“I’m looking for an electric car.” - “There are four of us and we often drive long distances.” - “And the dog has to come along.” - “Actually, I’d just like a test drive first.”
This is how people talk about a purchase decision. Not one of these sentences fits into a filter menu.
Two languages, one problem
Variant-rich products - vehicles, machinery, insurance, software packages - force customers to learn the provider’s logic before they can decide. Those who don’t want to drop out or pick up the phone.
Expectations come from elsewhere
Streaming, retail and banking have set how digital decisions should feel. Three expectations come up again and again in our projects:
This is the key point: guided advice does not replace expert mode. Some people want to be asked, others want to see every option. The experience has to do both - and let people switch between them.
What AI changes
For the first time, language models can reliably translate between the two languages. “There are four of us and the dog has to come” becomes body style, boot volume and range - and from that, a short list of matching variants.
For this to work, three things are needed:
- Real product data instead of general knowledge. Models, prices, rules and availability come from the provider’s systems, not from the model’s training.
- An interface, not just text. Comparisons, images and next steps belong on screen - the conversation leads, the interface shows.
- A handover into existing processes. From the conversation straight into configuration, quote or appointment - with everything already said.
What this looks like is shown in our references Advising where customers ask today and The assistant that knows every price list. The assistant on this website follows the same principle: it answers with tiles, not walls of text.
Who benefits
Wherever the offering is complex and the decision expensive: vehicles, capital goods, financial and insurance products, B2B software. Success is measured not in clicks but in fewer drop-offs and more qualified enquiries.
