When your chatbot goes from “meh” to medical-grade 

We all know chatbots. Sometimes they’re surprisingly helpful, other times they leave you wondering if you’re talking to a confused parrot. That was the case at Hartmann, the global healthcare company behind wound care and hygiene solutions in over 100 countries. 

Their chatbot “Eva” was supposed to help patients and healthcare professionals with reliable product and treatment information. The problem? Eva was only half-right. Literally. With an accuracy rate hovering at 50%, every second answer was off the mark—not exactly the level of precision you want in healthcare. 

Enter Dignify. Instead of treating Eva like just another tech experiment, we approached it like a medical-grade challenge. First, we ran a full audit to understand where things went wrong. Then, step by step, we rebuilt Eva’s brain: narrowing the focus to the most relevant medical documents, layering in retrieval-augmented generation (RAG) for contextual answers, and cutting the digital noise that confused the model. 

The result? Eva went from shaky intern to star performer. In just eight weeks, her accuracy jumped from 50% to an impressive 98%, passing compliance reviews and rolling out globally. 

For Hartmann, this wasn’t just about making a chatbot smarter. It was about trust. Patients and professionals can now rely on Eva to provide clear, compliant, and consistent information—whenever and wherever they need it. 

“Unlocking the true potential of AI isn’t just about hopping on the trend—it’s about expertise and precision. With Dignify’s guidance, we transformed our chatbot into a fully operational agent, seamlessly integrating Hartmann data and LLM comprehension. The result is a GenAI chatbot that delivers accurate answers in a highly regulated environment.” — Tamas Balog, Global Digital Marketing and Media, Hartmann Group 

And that’s the real win: when AI stops being a buzzword and starts making a tangible difference in people’s lives. 

Curious how we turned Eva around? Read the full Hartmann case study here.