Reference NEC article and spec section. Include: description of non-conformance, NEC/spec reference, required corrective action, deadline, and responsible party."
  • Field report summary: "Summarize these field observation notes for a weekly construction progress report. [Paste notes.] Include: work completed, items in compliance, open issues requiring contractor action, and items scheduled for next week."

Document every prompt used in formal engineering deliverables in your project QA file. If a submittal response or RFI answer is ever challenged during a construction dispute or permit audit, the prompt and AI output log serve as part of the derivation trail — showing that the engineer directed the AI with specific criteria and reviewed the output before issuing it.