Every job your crew finishes is a story: what was wrong, what you did, what it looks like now. Most of that story ends up in a camera roll, never seen by the customer. The tech uploads the photos, closes the app, and the job fades into a folder with a date on it.
This weekend you can build a workflow that reads those photos automatically and writes a professional, customer-ready completion summary — in minutes, without touching a word. The build: Google Drive receives the photos, n8n detects them and sends them to Gemini's vision model, and Gemini drafts a clean summary that lands in your inbox ready to review and forward. No ongoing subscription required.
Why this is worth building
Customers who receive a clear post-job summary — what was found, what was done, what it looks like now — are easier to work with. They're less likely to dispute an invoice, more likely to refer you, and more likely to leave a review. The problem isn't that contractors don't see the value. It's that writing the report takes time nobody has at the end of a long day on the tools.
AI vision models have become genuinely good at describing photos. Feed Gemini a photo of a replaced water heater, a cleared drain, or a rewired panel, and it produces a coherent, accurate description of what the image shows. Pair that with a structured prompt and an automation that fires without human input, and the report largely writes itself — leaving you a draft to review, not a blank page to fill.
What you'll build
The workflow has four steps:
- Trigger: A Google Drive folder (e.g.,
jobs/completed) is monitored by n8n. When a tech drops photos into a job-specific subfolder, the workflow fires. - Download: n8n fetches the new photos from Drive.
- Analyze: The photos go to Gemini via the n8n Google Gemini node with a structured prompt that tells it to write a professional completion summary in your voice.
- Deliver: n8n emails the draft summary to you. You review, add any details Gemini couldn't see, and forward it to the customer.
n8n already publishes a workflow template for tracking construction site progress with Gemini AI and Google Drive photos — you can import it and adapt the prompt to your trade. There's also a more general image and PDF analysis template that shows five different ways to pass files to Gemini. Start with one of these rather than from scratch — the plumbing is already done.
What you need
- Google Drive — free, and you almost certainly have it. Create a folder structure like
jobs/completed/[job-name-or-number]. Your techs drop photos into the right subfolder when the job is done. - n8n — the self-hosted Community Edition is free with no execution limits. If you have Docker installed, one command starts it:
docker run -it --rm --name n8n -p 5678:5678 -v ~/.n8n:/home/node/.n8n docker.n8n.io/n8nio/n8n. n8n Cloud also offers a free trial if you'd rather skip Docker for now. - Gemini API key — free from Google AI Studio. No billing setup required to get started. Check current free-tier limits before going live, as Google adjusts these periodically.
The prompt that makes it work
The Gemini prompt is where you customize the output for your trade. Here's a starting template — adjust the trade type and the voice to match how you actually write:
"You are writing a professional job completion summary for a residential service business. Based on the attached photos, write a 3–4 sentence description of the work shown. Focus on: what was repaired or replaced, the condition of the equipment or area, and any notable findings. Use plain English — no jargon. Do not mention pricing. End with one sentence confirming the work area was left clean and tidy. If you cannot determine specifics from the photo, describe what is visible without guessing."
Once you have a version that consistently produces useful output, save it in the workflow. When a report misses the mark, improving the prompt is almost always faster than retaking the photo.
What to do with the output
The workflow delivers a draft, not a finished letter. Before you forward it to the customer:
- Confirm the summary matches what was actually done — Gemini reads photos, not job notes
- Add any details it couldn't see: specific parts replaced, warranty information, anything that was ordered and is still outstanding
- Remove anything that sounds off or generic
Many shops pair the AI summary with their invoice link in a single email: one message closes the job professionally and tells the customer how to pay. Others use it as a leave-behind — printed and handed to the homeowner as you pack up. Either way, the hard part (turning photos into coherent sentences) is already done.
Honest limits
Gemini vision is capable, but it has real constraints. Dark or blurry photos produce vague descriptions. A photo of a clean finished wall with no before-context doesn't tell much of a story. And Gemini won't know your pricing, the specific model number of the part installed, or why you made a particular code-compliant choice — those details still need a human review pass.
The right mental model: treat the AI draft the way you'd treat a write-up from a capable apprentice who wasn't on the job. Read it before it goes out, every time. The goal isn't to eliminate your judgment — it's to eliminate the blank page at the end of a long day.
What to do this weekend
- Set up a Google Drive folder structure for completed jobs. Pull photos from three or four real recent jobs — those are your test cases.
- Get a free Gemini API key from aistudio.google.com — takes about two minutes.
- Start n8n locally with Docker, or sign up for the n8n Cloud free trial.
- Import the construction site tracking template from n8n's workflow library and adapt the Gemini prompt for your trade.
- Test with photos from those recent jobs. Read the output critically — if it's vague or wrong, adjust the prompt and run it again.
- Once you're happy with the output quality, drop in photos from your next real job and see what lands in your inbox.
By Sunday evening you'll have a draft pipeline. By the following week you'll know whether it actually saves you time — and that's a more valuable answer than any demo can give you. The photos are already being taken. This build just makes them do more work.