Less administrative time in healthcare thanks to secure AI conversation reporting

To achieve this, they enlisted our help, and together we started an AI proof of concept using LLMs (Large Language Models), publicly known as ChatGPT, Gemini, or Azure OpenAI. The goal: reduce administrative burdens and increase focus on clients.

Dedimo asked us to develop their initial idea into a functional prototype that would be safe, scalable, and usable in everyday practice. Afterward, we could implement the prototype into their platform, ADHDcentraal. As we had already been their Data & AI partner for some time, we were able to provide immediate support.

Approach

We began with discussions to determine what the solution should look like. Dedimo had some examples; we helped translate them into a clear, actionable approach. The core: a system that helps therapists generate reports while keeping them fully in control.

We worked alongside Dedimo’s partner, Zooma, during this process. We were responsible for the backend, transcription, and report generation, while Zooma developed the frontend in close collaboration with Dedimo.

Phase 1 – Building a prototype

In collaboration with Zooma, we developed a standalone application to test the concept, with client privacy at the forefront from day one. Through our implementation, conversations are automatically transcribed and then converted into reports. We consciously chose Microsoft Azure’s speech service to ensure all data remains within Dedimo’s own secure cloud environment. Additionally, it’s guaranteed that this data is not used to train models. Using sample recordings, we successfully tested the entire process, and the results formed the foundation for moving forward to phase 2.

Phase 2 – Integration into Dedimo’s existing platform

Dedimo decided to integrate the prototype directly into their platform, ADHDcentraal. Therefore, we placed our standalone application into their existing environment so it could be used for real client conversations.

Phase 3 – Monitoring and further development of LLM models

Because we work with LLMs, effective monitoring is crucial. LLMs can make mistakes, which need to be corrected by humans. Therefore, we embedded features in the application that allow Dedimo to verify whether reports are accurate and whether models need adjustments. This ensures high quality, even when models change due to updates from Microsoft, OpenAI (ChatGPT), or Google (Gemini).

Thanks to this approach, Dedimo can:

1. Monitor the quality of reports
2. Continuously improve LLM texts, prompts, and models
3. Quickly detect anomalies
4. Meet the need for a human-in-the-loop process (where humans review, assess, and guide LLM decisions).

By constantly monitoring and improving, the LLM conversation reporting remains reliable and future-proof.

Result

The solution is now fully integrated and being tested in real practice. Recent test conversations already show how much time therapists can save.

The biggest benefits of this implementation:

• Time savings: Therapists save time as they no longer need to manually write summaries after sessions.
• Overview: During the session, the LLM processes the client’s responses, allowing the therapist to see which questions remain open and which have been addressed.
• Control and flexibility: The LLM generates a complete final report, which the therapist can adjust or fine-tune by giving instructions to the LLM.
• More attention to clients: By reducing administrative work, therapists have more time to help additional clients or focus on other aspects of care.

Because sensitive data is involved, human oversight remains crucial. The process is deliberately not fully automated.

Furthermore, it’s important (and Dedimo attests this): this is not a “set and forget” solution. LLM transcription and reporting require ongoing attention, especially when models change or receive updates.

Future

After our implementation, we fully transferred the solution to Dedimo’s team. We carefully handed over the project, ensuring the team could independently monitor, manage, and further develop the LLM conversation reporting. As a result, Dedimo is now capable of taking full responsibility for the ongoing development of the solution.

Receive data insights, use cases and behind-the-scenes peeks once a month?

Sign up for our email list and stay 'up to data':