Introducing the AI agent function
Blog post from Knock
Knock has introduced an AI agent function designed to integrate AI model capabilities, including web search, into workflows for data enrichment, personalized messaging, and content summarization. When a workflow reaches an AI agent step, Knock combines a user-defined prompt with contextual data and sends it to the selected AI model, storing the response for use in subsequent workflow steps. Users can choose from different model tiers, each with varying credit costs, and utilize web search for additional data enrichment. The AI agent supports prompt creation with JSON schemas for structured data responses, ensuring usability for personalization and enrichment. Testing can be done without consuming credits, and the function operates on a credit model, with varying credit allocations depending on the plan. The AI agent function is now available for all non-enterprise accounts, offering free credits to developer and starter plans.