The AI capability 'suggested replies' in Deskhero utilizes advanced natural language processing and machine learning algorithms. Here's how it works:
Use case 1
Generating AI-powered suggestions - When a new ticket is created, Deskhero's AI analyzes the ticket content and generates embeddings vectors using OpenAI Embeddings API. These vectors capture the semantic meaning of the ticket and related data.
Use case 2
Searching for relevant data - The embeddings vectors are stored in a vector database, allowing quick retrieval and comparison.
Use case 3
Matching and suggesting replies - When a query or ticket comes in, the AI runs it against the embeddings database to find similar and relevant data. The relevant data is then passed to ChatGPT, an AI language model, along with the original question. ChatGPT generates a human-like reply based on the attached data, which is presented as a suggested reply to the agent.
Use case 4
Use case 4: Enhancing search capabilities - Additionally, the AI capability 'suggested replies' also powers the in-app search functionality. When users search for information within Deskhero, the AI analyzes the query and retrieves relevant data, providing phrased replies based on the matched customer data.
