Imagine searching for crucial information—whether it’s a company policy or technical documentation—buried within a labyrinth of documents and links. EmployeeGPT, an AI tool for internal productivity, emerged from the frustration of navigating company documents. Today, it’s transforming how knowledge is accessed at Delivery Hero. With growing interest and questions raised on global stages like AIAI Berlin, we chat with Timothy Low, the creator, to uncover how a personal project evolved into an essential AI solution for teams worldwide.
EmployeeGPT: Boosting Efficiency with a RAG-Powered Workplace Chatbot
EmployeeGPT, powered by Retrieval Augmented Generation (RAG) technology, integrates seamlessly with platforms like Slack to deliver fast, accurate answers from diverse sources like Google Drive, GitHub, and Confluence. By cutting through the noise and providing real-time, accessible knowledge, EmployeeGPT has become an indispensable tool, enhancing team efficiency and communication. See the image below for a detailed look at how EmployeeGPT operates.
Now, let’s hear from Timothy, the mind behind this incredible chatbot!
Q: Timothy, how did the idea for EmployeeGPT come about?
“It started pretty simply—I was trying to find out whether my shingles vaccine was covered, but the info was buried in too many documents. I fed the People Team’s docs into GPT4All, and when I got my answer, I realized that this tool could solve much bigger problems across our internal knowledge base.”
Tim’s passion is evident as he speaks. What began as a personal side project quickly turned into a tool that’s now enhancing knowledge discovery across Delivery Hero’s global operations. Powered by Retrieval Augmented Generation (RAG), EmployeeGPT is capable of scanning large data volumes and providing accurate, context-based answers to user queries.
Q: What challenges did you face during the development of EmployeeGPT?
“One of the hardest parts was making the bot accessible to everyone—regardless of language or document format. PDFs had strange encodings, and handling multiple languages added layers of complexity. Early users loved the speed but weren’t fans of the interface!”
Despite these challenges, Tim’s dedication and innovative use of RAG technology helped overcome the obstacles. By employing serverless, on-demand tools, they managed to keep latency low and costs as low as 40 EUR per month, making the tool scalable and efficient.
Q: How does EmployeeGPT avoid common AI pitfalls like hallucinations?
“RAGs are great because they only provide answers based on real documents. If the documents are good, the answers are good. We also made sure the documents we ingest are structured in an AI-friendly way.”
This focus on preventing AI hallucinations sets EmployeeGPT apart from many other AI solutions. By leveraging well-structured documents, Tim and his team ensured that the tool delivers accurate, reliable results—eliminating guesswork and improving productivity across the board.
Source: https://python.langchain.com/v0.2/docs/integrations/vectorstores/google_bigquery_vector_search/
Q: How does EmployeeGPT handle multiple languages and integrate with existing tools?
“Handling multiple languages is tricky, but since the bot sticks to the documents it retrieves, any issues tend to come from the documents themselves, not the language. As for integrations, we’ve connected EmployeeGPT with Slack, Google Drive, GitHub, and Confluence. As more integrations are added, scalability actually improves.”
EmployeeGPT uses BigQuery for vector storage, which allows the tool to handle integrations efficiently. The design makes it easy for non-technical users to interact with the bot via natural language queries, ensuring that everyone benefits from its capabilities.
Feedback from EmployeeGPT Users
Since its inception, EmployeeGPT has transformed how employees interact with company knowledge. Here’s what some of our users have to say:
- “As someone who regularly interacts with EmployeeGPT, I’ve found it to be an invaluable tool for obtaining relevant information. The chatbot’s intuitive design and quick response time have significantly reduced the time spent on routine queries, allowing our team to focus on more strategic tasks. EmployeeGPT provides personalized support, making our employees feel heard and valued. It’s been a game-changer for our team at foodpanda.” – Brian Arvind Chandran, Lead People Services
- “PeoplePal, a subset of EmployeeGPT, has transformed the way we engage with employees—offering quick, efficient, and personalized support 24/7. The collaboration in developing PeoplePal was marked by clear communication and a deep understanding of our team’s needs, ensuring that the final product is not just functional, but also aligned with our company values – huge shout out to our Pandora Tech team led by Timothy Low! Amazing work indeed!” – Jonathan Tam, Head of People Services
- “I’ve found EmployeeGPT to be a fantastic support tool, especially when onboarding the Treasury Central team. The bot’s ability to understand and answer queries based on our specific policies and procedures has been key in streamlining the onboarding process. We’re now working with the compliance team to leverage EmployeeGPT in a similar way, making sure everyone has easy access to and understands our company’s policies.” – Wiktor Jackowiak, Senior Manager, Transformation & Automation
From Side Project to Company-Wide Essential
What began as a personal side project—an attempt to answer a simple vaccine query—has now scaled into a company-wide tool that transforms how employees at Delivery Hero retrieve information. Tim originally built EmployeeGPT for fun, but once its potential was clear, it became an indispensable resource for teams worldwide.
Q: How did you realize this project had the potential to scale across the company?
“When I first built it, it was just to answer my own question. But once more people started using it, the benefits were obvious. It wasn’t long before it became the go-to tool for quick, accurate information retrieval across teams.”
Q: Then what’s next for EmployeeGPT?
“We’re developing channel defaults so teams can use the bot in Slack without extra setup. We’ve also launched an API to expand its platform reach. We’re always seeking ways to make it more efficient and useful.”
This transformation from a small project to a global tool underscores the power of thinking big, starting small, and scaling thoughtfully. EmployeeGPT’s adoption across Delivery Hero showcases both the tool’s scalability and Tim’s innovative approach to solving real-world problems with AI.
On to Global Recognition
As EmployeeGPT scaled across Delivery Hero, it caught the attention of key contributors like Raghav Bali, Staff Data Scientist, who joined forces with Tim to further develop the tool. At AIAI Berlin, Raghav presented EmployeeGPT, showcasing how the tool is transforming internal productivity. What began as a personal side project is now gaining recognition beyond Delivery Hero, proving its potential as a scalable, game-changing AI solution. Find more about the event on the AIAI Berlin Speakers Page.
Final Thoughts
EmployeeGPT isn’t just another AI tool—it reflects Timothy Low’s passion for solving real-world problems with smart technology. By boosting productivity for thousands of employees, it has become essential to Delivery Hero. With more integrations and updates planned, EmployeeGPT’s future looks bright.
Want to learn more about EmployeeGPT’s technical details? Read the original article EmployeeGPT: Leveraging AI for Internal Knowledge Discovery from the Delivery Hero Tech Blog. Or, if Tim’s journey inspires you, reach out for a coffee chat!
If you like what you’ve read and you’re someone who wants to work on open, interesting projects in a caring environment, check out our full list of open roles here – from Backend to Frontend and everything in between. We’d love to have you on board for an amazing journey ahead.
