01
The first problem was product shape.
A promise to help with business is too broad to make a useful interface. The customer needs to recognize their situation, understand what information is required and know what happens after they ask for help.
I split the experience into two clear lanes: people starting or growing a business, and operators trying to manage an existing one. Both routes lead to a scoped request rather than an open-ended chatbot.
02
AI prepares the case. An expert validates the recommendation.
The AI layer can collect context, structure the problem and prepare a first recommendation. It should not make regulated, contractual or high-consequence advice look final when an expert has not checked it.
The journey therefore shows the validation step as part of the service. The human is not hidden behind the interface, and the customer can see when a recommendation is prepared versus accepted.
03
Design the service after the answer, too.
A useful concept cannot stop at a conversational result. The customer also needs a quote, a payment step, a place for documents, a view of fulfilment and a clear sense of progress.
The 16-screen mobile concept connected those stages into one journey so the commercial and operational work was visible, not left as an undefined handoff behind the interface.
04
The prototype tests comprehension, not demand.
The design can test whether a customer understands the two lanes, the role of AI and the handoff to an expert. It can also expose missing states before engineering begins.
It cannot prove that customers will buy, that experts can fulfil the service profitably or that integrations will work. Those questions require a real pilot and operating evidence.