Blind mid-trip
An agency took full responsibility for a group trip the second a booking was confirmed. Once the trip actually left, that responsibility went blind. No system showed trip status, itinerary progress, or where the crew was. Every update travelled through a manual call chain, agency calls driver, driver calls guide, guide updates passengers, repeated across every trip running at once.
The visible symptom was wasted hours and missed updates. The real problem sat underneath it. Every role was working off its own version of the truth, updated at its own pace, with nothing connecting them.
Filter, then choose
Assigning a vehicle, driver, and guide to a trip meant manually cross-referencing capacity, license type, and compliance rules. High cognitive load, and violations that surfaced mid-trip when nothing could be fixed. I rejected full automation, since exceptions like exemption letters need a human call. I rejected staying fully manual too, since that just kept the same overload. The system now filters down to eligible options before an admin sees them. The admin picks from that shortlist.
Worth saying plainly. What shipped in Phase 1 is fully manual selection, without this filtering logic. Time and budget cut it. The guided version is designed and documented, not yet built.
One shared thread
Once a trip departed, updates moved through phone calls and personal WhatsApp threads, fragmented, private, and visible to no one but the two people on the call. I considered one-to-one chats between users and rejected it, since a private thread just rebuilds the same siloed problem. Every trip gets one shared channel instead. Agency, driver, guide, and passenger read the same thread, tied to that trip's ID.
One shared truth
Separate apps connected by an API looked like the cheaper option given the budget. I rejected it. Every API exchange is a place for the data to drift out of sync, which is exactly how a driver ends up at a location an admin already changed. A single system with one shared data layer removes that failure by construction, instead of trying to sync it away after the fact.
What shipped
Stakeholders reviewed and signed off the design with no major rework requested. GoWheels hasn't gone live. Rollout is on hold on a payment gateway decision the client hasn't finalized, which sits outside what I designed.
Approval tells me the design made sense on paper. It doesn't tell me whether the trip chat actually cuts WhatsApp reliance once people are mid-trip and under pressure, or whether manual verification holds up at higher volume. I've never seen this used by a real agency, driver, or guide under real conditions.