Front is a customer operations platform used by thousands of support and operations teams. Chat lives directly in our customers’ products and websites, the experience their end users interact with every day.
Chat had been part of Front for years, but it no longer worked for modern support teams. Built before AI changed what was possible, it didn’t match customer expectations, or how support leaders wanted to operate.
I led product design on the end-user experience, working closely with another designer who owned the admin setup.
The vision: resolve requests in real time with AI, and hand off to humans only when needed.
Previous experience
Chat adoption was low. Most teams couldn't staff a live chat team, and rigid bot flows created an experience that couldn't adapt to real customer needs.
The product assumed a 24/7 live chat team most companies didn't have.
Rigid, pre-mapped bot flows made it inflexible for customers and exhausting to maintain for teams.
From unstructured text to structured UI. We consistently observed how text-only AI interfaces added cognitive load.
We explored how to make the interaction more guided and less reliant on freeform input. We explored how to make the interaction more guided and less reliant on freeform input. The information being collected was defined in natural language by the admin setting up the widget and could cover any topic or scenario, so the pattern had to be scalable and adaptable.
We landed on focusing structured UI at the point of input, where customers were already entering information, while still giving the flexibility to type freely and change direction at any point.
Pain points of live chat
We iterated on language and placement to set clear expectations
Instead of keeping customers stuck in a live queue, we set clear response-time expectations and let them continue the conversation via email or SMS. When a human got involved, the full context became a shared ticket that teams could collaborate on - critical for complex B2B workflows.
We designed a setup flow that got teams testing in minutes
Guided logic for specific scenarios could be layered in as needed, and the whole experience was built on top of Front's AI pipeline so it scaled as the underlying models improved.
Customizing the widget to fit any product experience
We prioritised flexibility so teams could choose their form factor and define their own style. I built a system that generated a palette from their brand colour and applied it intentionally across all widget elements.
