AI & innovation
Conversational and agent experience design
Chat assistants and agents are services, and they fail for the same reasons other services fail: unclear purpose, poor handling of errors and no way to reach a human. We design the conversation, the hand-offs and the trust signals, then test them with real users.
Our background in content design, usability testing and service blueprinting is exactly what conversational products need and usually lack.
What we design
- Purpose and scope. What the assistant is for, what it will refuse and how it introduces itself.
- Conversation flows. Prompts, questions, confirmations and summaries written in plain language and tested with users.
- Error and uncertainty handling. What happens when the model is unsure, wrong or out of scope, and how a person takes over.
- Trust signals. Sources, confidence, transparency about AI use and clear routes to a human.
- Measurement. Task completion, containment, satisfaction and quality review, so you know whether it is working after launch.
Evaluation
We run moderated sessions where users complete real tasks with the assistant, and we review transcripts against a quality rubric covering accuracy, tone, safety and recovery. The findings feed directly into the prompts, flows and interface.
Organisations we have done AI and innovation work with
Design an assistant people will trust
Whether you are starting from scratch or fixing one that is live, we can help.