Four events, one goal: Making AI practical
The series built on itself step by step:
- October 10, 2024: How Do I Ask the Right Questions?
- October 17, 2024: How do I use ChatGPT and Copilot for quotes and record-keeping?
- October 24, 2024: Streamlined Claims Processing and Comparison of Policy Terms Using AI
- November 7, 2024: Answers to the questions that arose
Each session lasted one hour. Together, they formed a concise learning path covering everything from the basics of effective prompts to specific industry applications and open questions regarding data protection.
Asking the right questions is the foundation
The first part explored the possibilities and limitations of generative AI. Good results start with a clear task, relevant context, appropriate examples, and a desired output format.
Even a very good prompt cannot completely prevent hallucinations. It can reduce uncertainty and make the response easier to verify. For this reason, AI systems should not make technical statements that no one in the organization can evaluate.
From emails to meeting minutes
The second part focused on recurring communication. AI can organize incoming emails, draft replies and proposals, summarize meeting notes, or extract information from shared documents.
Custom GPTs or Microsoft Copilot solutions can incorporate internal knowledge. The key factors are up-to-date sources, clear access rights, and the distinction between drafts and final approvals.
Compare Claims and Conditions
The third section applied these fundamentals to typical tasks in day-to-day insurance work. Conversation transcripts can be converted into structured notes or next steps. Policies, applications, and insurance terms and conditions can be compared based on specified criteria. Longer email threads can be summarized and prepared for further processing.
An AI-powered comparison of conditions is an assistance service—not an automated technical or legal review. Sources, omissions, and interpretations must be verified by qualified individuals.
A separate unit for questions and boundaries
The fourth session addressed practical issues: personal data, data processing on behalf of others, costs and settings, transcription, custom GPTs, Microsoft 365 integration, SharePoint, Copilot Studio, and agents.
When it comes to insurance and health data in particular, a technical function alone is not sufficient. Before the data is used, factors such as the purpose, legal basis, data minimization, storage location, authorizations, retention periods, and contractual terms must be clarified.
AI as an assistant, not as a decision-maker
All four sections paint a clear picture: AI can help insurance brokers organize, summarize, and draft documents. It can reduce the time spent on research and help them better prepare for technical reviews.
However, it does not replace professional advice or the responsibility for customer information, recommendations, or claims processing. The more sensitive the content, the more important it is to have reliable sources, restricted access rights, logging, and a dual-control principle.
What Remains from Today's Perspective
Products, user interfaces, and model names are evolving rapidly. The value of this webinar series, therefore, lies less in individual features and more in the way we work:
- Clearly describe the task and objective,
- use only appropriate and approved data,
- Verify the results and sources in a transparent manner,
- leave binding decisions up to people,
- and document successful processes before scaling them up.
Webinar Recap · October 10 through November 7, 2024
From Proper Prompting to Comparing Conditions: AI in the Real Estate Agency
In the fall of 2024, I led a four-part webinar series for the Tyrol Insurance Brokers Association on the practical use of artificial intelligence. Using ChatGPT, Microsoft Copilot, and specific tasks from a brokerage office, the series demonstrated how communication, meeting minutes, and document comparisons can be prepared more efficiently—and why data protection and professional oversight remain essential.