AI route planning · Process platform
The app that is not an app.
Mobile teams receive their routes in their messenger, in their own language. Photos as evidence, questions as conversations, while dispatch keeps track.
From day-to-day operations
Mobile teams need their assignments and feedback directly on site. Cleaners and mobile service staff have no desk, no company laptop and often do not speak the language of head office.
This process platform grew out of day-to-day operations at a nationwide cleaning service provider and reverses the direction: the route arrives as a message in the messenger the employee already uses on their smartphone, in the language they speak. Questions become conversations; evidence takes the form of photos with time and location stamps. Feedback and photos are assigned to the case through a guided workflow. Text messages can be translated for head office; replies use the same messenger, via a Telegram bot or WhatsApp bot. The web interface is built for dispatching and controlling, for the roles that already work at a computer. The platform is AI- and MCP-enabled: dispatchers and controlling staff can also operate it from AI assistants via the Model Context Protocol.
Three reasons why the app stays empty
Attempts to digitalise mobile teams rarely fail on business requirements. They fail on three barriers that exist before the first line of code.
No device, no desk, no patience: Work happens on site, not at a screen. Access is through a Telegram bot or WhatsApp bot and the mini app embedded in it.
Language is the real interface: Mobile teams speak many languages, and head office speaks none of them fully. A German-only interface shifts the problem to the supervisor, who has to translate everything.
Evidence comes too late: Clients require proof of services delivered. If evidence is only recorded from memory in the evening, it is a record of a record rather than reliable evidence.
Four steps, no installation
Employees do not need to install an app or create a separate account. The process starts and ends in the messenger:
The route arrives: Sites, sequence, time slots and tasks are sent as a message in the employee's stored language.
Arrival is recorded on site: A QR code at the site is scanned using the live camera in a lightweight mini app directly in the messenger. QR check-in documents registration at the site.
The evidence is a photo: The mini app captures photos with a server-side timestamp and, where location permission has been granted, the device's location. Missing or unusual location details are flagged for review. Unusual images are flagged; images assessed as faulty can trigger a review ticket.
The question is a conversation: Anyone who writes receives a reply in the same channel in their own language. When someone reports a defect or an additional service, it becomes a structured case for head office.
Evidence, not surveillance
Attendance records and photographic evidence serve the same purpose as a signature on a traditional job sheet, but generate digital data. Their purpose is therefore clearly limited.
QR check-in confirms that a service worker was at the site at a particular time, not where that person was before or afterwards. The evidence photo documents the condition at the site and the time of capture, not an individual's behaviour. Both serve to support acceptance of services and billing to the client.
Employee data protection and co-determination issues are clarified with employee representatives and data protection officers before introduction.
Image analysis and guided workflows
Image analysis checks evidence photos for anomalies in the background and supports preliminary assessment for the back office. Unusual images are flagged; images assessed as faulty can trigger a review ticket for dispatchers.
This works because the underlying back-office data is cleanly structured: work sites, locations, time slots, roles and qualifications are clearly recorded. The operational decision remains with the human dispatcher: reported defects and anomalies are presented as tickets and assessed there.
Features
Built and proven in live operation.
The platform is in day-to-day use at a nationwide cleaning service provider.
Telegram bot / WhatsApp botMultilingual tasks, questions and status messages via Telegram or WhatsApp in 13 bot languages, with German as the fallback language.
QR check-in & mini appOn-site recording via the Telegram Mini App with mandatory live camera use, without a separate app installation.
Evidence photos with verification stampsThe mini app captures photos with a server-side timestamp and device location where location permission is granted; unusual details are flagged.
AI anomaly detectionAutomatic preliminary assessment of after-service photos in the background; images assessed as faulty can trigger a review ticket.
Automatic route suggestionsAutomatic route suggestions with transparent assessments based on geography, qualifications, regular teams and priorities.
Controlling & proof of serviceWeekly service records per customer and cost centre, planned/actual comparisons and net working time calculation.
Role model & multi-tenancyRole-based access control (RBAC) for dispatchers, controlling staff, admins and supervisors, plus separate tenants with their own branding.
Change logSelected changes and sign-in events are logged in the system.
Target sectors
Where the pattern fits.
Potential areas for adapting the platform:
Building cleaningThe platform's starting point: fixed sites, time slots, multilingual cleaning teams, photographic evidence and defect reports.
Mobile care servicesFor mobile care services, we are examining routes, team coordination and visit documentation as a use case.
Facility services & tradesFor cleaning and facility services, we are examining assignments, photos and feedback on site.
Cross-regional team coordinationFor cross-regional team coordination, we are examining central dispatching and role-based controlling across several locations.
