AI Assistant: secure GenAI chatbot for PZU employees Qorus-NTT DATA Innovation in Insurance Awards 2025

The AI Assistant project was initiated to create a chatbot based on generative artificial intelligence, aimed at supporting employees in their daily tasks. This initiative is a direct response to our employees' expectations and reflects PZU's commitment to equipping our workforce with the latest tools. The primary goal was to ensure the highest level of security and control over user-provided data, which is crucial for a financial institution like PZU.

Recognizing the potential of such tools in our employees' work, we understood that the nature of tasks in the insurance industry involves handling a large amount of confidential information, making it impossible to use publicly available tools. Therefore, the tool was designed to enable employees to use modern technologies in a safe and compliant manner. Additionally, we aimed to explore applications and enhance employees' skills in using artificial intelligence in their daily duties. The pilot also aimed to gather experiences to implement such solutions on a large scale across the organization.

Shortly after the launch of ChatGPT at the end of 2022, PZU recognized the need to create its own conversational tool. Publicly available generative AI tools at the time posed several risks that were unacceptable for a financial institution. Therefore, PZU decided to develop an internal solution that met the required legal standards for data protection and information security.

The AI Assistant project was carried out by GPT Lab, which started its operations at PZU in June 2023. The tool was tested by 1,000 employees from all divisions of the organization, allowing for verification of its effectiveness, response quality, and ease of use. An "unintended consequence" of creating the tester community was the formation of an informal circle of AI enthusiasts and Ambassadors, who are local leaders in their departments, sharing knowledge and being the most active users. The GPT Lab team prepared a special Teams channel for the community, which became a place for discussions, inspiration, and problem-solving. Educational materials on Gen AI and prompt engineering, such as instructions, webinars, and an educational website on the intranet, were used by the team not only in contact with testers.

Main results so far:

1. Education and training: Thanks to the knowledge gained from the pilot and training employees who use AI more frequently, we are currently rolling out the generative artificial intelligence solution across the entire organization: 15,000 employees with a well-thought-out educational program in this area.

2. Legal compliance preparation: By implementing the solution and starting educational activities, gaining shared experiences among employees, and consulting with the legal, security, and compliance departments, our ability to ensure compliance with AI Literacy requirements resulting from the AI Act is much easier to achieve.

3. Data security and control: The tool ensures the highest level of security and control over PZU's data, which in 2023 was difficult to achieve with publicly available tools.

4. Employee support: The AI Assistant supports employees in their daily work, enabling, among other things, summarizing content, analyzing provided materials, and answering questions.

5. Work optimization: We estimate that the tool saves approximately 144,000 hours annually by optimizing employees' workflow and reducing the time spent on repetitive tasks (estimate for roll-out based on pilot results).

6. Cost reduction: Automating routine tasks helps lower operational costs, including reducing the absorption of full-time equivalents (FTEs), thereby freeing up people's potential to perform more creative work rather than routine and repetitive tasks.

7. Operational risk reduction: The secure and controlled environment of the AI Assistant ensures compliance with regulatory standards, thereby reducing operational risk.

The PZU AI Assistant project is an example of an innovative approach to using generative artificial intelligence in a safe and compliant manner, contributing to increased efficiency and safety in the organization, as well as comprehensive enhancement of employees' skills in using artificial intelligence.