Artificial intelligence for the insurance broker

The problem

An insurance broker’s work is largely made of documents: offers, policies, conditions, correspondence with the companies. The information needed is there, but it sits inside texts written by different senders in different formats, and to compare it someone has to read it, extract it and bring it into a common form. The same goes for requests for offers: each company receives and returns information in its own way. The time goes into transcription and standardisation, not into advice.

The solution

I set up a project dedicated to introducing artificial intelligence into these processes, as research and experimentation before anything like a finished product. The areas I work on are document analysis, information extraction from documents, offer comparison, standardisation of the information exchanged between broker and companies, AI-assisted workflows and the automation of requests for offers. Each area is tested before it enters the processes: what works stays, and what does not says something about the next step.

Technologies

The project is experimental by nature and the tools change with the tests; these are the technical areas.

• Document analysis and information extraction
• Offer comparison
• Standardisation of information between broker and companies
• AI-assisted workflows
• Automation of requests for offers

The result

The project is ongoing and I present it for what it is: a path of research, experimentation and digital transformation, not a closed system. The result, today, is that the areas where artificial intelligence can enter the broker’s processes have been identified and tested one by one, and the processes receive assisted steps where the test justifies it.