Pega Knowledge Buddy

Pega Knowledge Buddy is an innovative product designed to provide clear and precise answers to specific questions in any web application or online experience. The Knowledge Buddy tackles the challenges of time-consuming searches and the risk of human error by delivering quick, concise answers to questions that would typically require browsing through multiple documents and systems. Knowledge Buddy employs Pega GenAI™, a cutting-edge technology that combines the use of the Generative Pre-trained Transformer (GPT) and Retrieval Augmented Generation (RAG) framework to offer accurate and relevant responses to user queries, based on the provided knowledge base. Furthermore, Knowledge Buddy customizes responses according to the audience based on their respective access permissions, catering to employees using Pega applications, self-service users seeking information during a customer service journey, or back-office agents addressing complex scenarios.

How does Pega Knowledge Buddy work? First, you create or identify the content that you want your buddy to use as a basis for its responses. This content then serves as a data source. Once you have configured one or more data sources, you can create a buddy – a Large Language Model-based tool that employs RAG to answer questions using the knowledge you provided. Pega Knowledge Buddy enables you to create multiple buddies, each with a unique set of data sources, and each with a different query purpose, including sales, customer service, marketing and more.

Let's examine the data ingestion in more detail.

Customers can input any text-based content into Knowledge Buddy using the REST API service. When used in combination with Pega Knowledge Management software, data ingestion can be run automatically whenever a knowledge article is published. Once the content is uploaded, Pega Knowledge Buddy generates a set of content chunks and processes them through the Pega GenAI gateway. This step creates embeddings – segments of content with detailed semantic features attached to them. Embeddings are finally saved in the Pega GenAI Vector Store.

Now, let's explore what happens when a user submits a question through one of the available channels.

Pega Knowledge Buddy receives the question from the application through the corresponding REST API service. The question is then passed through the Pega GenAI gateway, where it is converted into embeddings. Next, Pega Knowledge Buddy runs a similarity search in the vector database to find relevant content embeddings that match the question. Once the matching content is identified, Knowledge Buddy adds it to a specially formulated prompt. This prompt, along with the user role information, is sent to the Pega GenAI gateway again. The Large Language Model within the GenAI gateway processes the prompt and generates the answer to the user's question.

You have reached the end of this video.