Back to AI LabTry it live (opens in a new tab)

AI Lab · 02

alligo. Right there with you.

A roadside-assistance platform that insurers offer their policyholders: the driver asks for help from her phone and an agent at the insurer coordinates it from a console. Both sides of one case, designed for a moment of peak stress.

alligo insurer console: requests sorted by most urgent, with an urgent notice from the driver at the topalligo driver app: help on the way, with the tow’s arrival time, driver and plate
Role

Product Builder · strategy, UX/UI, brand, design system and the build

Platform

Driver app + insurer console, sharing one case

Built with

Claude, React 19, TypeScript, Vite, Tailwind, Leaflet, OpenStreetMap

Status

Live: driver app and insurer console

Context

Someone stopped on the hard shoulder.

The challenge was to design both sides of a single case for a moment of peak stress. Three people need different things from it.

How it’s built

A working product, not a mock-up.

The full journey works across two tabs: a pure, tested domain (state machine, deadlines, notices, urgency), a data layer that handles latency, failures and duplicate sends, and real services: browser GPS, OpenStreetMap addresses, Leaflet maps, notifications and an installable app.

Unit tests
29
End-to-end journey tests
21
Design decisions documented
40+

Two products, one case

Same case. Two sides.

A first approach with a role switcher felt artificial; two products, each with its own entry point, look like what would really exist. A single state machine decides which change is valid and who can make it.

alligo: lifecycle of a roadside-assistance caseSwimlane diagram. The driver first answers whether anyone is hurt; yes calls 112. Otherwise she sends a four-step request, queued if there is no signal. The case moves from received to reviewing, to waiting for an answer and back when the agent asks for a detail, then assigned, en route and completed; the driver can withdraw or the agent cancel from any open status. The agent starts the review within 5 minutes, assigns a provider within 15 and marks it on the way within 10. Exceptions: no provider yet, the tow running late, and driver notices the agent handles.ALLIGO · CASE LIFECYCLEOne case, two sides.How a request for help moves between the driver app and the insurer console.Status changeAction that moves the caseWhen something goes wrongEmergencyWhat the driver seesDRIVERapp on her phoneCASEone state machineAGENTinsurer consoleSix statuses, plus cancelled.A single state machine decides whichchange is valid and who can make it.Cancelled: the driver withdraws or theagent cancels, from any open status.ReceivedReviewingWaiting for an answerAssignedEn routeCompletedCancelledAnyone hurtor in danger?Call 112 · one tapYes4-step requestGPS location, whathappened, 2 photosNoSendqueued withno signalAnswers one questionTrackingname, plate, mapand countdownRates the helptwo questionsNOTICESends a notice, never editsurgent if the tow is on its wayA deadline for every step.Past it, “Overdue” shows on both sides,so driver and agent tell the same story.Starts the reviewwithin 5 minAsks forone detailAssigns a providerwithin 15 minMarks it on the waywithin 10 minCompletesthe serviceEXCEPTIONNo provider yetdriver sees “Still looking for a tow”;assign as soon as one is foundEXCEPTIONTow running latedriver sees “Running late”;the agent sends the new timeNOTICEHandles the noticemark as read, editdetails or callThe inbox puts unread urgent notices first, then what depends on the agent, then what waits on the driver.An expired policy warns, but never blocks a request for help.

alligo · case lifecycle

One case, two sides.

  1. DriverAnyone hurt or in danger?
  2. Yes Call 112 · one tap
  3. No 4-step request: GPS location, what happened, 2 photos
  4. Send queued with no signal
  1. Received

    AgentStarts the reviewwithin 5 min

  2. Reviewing

    AgentAsks for one detail

    Waiting for an answer

    DriverAnswers, and it returns to Reviewing

    ExceptionNo provider yetDriver sees “Still looking for a tow”; assign as soon as one is found.

    AgentAssigns a providerwithin 15 min

  3. Assigned

    AgentMarks it on the waywithin 10 min

  4. En route
    ExceptionTow running lateDriver sees “Running late”; the agent sends the new time.
    NoticeA notice, never an editUrgent while the tow is on its way; the agent marks it as read, edits the details or calls.

    Driver seesTracking: name, plate, map and countdown

    AgentCompletes the service

  5. Completed

    DriverRates the helptwo questions

CancelledThe driver withdraws or the agent cancels, from any open status.

Past a deadline, “Overdue” shows on both sides. The inbox puts unread urgent notices first, then what depends on the agent, then what waits on the driver. An expired policy warns, but never blocks a request for help.

One case, two sides: the statuses, who moves them, and what happens when something goes wrong.
A driver beside her broken-down car on the hard shoulder, opening alligo on her phoneDriver app: the first screen asks whether anyone is hurt, with 112 one tap away

Driver app

María, a policyholder stopped on the hard shoulder

“Is help coming, and when?”

Device
Phone, one hand, installable
Sign-in
None: her insurer’s link identifies her
An insurer’s assistance agent at his desk reviewing a request in the alligo consoleInsurer console: a case with an urgent notice, the next step and the timeline

Insurer console

Javier, an assistance agent at Solana Mutual

“What do I handle first, and can I decide without calling?”

Device
Desktop and mobile
Sign-in
Work account

Key design decisions

Less reading, fewer calls.

Driver app, first screen: “Is anyone hurt or in immediate danger?” with “Yes · Call 112” firstDriver app: the request reviewed before sending, in four steps

Also: sending with no signal (the request queues and sends itself, without duplicates) · two suggested photos, none required · child seats asked only when children are ticked · a “Next step” block with only the valid actions · a two-question rating at the end.

Hypotheses and next steps

Built to be tested.

Several decisions rest on assumptions to validate with the insurer and users before scaling it: the tow cab’s passenger space, the review, assign and depart deadlines, and the child-seat rule for taxis. Next: test with 5 drivers in a simulated breakdown and 3 agents on “Most urgent”, and measure time to assignment, clarifications, notices per case and rating.

Behind the work

Want the walkthrough?

alligo is live, so you can try both sides yourself. I’d also be happy to walk you through the product decisions, the build with AI and the brand in an interview.

Let’s talk