Introduction
The impact of an Internal Chatbot on workplace prevention provider
Client
Workplace Prevention & Occupational Health
Client since
Services
Solutions
Technologies
The problem
Creating an accessible, accurate & responsible Internal Chatbot
For one of our clients, a workplace prevention and occupational health service provider, it's crucial that their professionals have immediate access to the most recent and accurate information. This is due to the fast changing and competitive nature of the market, where legal and regulatory context are always evolving. This case study highlights how we helped our client revolutionize its service offering by centralizing and optimizing access to up-to-date expertise for their more than 1000 advisors. As part of their information management program, our client asked us to develop a chatbot-like interface.
The chatbot should have direct access to their internal knowledge base with several requirements:
Accessibility
- Simplify and speed up the accessibility of information
- Get rid of the dependency of having to know the right specialized search term to find the relevant information
- Support a large user base of more than 1000 advisors and more in the future
- Communicate in both Dutch and French to support the client’s strategy to increase its market share in different regions
Accuracy
- Have access to the full context of each conversation so that user's questions can build on previous inquiries
- Incorporate near real-time updates from source data, ensuring that the most current information is always available
- Integrate various data types, including text, images, tables, and diagrams
Responsibility
- Comply with company policies
- Secure (role based) access to data
- Use of PII following legal requirements
How we solved it
An Iterative approach with all different stakeholders involved
After clearly defining all business requirements and expectations and analyzing the available data, we started with a small proof of value, this consisted of a setup connecting to the client’s information base, using a RAG architecture. RAG stands for Retrieval Augmented Generationand allows us to connect existing LLM's (i.e. Large Language Models, the hidden force behind chatbots like ChatGPT) with our own (private) company data.
Once the proof of value was delivered and accepted, we listed different solution scenarios for further development. In doing so we compared different building blocks with their alternatives, laid out the pros and cons of different possibilities and set priorities. After choosing the best solution in collaboration with both business and IT stakeholders on the client side, we defined a highly scalable but cost-efficient target architecture. Our vendor-agnostic approach allowed the customer to reuse existing technical capabilities, while introducing new cutting-edge technology where needed to support our solution. As part of the solution, we defined an AI governance framework, consisting of integrated guardrails and best practices, to ensure that the solution adheres to the company’s ethical policies.
By using an iterative approach during the development with different stakeholders involved from the start, it’s possible to test results early in the process, adjust quickly, and deliver a minimum viable product with a short time to market. At the same time, the business can already start adopting business processes that are required to ensure that data is validated at the source. The latter is a key success factor to secure accuracy of the answers to the questions that are asked by the users.
An Iterative approach with all different stakeholders involved
After clearly defining all business requirements and expectations and analyzing the available data, we started with a small proof of value, this consisted of a setup connecting to the client’s information base, using a RAG architecture. RAG stands for Retrieval Augmented Generationand allows us to connect existing LLM's (i.e. Large Language Models, the hidden force behind chatbots like ChatGPT) with our own (private) company data.
Once the proof of value was delivered and accepted, we listed different solution scenarios for further development. In doing so we compared different building blocks with their alternatives, laid out the pros and cons of different possibilities and set priorities. After choosing the best solution in collaboration with both business and IT stakeholders on the client side, we defined a highly scalable but cost-efficient target architecture. Our vendor-agnostic approach allowed the customer to reuse existing technical capabilities, while introducing new cutting-edge technology where needed to support our solution. As part of the solution, we defined an AI governance framework, consisting of integrated guardrails and best practices, to ensure that the solution adheres to the company’s ethical policies.
By using an iterative approach during the development with different stakeholders involved from the start, it’s possible to test results early in the process, adjust quickly, and deliver a minimum viable product with a short time to market. At the same time, the business can already start adopting business processes that are required to ensure that data is validated at the source. The latter is a key success factor to secure accuracy of the answers to the questions that are asked by the users.
The results
Information directly in the hands of all advisors
The adoption of the internal chatbot had several direct consequences:
- Advisors can provide quicker and more accurate advice based on the latest information, leading to more efficient and effective client interactions.
- By adhering to strict data policies, the platform maintains high standards of trust and reliability.
- The modular architecture of the platform allows our client to seamlessly extend its user base, adding additional data sources and exposing the interface in multiple environments.
Our client plans to extend the platform's reach by making the interface directly available to clients, further enhancing the advisory experience. This forward-looking approach positions our client at the forefront of the advisory services industry, ready to adapt to future changes in the legal and regulatory environment.
Want to create solutions with AI? Contact our colleagues to see what we can do with the power of AI at your organisation.
Key Learnings
More case studies
We deliver impact where it matters most.
Unlocking Viewer Insights with a Scalable Data Platform
Streamz is the Flemish video-on-demand platform offering local and international series and films through a subscription-based model. Streamz focuses on delivering high-quality content tailored to local audiences, supported by a strategic partnership with Paramount.
