roadsurfer · Self-service · AI · Experimentation
Solving a flexibility problem by building an AI experience customers recognise and use.
The existing chatbot solution limited how much we could tailor the experience. As Product Owner, I led the move towards an in-house n8n-based solution and shaped the experience so customers would recognise it as an advanced assistant rather than another generic support bot.
Problem
We needed both a better foundation and a better customer signal.
The existing solution gave us limited flexibility and little room to use customer context. I led the product work to introduce our own chatbot in the customer account, including authentication so the assistant could access booking-related data. That created the foundation for a more useful, contextual product.
Product decision
Make the AI visible, contextual and easy to start.
I led the product direction for an experience that feels like a modern AI assistant: an exposed prompt bar instead of a hidden entry point, plus suggested prompts based on the customer's current booking status and the page they were visiting. We then A/B tested the exposed prompt bar against a floating action button. The visible prompt bar produced 2.5× more interactions, giving us a clear behavioural signal to scale the direction.
Outcome
Chatbot interactions increased 2.5×.
The product direction increased engagement and helped scale self-service to 44% of customer contacts. The combination of a more flexible technical foundation, contextual prompts and a recognisable AI interaction made the assistant easier to discover and more useful in the moment.