The challenge
Doccle, Belgium’s largest digital administration platform, services more than three and a half million users, reaching one-third of Belgian households. With this massive user base, Doccle faces daily inquiries, document submissions, and invoice handling. To maintain top-tier customer service, Doccle knew it needed a digital solution capable of addressing a high volume of questions efficiently and accurately. Enter Doccy, the virtual assistant that not only meets today’s needs but is now more advanced than ever.
"In time, we want to ensure that Doccle can manage everyone’s administration fully automatically." Bram Lerouge, CEO, Doccle
Already in 2018, Doccle wanted to tackle their growing pains resulting from exponential growth. In a strategy workshop, we identified the main issue: as Doccle had to cope with an enormous amount of questions from their ever-expanding customer base, a virtual assistant — or chatbot — was needed to lighten the team’s workload.
Together, we started building the chatbot on our Oswald platform. In February 2019, the virtual assistant, eventually named Doccy, was ready to go: with its 24/7 availability, the assistant started providing answers to frequently asked questions. We tailored the chatbot exactly to the audience’s needs, with questions ranging from “How do I create an account?” to “How do I add a company?”. Forgot your password? No problem. Doccy had your back.
Although the virtual assistant reduced customer service workload significantly, we decided to take Doccy a step further in 2024. With the rise of generative AI, we rethought the way our Oswald platform works and decided to incorporate the power of large language models (LLMs). By adding this functionality, we can combine the creativity and adaptability of LLMs with strict, company-specific controls in the more traditional dialogue flows. For Doccle’s next-generation chatbot, called Doccy 2.0, we decided to make the virtual assistant 90% generative by integrating GPT-4o mini with Doccle’s own helpdesk as its knowledge base.
This upgrade allows Doccle to respond to a broader range of user questions with even greater accuracy. With our hybrid approach, we can still control critical questions the old-fashioned way and keep using our integrations (e.g., resetting passwords in the chatbots), while also leveraging the LLM’s power of really understanding questions and responding in a nuanced way. This controlled environment prevents issues that are common in generative chatbots and, of course, we still have human agents are always on standby for seamless intervention if needed, ensuring every interaction is helpful.
The LLM integration significantly enhances Doccy’s scope and responsiveness while keeping token usage efficient. Routine inquiries are addressed by Oswald’s standard chatbot functions, reducing unnecessary token consumption and optimizing cost efficiency. Our partnership with Microsoft Azure also means data privacy and security remain a top priority; Doccy’s interactions are GDPR-compliant and processed in a secure, controlled framework.
Over the years, Doccy has processed more than one million messages — always with 24/7/365 availability. Now, we expect Doccy to give even better, more nuanced answers. By going beyond the proof-of-concept phase and actually moving our GPT-cases to production, we prove that this relatively new technology can be used in customer-facing solutions. We make sure this groundbreaking technology doesn’t just stay on the drawing board. We take it further — thanks to amazing clients like Doccle.
Curious to test Doccy yourself? You can do so via www.doccle.be.
"The power of a digital administration lies in its 24/7 availability. With this virtual assistant, we guarantee that our users always get a quick answer to their questions, even outside office hours. The ultimate goal is that Doccle will eventually be able to manage everyone's administration fully automatically. That is why we continue to invest in innovation and AI, with this chatbot as the first step."
Want to know more?
