Security guard company · Canada · 60+ client sites
15 minutes of work. Done in 30 seconds. 60 times a day.
A security company's guards report their whole shift in WhatsApp groups: patrol check-ins, crews arriving and leaving, vehicles, alarms, incidents and hundreds of photos. Every shift, someone in the office turned that into a professional client report by hand. We built the tool that does it in about 30 seconds, and a person approves every one.
Press generate to watch it read a real night shift
Site 14 · Night shift group
WhatsApp export · 6:00 PM to 6:00 AM · guard, office, monitoring
What the guards send
Client letterhead
Daily Security Patrolling Log
Site 14 · Construction · Night shift 18:00 to 06:00
0
events found
0
messages excluded
0
photos archived
The tool only ever creates drafts. A reviewer approves each one.
per report by hand, before
per report now
reports a day
of office time back, every day
The problem
Sixty reports a day, every one written by hand.
Each client site has a WhatsApp group. Through a shift the guard posts patrol updates, who is on site, contractors arriving and leaving, vehicle movements, alarms, damage and photos. A busy construction site produces 600+ photos a night. Office staff ask questions in the same group.
Clients expect a daily report per shift in a consistent format, hour by hour, with the real events called out and photo evidence available. Building one meant reading the entire chat, separating real events from routine check-ins and office chatter, writing every entry in the right tone, and saving every photo. Ten to fifteen minutes per report, often longer for the photos. At about 60 uploads a day, that is a full-time job that nobody was hired for.
Before and after
What a report used to cost. What it costs now.
Before
- Read the whole chat, hundreds of messages and photos, and pick out what actually happened
- Separate guard updates from office questions and monitoring check-ins
- Write every half-hour entry by hand in the client's format and tone
- Save and sort every photo into the right folder
- 10 to 15 minutes per report. Photos often took longer than the report
After
- Upload the export and pick the site
- Read the finished report, every event already in the right slot
- Fix anything, then click Approve
- Email draft and branded PDF created, photos filed, automatically
- About 30 seconds per report
At 60 reports a day, that is roughly 12 hours of office time returned every day. At a $25-an-hour office wage, about $6,000 a month. The AI cost per report is around five cents.
What we built
Upload the chat. Get a finished report.
Six steps, one of them human.
Upload
Staff drop the WhatsApp chat export (a .zip with photos, any size) and pick the site. Photos go straight to cloud storage from the browser, so a 600-photo night uploads in seconds instead of being saved one by one.
Sort into shifts
Every message is read, timestamps are converted to the site's timezone, and messages land in the right shift. Overnight shifts, weekday-specific hours and 24-hour sites included.
Classify
AI separates the guard from office staff, the monitoring station and contractors, then tells routine patrol check-ins apart from real events: incidents, equipment faults, crews on and off site, vehicles.
Write
Every half-hour of the shift is filled. Quiet periods use the site's own approved wording, rephrased so reports do not read copy-pasted. Real events go into the exact slot where they happened, highlighted for the reviewer.
Review and approve
A person checks the report in a Word-like editor and clicks Approve. That creates an email draft to the client (never auto-sent) and files a branded PDF to the company's Drive.
Archive photos
Every photo is filed under Site / Year / Month / Day, in the order the guard sent it. No more dragging photos into folders.
Why it holds up
It understands how guards actually write.
Guards write the way people text: fast, informal, full of photos. The hard part was never generating text. It was understanding exactly what happened, when, and who said it, then producing something a client can trust.
- A guard's short reply "Not working" to an office question about the light tower is recorded as "light tower not working".
- Uses the time written in a message ("At 11:10 pm saw two people") over the time it was sent, so events land in the right hour.
- Separates the guard from office staff, monitoring stations and contractors in the same group.
- Handles iPhone and Android exports, 12 and 24-hour clocks, multi-line and deleted messages, duplicates and system notices.
- Overnight shifts, per-weekday hours and 24-hour sites merge into one correct report.
- Rephrases, never invents. Uncertain items are flagged for the reviewer. Checked against a real two-night chat of 1,315 messages: every real event landed in the correct time slot.
A real night, anonymised
661 messages. 608 photos. One report.
At a construction site, a guard's 14-hour night shift produced 661 messages, 608 of them photos. The tool:
- identified the guard and ignored the monitoring station's check-ins;
- logged every contractor crew arriving and leaving by company and name, in the right half-hour;
- captured a small fire the guard found and put out in the back alley, at the time stated in his incident report;
- filled every other half-hour with site-appropriate patrol entries;
- archived all 608 photos in order.
Time for a person to write that report by hand: the better part of an hour, once the photos were sorted. Time for the tool: under a minute.
Keeping a human in the loop for every report was a design choice, not a limitation.
B4hours, on the build
How custom automations workHave a workflow like this?
If your business runs on chats, emails or spreadsheets that a person has to turn into something else, we can probably automate it. Tell us the workflow and we will say yes, no, or partly.
30-day money-back guarantee · Live in one week · Toronto & GTA