Getliņi EKO, a waste management company in Latvia, required a solution to reduce the workload on their customer support team by automating responses to routine questions. However, they also wanted to ensure that the quality and accuracy of the responses were not compromised.
To meet these needs, they engaged Tilde to develop Getlīna, a virtual assistant powered by Large Language Models (LLM), capable of providing precise answers on various topics, including waste disposal and recycling.
Client: Getliņi EKO
Industry: Waste management
Trash in, trash out – the correct data is key
To ensure that Getlīna could leverage the capabilities of Large Language Models (LLM), it was crucial to establish a well-defined knowledge base. A knowledge base is a curated collection of documents, manuals, web content, and other relevant information. This serves as the foundation from which generative AI helps the chatbot retrieve information to create accurate and dynamic responses.
Without a well-prepared knowledge base, Getlīna would struggle to provide accurate or helpful information, resulting in poor-quality support. To avoid this, Getliņi EKO carefully selected the right documents for the knowledge base, and Tilde ensured that they were processed and indexed accurately.
More than just an FAQ bot
As Getlīna is powered by LLMs, the bot can comprehend user intent and derive an accurate answer from its knowledge base, which sets it apart from traditional bots that rely on pre-defined decision trees. When users open the chat window, Getlīna proposes three key topics – waste disposal, storage of products, and guided tours – while remaining sufficiently flexible to handle unpredictable questions.
For example, Getlīna can respond to more unusual queries, like, “Will you take my toaster?” or “How much does it cost to dispose of my sofa?”. While a traditional bot might simply provide a generic price per tonne for sofa disposal, Getlīna calculates the exact cost based on the actual weight of the sofa. LLMs allow Getlīna to make such calculations and generate dynamic, personalised responses based on the details provided.
How LLM makes Getlīna smarter
Aside from the ability to understand and respond naturally to user queries, Getlīna’s other LLM-powered capabilities include:
- Differentiating between users (legal entities vs individuals) and tailoring responses on pricing and business hours accordingly;
- Helping users determine if they can visit the store or dispose of waste based on the current date and time;
- Sending product images when users inquire about specific items, such as plants sold in the store;
- Providing links to relevant sources to ensure that users have access to accurate information.
Voice assistance is on the way
Tilde is now working with Getliņi EKO to expand Getlīna’s functionality by introducing voice interactions. Soon, users will be able to engage in real-time voice conversations with the virtual assistant, which will generate and read responses aloud. Additionally, users will have the option to receive a transcript of any conversation, whether voice or text-based, via email as a follow-up.
Faster and more efficient customer support with AI
Getlīna has proven to be a valuable asset for Getliņi EKO by delivering contextually relevant and accurate responses and successfully reducing the support team’s workload.
Companies seeking to optimise their customer service operations can greatly benefit from AI chatbots, ensuring both efficiency and customer satisfaction. With the right development and training, AI can make all the difference.
Looking back at the Getlīna project, I can say it’s been a journey full of knowledge, benefits, and challenges. I am grateful to have worked with Tilde, who diligently addressed our requests and made regular adjustments. Thanks to this collaboration, we’ve created our amazing virtual assistant, Getlīna. Thank you!
Santa Peiraga, Client Service Specialist
