How the AI fits into the workflow
Think of the AI as a drafting assistant, not an autopilot. In the field, it writes a narrative or extracts a reading so you can move to the next room. At the desk, it drafts anything left blank, cleans up language and flags anything that does not match the standard. You still approve every word. A typical path through an inspection uses the AI in three waves:- In the field, as you capture: AI Generate on tricky findings, Vision Scan on findings where the photo tells the story, Scan Label / Reading for every nameplate and gauge, Smart Search for anything you have already written up before.
- On the drive back: Finish & Draft Report runs one batch call on the server that drafts every blank narrative.
- At the desk: Review flagged narratives, Polish anything you want tightened, draft System Performance Opinions, publish.
Field capabilities (mobile)
AI Generate
Writes a finding’s narrative from your caption, the field note and the inspection standard. Produces a Condition / Analysis / Recommendation (CAR) narrative that you can edit on the spot.- Where: Capture screen on the finding
- Best for: Common findings where your caption and notes already describe the problem
- Tips: The more specific your caption, the better the narrative. Make sure the section and subsection are correct before you tap AI Generate: the AI uses that context to calibrate severity and recommended action.
Vision Scan
A modifier on AI Generate, not a separate button. Arm Vision Scan, then tap AI Generate. The photo goes up with the caption and field note, and the AI drafts the narrative from what it sees in the picture. The Co-Inspector pass runs on the same photo and flags up to three additional findings you did not note.- Where: Capture screen on the finding
- Best for: Findings where the photo tells the story better than words can
- Co-Inspector output: Accept each suggestion into a new finding or dismiss it. The suggestions are suggestions only. You are the licensed inspector and your judgment is final.
Scan Label / Reading
Reads values off a photo (a furnace data plate, a thermometer, an app screenshot) or routes a typedfield label: value into the matching field on the Ratings screen. A live scan shows a confirm screen before applying. A scan that was queued (for example, taken offline) applies its results automatically when it finishes and shows a review pill so you can check what it filled.
- Where: Capture screen
- Best for: Any time you would otherwise type a reading. One screenshot of a Fieldpiece or similar tool extracts every reading on screen in a single call.
- Match is forgiving: “Furnace supply air temperature” finds “Furnace Tested Supply Air Temperature” without a perfect match.
Smart Search
A free keyword search of your narrative library, cached on the phone so it works offline. Pulls a narrative you have already written and refined. Faster than regenerating when your library has a close match.- Where: Capture screen on the finding
- Best for: Repeat findings where the wording you have already approved is better than anything the AI would write fresh
Add to Summary
Not AI, but it works hand in hand with the AI. Marks a finding for the report’s summary section. Use it on safety items and potentially costly repairs so the client sees them front and center.Finish & Draft Report
The batch call. One tap syncs everything from the phone and asks the AI to draft every blank narrative on the server. When the drafts are back, a short banner appears on the Review screen. There is no push notification. A full report can take a while to draft, so start it and carry on; you don’t need to wait on the Review screen.- Where: Review screen
- Best for: The drive back to the office. Most inspectors leave light drafting for the field and let Finish & Draft handle the rest.
System Performance Opinion (on mobile)
Drafts a section-level professional opinion (for example, Foundation Performance). You can write it by hand if you prefer.Property Baseline (included)
A pre-built snapshot of the inspected property pulled from its address. In the web admin the panel is titled Property Baseline and splits into two subsections: Site Conditions (soil map unit, dominant component, shrink-swell potential, FEMA flood zone, EPA radon zone, termite pressure and DOE/PNNL climate zone with recommended attic insulation) and Construction-era materials (materials typical for this build year and region, as candidates to look for). Sources: USDA NRCS SSURGO, FEMA NFHL, EPA Radon Zones, IRC R301.2(6) and DOE/PNNL Climate Zones. Appears on the inspection setup screen on mobile and on the order in the web admin. Visible to staff only.Offline AI queue
Most paid AI calls defer when you’re offline. Scan Label / Reading photo scans queue on the device and finish automatically when you reconnect; narrative drafts you queue up also defer and run on reconnect. AI Generate and Vision Scan tapped live still require a live connection at tap time; use Finish & Draft on the drive back to batch any that weren’t drafted in the field.Office capabilities (web)
AI Draft
The web Reviewer version of AI Generate. Fills in every blank narrative on a finding one at a time, or in a batch.- Where: Report Workspace, per finding or in batch across the whole report
- Best for: Reviewing a mobile-captured inspection where the inspector left narratives blank for the office
AI Draft and Polish (single click)
Drafts every eligible finding, fills required performance opinions, then runs Polish Report. One button, one trip.- Where: Report Workspace
AI Polish (single narrative)
Rewrites one narrative for tone and clarity. Keeps the facts, polishes the wording. The web button label is AI Polish.Polish Report
Whole-report polish pass with a before and after review. Flags absolute claims so you can soften anything the standard requires you to hedge.Spell check
Fixes spelling and dictation stutters across the report. Limited to spelling so it never changes your findings.Vision Analysis (web)
The web equivalent of mobile Vision Scan. Drafts a narrative from a photo attached to a finding.Label and nameplate reading
The web equivalent of Scan Label / Reading. Fills equipment fields from a plate photo or an uploaded screenshot.Bundle Findings
Merges several related findings into one narrative. Useful when several captures describe facets of the same issue.System Performance Opinion
Drafts a professional opinion for a system from every finding in that section. Appears on the report summary.VoiceDNA
Learns your company’s writing voice from sample reports (up to 8). Future AI Generate calls produce narratives that match that voice. Adjust tone and length with simple dials, and choose how narratives are laid out in reports under Narrative format in reports. Owners and admins run VoiceDNA.Repair Request narratives (included)
When a buyer or buyer’s agent builds a repair request from the client-facing report, the AI rewrites each selected deficiency in a plain, actionable, contractor-ready voice. No separate credit cost.Business Insights (included)
Dashboard commentary written from three viewpoints: CEO (strategy), Operations (throughput and quality) and Finance (revenue and margin). Refreshes three times a day. Appears on the Dashboard.What keeps the AI honest
These are rule-based checks written in code, not AI. They are what let you ship AI-drafted narratives without worrying about the AI inventing details.Specialist writing rules
HVAC, electrical, plumbing, site and grading, gas and fire safety findings each get their own writing rules. The AI uses the right voice for the trade and refers the right trade professional for the work.Standard-of-practice checks
Every draft is checked against the rules of the selected standard and flagged for review if something is off. Examples:- Numbers must come from your captured readings. The AI cannot invent an amperage or a temperature.
- Causes cannot be stated as certain. “Possibly caused by” stays “possibly caused by.”
- Your hedging stays. If you wrote “appears to,” the AI will not promote it to “is.”
Plain-language pass
As narratives are drafted, jargon is replaced with plainer words. Your client reads the report.Computed, not guessed
HVAC temperature-split verdicts, equipment service life and roof-covering limitations are calculated from the readings you captured. The AI only words them. If you capture a cooling supply-return split outside the normal 15–22°F window, the finding reports “outside expected range” because the math says so, not because the AI guessed.Pre-publish checks
The report cannot be published until the required items are complete. Before publishing, you review every flagged narrative, and every finding without a narrative must be drafted or explicitly signed off (with a record of who signed it off).Finding types and the AI
The AI adjusts how it frames findings for each inspection type. The service or order tells it which type to use (not the template).Where this does not go
Features in this list are not available even if you have seen them mentioned in product screens or older docs:- Voice dictation is not built in. The app cleans up stutters and inspection-term typos (for example, “freeze” to “frieze”) in whatever your phone keyboard dictates, but there is no 1nspecT voice engine.
- There is no in-app report viewer on the mobile app. Reports are mobile-first web pages and PDFs that open in any browser from the client link.
- The AI does not cite specific building codes. Code compliance depends on permit date, local amendments and the authority having jurisdiction. If your standard requires a code reference, add it yourself.
Related articles
- AI Credits — what each capability costs
- AI During Capture — the mobile buttons step by step
- Report Workspace — AI in the office
- Narrative Library — building the library the AI pulls from