Saxana is an AI chatbot that replaces traditional multi-step wizard forms with a natural conversation. Instead of clicking through dozens of screens and fields, the customer simply describes their situation — the AI guides the dialogue, collects all necessary information, answers questions, and extracts data from uploaded documents automatically, in any language. The current implementation handles life insurance claim reporting (FNOL) for ČSOB Pojišťovna.
Core competencies
Technologies
Problem definition and goal
Filing an insurance claim after an accident or illness is one of the most stressful interactions a customer has with their insurer — and the traditional approach, a complex wizard form with strict ordering and dozens of mandatory fields, makes it worse. Customers must follow a rigid process regardless of their situation, cannot ask questions, and uploaded documents are not understood by the system: many give up mid-process or end up calling the support line, and data from medical reports gets re-typed by hand.
The goal: let the customer simply describe what happened in their own words while the AI handles the rest — collecting data, answering questions, validating inputs, and reading documents. No fixed step order. Any language.
Insurance claim data is complex — personal details, incident circumstances, medical information, policy data. Collecting all of it conversationally means the AI must keep the dialogue natural and empathetic, handle topic switches, understand vague descriptions, and stay focused — all in whatever language the customer chooses to use.
A conversational interface is bidirectional. Customers don't just answer — they ask: "What documents do I need?", "Will this be covered?", "How long will it take?" The system must switch seamlessly between collecting information and providing knowledgeable answers, without losing track of overall progress.
Customers often have documents — medical reports, discharge summaries, ID cards — that already contain much of the needed information. The system must read these documents from photos, scans, or PDFs, understand their content, and extract the relevant fields automatically, so the customer doesn't have to type what is already on paper.
Saxana organizes data collection into logical topics — for example "Policy information", "Insured person", "Accident details". The AI focuses on one topic at a time, asking related questions together rather than jumping between unrelated areas; when the customer mentions something from a different topic, the system accepts it flexibly. A visual progress indicator shows which areas are covered and which remain, giving both the customer and any assisting agent a clear overview at any point.
Saxana is not a generic chatbot. For life insurance it knows the different claim types — accident, hospitalization, illness, death — the specific data required for each, and the business rules that apply, so it validates inputs the way a trained professional would. It also has access to a knowledge base built from the actual insurance policy documents: when a customer asks about coverage, exclusions, required documents, or timelines, the system provides accurate, policy-grounded answers.
When a customer uploads a document — a photo of a medical report, a scanned PDF, an ID card — the system reads it using specialized AI vision models, extracts the relevant data, and fills in the corresponding fields automatically. What can be extracted from a document doesn't need to be asked. Extracted data is transparently shown to the customer and can be confirmed or corrected during the conversation.
While the current implementation focuses on life insurance claims, the underlying platform is domain-agnostic. Any process that collects structured data through interaction — car insurance claims, property damage reports, loan applications, customer onboarding — can be built on top of Saxana by defining the target data model and the domain-specific knowledge.
We help individuals grow and companies deliver quality technology solutions.