In designing the Turo support workspace, we put the vehicle, booking status, trip details and the customer conversation together. Pickup and return questions often need that context; anything involving charges, disputes or exceptions needs an operator’s judgment. The two kinds of work call for different handling, and a support system has to make that difference visible.
Even standard questions start by checking this trip
“How do I pick up the car?” and “I’m here now, how do I pick up the car?” look alike, but the information needed to answer may differ. Before shaping a reply, you need to know which car and which booking the customer means, and what stage they’re at. Putting this next to the conversation saves the person handling it several trips back and forth.
Templates suit stable guidance, such as pickup and return steps or the photos required. The actual reply still needs checking against the trip; a familiar question doesn’t justify pasting a stock answer.
When judgment is needed, a person takes over
Whether a charge is reasonable, whether a condition record is disputed, how an exception should be handled: these questions involve responsibility and specific facts. Being able to generate a fluent explanation doesn’t mean the system has grounds to make the decision.
So the workspace separates standard replies from matters a person handles. For the latter, the design focuses on keeping the question and the information already gathered, so the operator can verify it and then decide how to respond. A natural tone helps communication, but it can’t replace authority and judgment.
The handover is a process that needs designing too
When designing the handover, we focused on what the person taking over needs to see: the trip in question, what the customer asked, what is already known and what still needs checking. If all of that has to be gathered again after the handover, the work has simply restarted with a different person.
The current demo uses simulated trips and template replies, so it can’t show real service efficiency yet. When it’s connected to a real business, the things worth checking are: is the context complete, is it easy for a person to take over, and are unverified matters being mistaken for conclusions?
We want a support system that handles the repetitive work and also puts the matters that need judgment, with the necessary information, into the right person’s hands.
Related project
This article is based on the design of an existing project.See the AI Support Agent case study, interface and notes ↗