Hotel call center agents were managing complex guest interactions while juggling 4–5 separate tools — booking platforms, loyalty systems, CRM, case management, and performance dashboards — all open at once, none of them talking to each other.
The result was high cognitive load, slow resolution times, and agents who knew the guest's name but couldn't surface their last three stays, open complaints, or loyalty status without tabbing through three different apps.
We conducted contextual inquiry with call center agents — standing beside them during live calls, watching real interactions in real time. The workarounds agents had invented told us more than any survey could have.
"The workarounds agents used told us more than any interview could."
Observing agents mid-call surfaced friction that would never appear in a structured session. We saw where systems broke down, what shortcuts agents invented, and where genuine failures occurred under pressure.
Different agents, different contexts, the same problems. These three themes surfaced repeatedly and shaped every design decision that followed.
Agents rely on manual workarounds to complete tasks that should be automatic. Every extra step is time taken away from the guest.
Customer information is scattered across systems with no consistent link between them. A full picture requires visiting multiple tools.
When a situation escalates or an exception arises, agents are on their own. No guidance. No context. Just memory and instinct.
We reframed away from "how do we improve each tool?" and toward "how might we design an experience that makes the right information available at the right moment?"
The brief called for a better interface. But the research made clear that better UI on top of fragmented data wasn't the answer. We had to rethink the information model before we could design anything meaningful.
Every project has moments where a decision constrains everything that follows. These are the ones that mattered on this one.
Agents didn't want more information — they wanted the right information at the right moment. AI had to earn its place on screen.
The biggest lesson was restraint. In AI design the instinct is always to show more — more suggestions, more data, more visibility. But in a high-pressure service environment, more is noise. The real work was figuring out what not to surface, and when to surface nothing at all.
The full case study includes the actual screens, research synthesis, and detailed iteration history. Password-protected for client confidentiality.