{"id":9207,"date":"2024-08-16T15:10:54","date_gmt":"2024-08-16T12:10:54","guid":{"rendered":"https:\/\/tehisintellekt.ee\/ai-assistendi-arendus-muutis-iizi-kindlustusspetsialistide-too-efektiivsemaks\/"},"modified":"2026-02-06T10:56:22","modified_gmt":"2026-02-06T07:56:22","slug":"ai-assistant-development-made-iizi-insurance-specialists-work-more-efficient","status":"publish","type":"post","link":"https:\/\/tehisintellekt.ee\/en\/ai-assistant-development-made-iizi-insurance-specialists-work-more-efficient\/","title":{"rendered":"AI assistant development made IIZI insurance specialists’ work more efficient"},"content":{"rendered":"\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t
<\/div>\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Image generated by Ideogram.ai<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t\t
<\/div>\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

IIZI Insurance Broker is one of the leading companies in the Estonian insurance sector. The company has 11 offices across Estonia and more than 140,000 active private clients. <\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

To efficiently serve their large customer base, the company has always been open to modern technologies. This year, they took the first steps in adopting artificial intelligence to make the work of insurance specialists faster, more accurate, and more efficient. What exactly was done, and what is the outcome of the project? <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

Employee time and knowledge are limited<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

The daily work of IIZI’s insurance specialists (about 30-40 people) involves identifying the client’s insurance interests and needs and advising them on the right choice.<\/p>

When a client asks about insurance terms (e.g., “Is the repair cost of a robotic lawnmower covered?”), the specialist has three ways to answer: 1. use previous knowledge, 2. search for the answer in the insurance terms, or 3. ask a colleague for help. <\/p>

All of these options have certain drawbacks:<\/p>

  1. A specialist’s knowledge is always limited, and memory can fail when insurance terms change.<\/li>
  2. Searching for answers in insurance conditions is very time-consuming, especially when comparing terms from several insurance providers. <\/li>
  3. Asking a focused colleague for help interrupts their train of thought and wastes their valuable working time.<\/li><\/ol>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
    \n\t\t\t\t
    \n\t\t\t\t\t

    Achieving high-quality information search takes work<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
    \n\t\t\t\t
    \n\t\t\t\t\t\t\t\t\t

    We decided to create an AI assistant that would analyze and compare the terms of insurance companies and find the information the client needs.
    First, we had to process the source data. This meant that long insurance policy documents had to be broken down into smaller logical parts. Several factors made this task difficult: <\/p>

    1. Insurance companies use different formats to describe their terms.<\/li>
    2. Descriptions of coverage and exclusions can be located in different parts of the document.<\/li>
    3. One clause may refer back or forward to other points in the terms.<\/li>
    4. Terms are usually in PDF format, and some data may be lost during automated reading.<\/li><\/ol>

      Two key points became essential for ensuring high-quality search results:<\/p>

      1. Smooth cooperation with IIZI specialists, who helped define the logic for dividing documents into sections;<\/li>
      2. An efficient search system using semantic hybrid search and neural re-ranking to find information segments most likely to contain the required answers.<\/li><\/ol>

        The second central issue of the project concerned the development of comparison functionality. In the initial testing phase, the AI assistant answered questions about only one insurance company at a time. The user had to ask questions like “Does ABC cover robotic lawnmower repairs?”, where ABC referred to a specific company.

        Eventually, we developed a comparison solution where the AI assistant provides answers for multiple insurance companies at once. To achieve this, we created a workflow where the specialist asks a question and selects the companies they want answers for. The AI assistant then generates an answer for each company separately and finally combines them into a single response, which is delivered via the chat window. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

        \n\t\t\t\t\t\t\t
        <\/div>\n\t\t\t\t\t\t\t
        \n\t\t\t\t\t
        \n\t\t\t
        \n\t\t\t\t\t\t
        \n\t\t\t\t
        \n\t\t\t\t\t
        \r\n