How AI Enhances Insurance Telematics

Exploring the Role of Artificial Intelligence in Modern Risk Management

In this episode of the Insurance Telematics Podcast, Yasmin welcomes back Harald, the CEO and founder of Dolphin Technologies, to explore how artificial intelligence (AI) is transforming the insurance telematics industry. They dive into practical applications of AI in automating workflows, enhancing customer support, and mitigating risk through intelligent systems.

Levels of AI in Use

Harald outlines four levels of AI application:

Basic AI Use: Tools like ChatGPT or Claude for writing or data tasks.
Workflow Automation: Formatting data, summarizing, translating — deterministic and structured processes.
AI Agents: Tools that autonomously choose actions across systems like calendars and email.
Agentic Systems: Multiple agents collaborating within structured frameworks.

The episode focuses on level two, where AI automates structured workflows in insurance telematics.

Real-World Applications of AI at Dolphin Technologies

Weather-Based Risk Alerts

Risk Avoidance: Alerts if a car is in a high-risk area (e.g., hailstorms, theft-prone zones).
Data Structuring: Uses large language models to convert unstructured weather data into machine-readable JSON.
Localized Communication: AI translates alerts into the user’s language while maintaining the insurer’s tone.

AI-Powered Customer Support

Proactive Issue Resolution: Detects background app issues caused by user settings or device limitations.
Tailored Instructions: AI scrapes web resources and generates customized guides based on the user’s phone model and language.
Outcome Monitoring: Tracks whether the issue is resolved by checking if trip data resumes.

The Future of AI in Insurance Telematics

Harald emphasizes that AI should handle repetitive tasks so humans can focus on creative and impactful work. He sees AI as a valuable tool in:

Coding and Development
Data Science
User Interface Design
Quality Assurance

Closing Thoughts

AI is already significantly improving processes in insurance telematics, from weather-based risk warnings to smart customer support. This technology is not just a futuristic concept — it’s being actively used to create safer, smarter, and more efficient insurance solutions.

In the next episode, Yasmin and Harald will delve into driver scoring and driving behavior, and why these metrics are crucial for insurers. Stay tuned!

Yasmin
Hi everyone, welcome back to the Insurance Telematics Podcast. In our last episode, we covered the topic of what really is insurance telematics with Harald, the CEO and founder of Dolphin Technologies. I’m happy he’s back here with us today to discuss a really exciting topic, which has been all over the news recently—AI. We’ll specifically be speaking about the role of AI in our industry, insurance telematics.

Yasmin
Harald.

Harald
Hi, Yasmin. How are you doing?

Yasmin
Hi. I’m good. How are you?

Harald
I’m great. I love AI. I’m not an AI specialist, but I know how we use AI and how we leverage it to make a great customer experience.

Yasmin
I’m excited to learn more about your insights and how we use and leverage AI to build a great customer experience. So, tell me your first thoughts—what is AI and what is the role of AI in insurance telematics?

Harald
Before I answer, let me talk about the different levels I see in AI.

Level one is the base level, how I and you use AI—we go to a website like ChatGPT, Claude, or DeepSeek and interact with a large language model. We give it tasks, let it write text, control data, and so on.
Level two, which we use a lot, is automations—structured workflows where AI brings data into the right format, adds summarizations, helps with translations, etc.
Level three is agents—tools that have access to other tools like emails, calendars, Notion or Airtable databases. The agent decides which tools to use for a task.
Level four would be agentic systems—multiple agents working together in a horizontal or vertical system.
Today, we’ll talk a lot about level two: how we incorporate automations with AI into our workflows.

Yasmin
That sounds really interesting. AI helps us automate processes as you clearly just stated. And I know we use it for weather alerts, right?

Harald
We do.

Yasmin
So, how do we use it in the case of weather alerts?

Harald
First, we want to avoid risk. For example, if your car is parked in an area at risk of hailstorms or theft, we want to inform you. Weather is a big risk element.
We check databases and use APIs to get data about severe weather events—hailstorms, ice, wind, heavy rain. This data is unstructured and varies from one provider to another.
We use a large language model to structure this data into a machine-readable JSON format and store it in our database.
Next, we need to inform the user. So, we send a message, like “Your car is parked in a hailstorm area” or “Roads might be icy tomorrow.” The message must be in the user’s language, with the right tone of voice, depending on the insurance company’s style—casual or professional.

Yasmin
Very strict, professional.

Harald
Absolutely.

Yasmin
Wow, this sounds really great. AI is obviously becoming huge…

Harald
I don’t know, this is not my mother tongue.

Yasmin
No, I’m joking. But you can see how it becomes bigger every day. I think it’s great that we’re leveraging AI. For example, if I’m a customer of an insurance company and need support via an app, is that optimized with AI or does it still require human contact?

Harald
What you’re referring to is bots or chatbots, which use AI and have for a long time. If you talk to support, it might have been a computer, not a person.

We never do first-level support, but we face other challenges. For example, to track mobility, our app needs to wake up—even if it’s closed. This requires permissions from the operating system.
But phone manufacturers want to preserve battery, so they often close background apps. Sometimes users unknowingly revoke permissions.
For instance, iPhone may prompt you: “This app has used your location 20 times—allow or reduce?” Many users choose “only when in use,” which stops our tracking.
With Android, it’s even more complex due to multiple manufacturers and models. Each phone has different settings.
So, we look into our database for a solution. If we don’t have one, we tell an AI to scrape the web.
Then, we generate a structured, step-by-step message for the user in their language and specific to their phone.
Finally, we monitor whether trips are still recorded to confirm if the advice worked.
That’s how we avoid support issues before they even arise.

Yasmin
I like that. Solving a problem before it even becomes a problem.

Harald
Yes, that’s what we do.

Yasmin
Thank you so much for your insightful answer. It’s amazing to see how AI saves time and resources while optimizing processes. Just one last question: how do you see the future of AI in insurance telematics?

Harald
Skynet, they will take over. No—seriously, I try to focus on what AI can do for us.

We use it in coding, data science, writing, UI design, quality assurance.
My goal is to work efficiently and effectively. Anything repetitive that can be done by a machine should be done by a machine.
Free people’s minds for exciting, fulfilling tasks, and let machines handle the rest.

Yasmin
Really great answer. I know that sounds biased because I’m here filming with you, but it genuinely is a great answer. Thank you so much for joining me again.

Harald
Thank you.

Yasmin
I’m excited for the next episodes. Do you want to give the teaser or should I?

Harald
You go.

Yasmin
The next episode, we’re going to be discussing driver scoring and driving behavior and why this data is so valuable to insurance companies. I’m really looking forward.

Harald
Me too.

Yasmin
Yeah. We’ll see you very soon. Thank you for watching and feel free to reach out if you have comments or topics you want us to cover. We’ll see you soon. Thank you.

Harald
Thank you. Bye bye.