How to Turn Photos and Voice Memos into Job Notes with AI
Once the phone is covered, the next AI win for a trades crew is in the truck: a 20-second voice memo becomes a job note, and a nameplate or wet-ceiling photo becomes a triage note you can edit. ChatGPT (or a voice tool like Otter) turns messy field input into something the office or the invoice can use — without inventing model numbers, square counts, or prices you did not read. This guide is for flagship trades work on the job, not desk writing.
Why it matters: Typing a clean note with greasy gloves loses the detail you just saw. A short voice memo or photo caption, pasted into AI with a strict "missing = missing" rule, is faster than rebuilding the job from memory at 8pm — and safer than letting the model guess.
Step-by-step guide
Follow these steps to get the best results from AI for this task.
Cover missed calls before you optimize notes
Photo and voice workflows do not catch the lead that went to voicemail. If after-hours rings still die, set up answering first: /how-to-use-ai/missed-calls and your trade /ai-receptionist-for-* page. Then come back here for truck notes.
Record a short voice memo before you leave the driveway
Twenty seconds is enough: customer, address, what you found, parts to order, labor feel, and one sentence the homeowner can read on the invoice. Do not wait until you are back at the shop. Otter or ChatGPT can transcribe; the prompt below turns the mess into a structured job note. Paste the cleaned note into Jobber or your FSM before you drive — AI drafts, the FSM holds the record.
Turn this messy voice transcript into a trades job note with: customer name, address, what we found, parts to order, labor hours if said, and a one-sentence invoice description in plain English. If a field was not said, write missing — do not guess. Transcript: "{transcript}"Photo the nameplate; confirm every number against the plate
For water heaters, RTUs, and attic units, shoot the data plate (or read the model string into the same voice memo). Ask AI for brand, model, serial, and only the fields that are actually readable. Label anything unclear as missing. Do not let the model invent recovery rate, refrigerant charge, or warranty years you did not read. HVAC and plumbing hubs on this site use the same rule.
Turn this equipment data-plate text into a job note: "{nameplate}". Output: brand, model, serial, fuel or refrigerant if stated, capacity if stated, and questions I should still verify on site before I quote. If a field is missing, say missing. Do not guess efficiency ratings or warranty years.Triage homeowner photos before the truck rolls
A dark wet-ceiling shot or a curb-access photo can draft a triage note: active drip vs stain, access limits, suggested first truck, and unknowns. You still decide who rolls. Do not invent square, pitch, or a claim number. Roofing examples live on /roofing; receptionist coverage for storm week is /ai-receptionist-for-roofers.
Write a leak or access triage note from only what I describe. City: {city}. What the photo shows: {photo_notes}. What they said: {call_notes}. Output: active drip / stain only / unclear (or access limits), suggested first truck, and unknowns. Do not invent square, pitch, shingle type, or a claim number.Voice memo to a first-pass quote outline — prices from your book only
After the note is clean, ask AI for a quote outline: line items as labels, not invented dollar amounts. Paste prices only from your price book or FSM. If a price was not provided, leave it blank. Honesty: the model drafts structure; you own the number.
Turn this job note into a first-pass quote outline for a {trade} shop. Job note: {note}. List line items as labels only. Leave price blank unless I paste a price-book amount. Flag unknowns. Do not invent labor hours, material costs, or discounts. Plain language the homeowner can read.Keep proof in the tool you already use
AI drafts the note; Jobber, Housecall Pro, CompanyCam, or your FSM holds the photo and the signed invoice. Do not buy a second product just to store pictures if your current stack already does. For phone gaps that remain after notes are clean, return to /how-to-use-ai/missed-calls.
Copy-paste prompts
Use these in ChatGPT, Claude, or Gemini. Replace the [brackets] with your own details.
Turn this messy voice transcript into a trades job note with: customer name, address, what we found, parts to order, labor hours if said, and a one-sentence invoice description. Missing fields stay missing. Transcript: "{transcript}"Turn this equipment data-plate text into a job note: "{nameplate}". Brand, model, serial, fuel or refrigerant if stated, capacity if stated, verify-on-site questions. Say missing when unread. Do not guess efficiency or warranty.Write a roofing leak triage note from only what I describe. Photo shows: {photo_notes}. They said: {call_notes}. Active drip / stain only / unclear, room/story if known, first truck, unknowns. No invented square, slope, or claim number.Write a curb-access triage note from only what I describe. Photo shows: {photo_notes}. They said: {call_notes}. Gate/fence, parked cars, overhead lines, ladder side, first truck, unknowns. No invented pitch or square.Turn this job note into a first-pass quote outline for {trade}. Note: {note}. Line items as labels only. Leave price blank unless I paste a price-book amount. Flag unknowns. Do not invent costs.Frequently asked questions
Can AI read a nameplate photo accurately?
It can draft a unit record from what is readable in the photo or from text you typed off the plate. Confirm every number against the plate before you order parts or quote a changeout. Mark unread fields missing — do not trust guessed serials or warranty years.
Can AI tell me which truck to send from a wet-ceiling photo?
It can draft a triage note from what the photo shows and what they said. You still decide tarp vs look-only. It does not invent square counts or a claim. Storm-week phone coverage still comes first: /ai-receptionist-for-roofers.
Will AI invent prices on a voice-to-quote prompt?
It will if you let it. Tell it to leave price blank unless you paste a price-book amount, and to flag unknowns. Your FSM or price book owns the number — same honesty rule as the flagship hub FAQs.
Should I do photo notes before an AI receptionist?
No. Missed calls lose the job before a note matters. Start with /how-to-use-ai/missed-calls and your trade receptionist page, then use these prompts in the truck.
Do I need Otter if I already have ChatGPT?
Not required. Otter (or any voice tool in the catalog) helps when you want a clean transcript from a long driveway memo. ChatGPT alone works if you paste the transcript or type the plate string yourself. The honesty rule is the same: missing stays missing.
Can AI map a panel or EV charger from one photo?
It can draft a triage note from what is readable. Confirm every breaker label and amp rating on site before you quote a service upgrade. Electrical hub FAQs on /electrical cover the same rule; phone coverage for panel calls is /ai-receptionist-for-electricians.
Where do the finished notes and photos live?
In the tool you already pay for — Jobber, Housecall Pro, CompanyCam, or your FSM. AI drafts; it is not a second photo vault. If rings still die after notes are clean, return to /how-to-use-ai/missed-calls.
More step-by-step AI guides
How to Use AI for Missed Calls — Step-by-Step for Trades
For a trades shop, the first AI hire is the phone — not another writing bot. After-hours b…
How to Use AI for Meeting Notes
Meeting notes are one of those tasks that eats time but rarely gets done well. Either you'…