Staff attendance with a WhatsApp message
A company that works seven days a week coordinated its staff in a WhatsApp group. Today the employee writes “in” or “out” on the same WhatsApp and the system records attendance; the administrator sees everything and has the last word. In daily use since 3 September 2026.
Before
Day and night shifts, weekends, full-time and part-time staff. Attendance, changes and tasks went through a WhatsApp group and people’s memory. Knowing who was in and how many hours had been worked took time and manual counting.
What I did
I kept the tool everyone already used, WhatsApp, and built a system behind it. The employee writes naturally; the AI understands the message and proposes the record; a rules engine checks it; the administrator confirms doubtful cases. The AI never decides on its own.
Then I analysed 36,054 messages from four years of work (1,196 days) to reconstruct how the organisation really works: roles, rules, exceptions. This produced 69 written scenarios describing what the system must do and must never do, and the next modules.
After
- Attendance recorded via WhatsApp on the company number, every day, since 3 September 2026.
- Admin panel with roles and permissions.
- Weekly rotas, planned absences, “who is in today” asked in chat, daily tasks, transcribed voice messages: ready and tested in the lab, released in blocks.
What it can do for your company
It works wherever work is coordinated by chat and nobody has time to learn new software: restaurants, clinics, logistics, maintenance, hotels, field services.
Technical details, for those who want them
- WhatsApp Business API with identity tied to the phone number and signed messages; secure tunnel to the server.
- Architecture: the AI produces a structured proposal, a deterministic engine validates it, the administrator confirms; append-only log.
- Python; 27 versioned schema migrations; about 6,000 automated tests (6,069 at 30/9/2026); 157 revisions in 30 days.
- Separate development, lab and production environments; live rehearsals in the lab before every release.
- Not built yet: automatic rota generation.
The other systems I built.
Google Ads and landing pages: −38% cost per lead
Beforecampaigns run by hand, occasional analysis, a generic landing page.
Aftera new landing page; every day the AI analyses and proposes, changes go live only after approval.
Websites, online shop and digital contracts
Beforeold sites hard to update, a messy catalogue, paper contracts.
Aftera custom platform for several brands, a clean catalogue, contracts signed on a tablet.
Do you have a similar problem?
Tell me in two lines: I will tell you whether the same approach applies to your company and what it would take to start.