Understand the experience before deciding on the interface.
I want to understand what customers are trying to achieve, what they know and where the service makes things difficult. This gives the team a basis for deciding what needs to change.
Make the journey something the team can discuss
I use task flows, customer journeys and experience requirements to work through decisions with product and engineering. This includes the parts that happen outside the interface, such as a review by a specialist or a conversation with an adviser.
The diagrams help us discuss priorities and technical constraints before those decisions become screens. I try to involve the people responsible for delivery throughout the work.
Choose research around the decision
I define what we need to learn and use that to plan the study and prototype. I've used moderated sessions to examine complex quoting journeys and unmoderated testing to evaluate an AI marketplace. In both cases, observing people changed parts of the experience I had designed.
Sometimes the recommendation is to stop. On an SME project, evaluating a working concept showed that the proposed tool needed a level of digital maturity our customers did not yet have. The business discontinued the initiative after reviewing our findings.
Use AI to support the existing process
AI-assisted prototyping helps me turn an agreed structure into a functional experience people can test. On the marketplace project, this supported more time for task flows, experience definition and stakeholder alignment before moving into interface work.
The user-centred design process stayed the same. Faster prototype production helped us resist pressure to produce screens before we had worked through the experience. My team has begun applying elements of this approach on other projects.
Be clear about what the evidence supports
I separate what we observed from what we recommend and what has been delivered. A revised design informed by research still needs evaluation. Feedback from a project team is useful, but it is different from an outcome I measured myself.