The challenge
The retailer sold a technical, configurable product range where the right answer depended heavily on the customer's situation. The legacy webshop did not capture that nuance. Customers either left without buying or contacted the small advisory team for clarification, which created a queue, slowed sales and made staff costs the binding constraint on growth. Add-to-cart conversion sat below industry benchmarks and the team felt that scaling marketing spend would only make the queue worse.
Leadership was clear that they did not want a generic shop with a chatbot bolted on. They wanted a platform where AI sat inside the buying journey itself — helping customers reach a correct, confident specification without always needing a human, while keeping every recommendation transparent and traceable.
Our approach
We delivered the project through our e-commerce development and custom SaaS development practices, with AI capabilities scoped by our AI integration consulting team. The platform was rebuilt on a modern headless commerce stack and a Swiss-hosted application layer so the retailer kept full control over data, performance and integrations with their ERP.
The AI work focused on three concrete moments in the journey. First, a guided specification flow used a constrained model to translate plain-language questions into the correct technical configuration, citing the source rule for every answer. Second, dynamic pricing logic combined catalogue rules and contextual factors so the displayed price was always defensible and the team could trace any deviation. Third, an in-context advisory dialogue allowed customers to ask follow-up questions on a chosen configuration, with the model strictly grounded in the product knowledge base and quietly escalating to the human team when confidence dropped.
Every AI surface shipped with logging, evaluation harnesses and a kill switch. No recommendation reached a customer without an audit trail.
Outcome
The relaunch shifted the unit economics of the business. Conversion on the most configured product lines rose meaningfully against the baseline, and a measurable share of advisory questions were resolved without involving the human team. The team reported that the work that did reach them was higher quality — deeper questions, warmer customers — rather than basic clarification.
The platform is now the channel through which the retailer absorbs growth without proportional staff cost. Future iterations are planned around personalisation and post-purchase support.
Stack used
Duration
approximately 6 months