Chatbots that belong to the brand.

Discovery

We analyse the data the chatbot should know, the questions it should answer, and the lines it shouldn't cross. What's allowed and what isn't. The output is a short scope brief the client agrees to before any build work starts.

Persona

Name, voice, tone and refusal rules. Evín on amka.is is a wine advisor, not a product search box, so she politely declines questions outside that scope. The system prompt reads like a brief for a new hire, not a regex.

Answers grounded in your data

The chatbot looks things up in your own data, product catalogue, PDFs, help pages, whatever you want it to know, before it replies. When someone asks a question it first finds the relevant passage in your own records, then cites the source directly, the way a staff member checks the manual before answering. No made-up facts, only real content from your own data. The technique is called RAG (Retrieval-Augmented Generation) in technical terms.

Integration

The chatbot lives inside your site, same colours, same type, same tone. Streaming responses, fallback to email when Claude isn't sure, conversation analytics, and a prompt-tuning panel for staff so you can adjust without redeploying.