AI Agent AI Flows

ai flows enable enjo ai agents to resolve complex support requests by executing multiple steps in sequence they are designed for cases that require conditional logic, backend lookups, or coordination across systems each flow orchestrates one or more docid\ b8vmn1t7c4ofw1 avvddp , guided by natural language instructions , so the agent can branch logic, make decisions, and complete workflows end to end—without human intervention when to use ai flows a single action or knowledge response is insufficient example verify user identity → fetch account → branch logic based on account type key capabilities sequential execution of multiple ai actions natural language instructions to define flow behavior condition based branching (e g , premium vs non premium users) reusable ai actions across different flows structured orchestration for specific request types ai flow structure field description name internal identifier for the flow description purpose of the flow instruction natural language guidance for execution ai actions list of actions (lookups, updates, integrations) question variations sample user queries to trigger the flow creating and using an ai flow step 1 create a flow navigate ai agent studio → flows click new flow provide name (e g , update insurance policy) description (e g , handles insurance policy updates by account type) save the flow step 2 add ai actions go to ai agent studio → actions create a new action or reuse an existing one define action name input/output fields api/backend integration details test the action independently return to the flow and add the action(s) step 3 define flow instructions in the flow editor, add instruction a clear description of execution steps instructions should specify order of operations conditional logic escalation or fallback rules example instruction “when a user wants to update their policy, first verify identity using email then retrieve account details if premium, confirm policy and allow updates if not, inform the user and escalate to ticketing ” step 4 add question variations provide examples of user queries to train detection “i want to change my insurance policy” “need to update my coverage” “modify my plan” step 5 test the flow use preview agent submit queries matching variations validate that the agent maps queries correctly executes all steps follows branching logic falls back to ticketing if needed best practices keep instructions precise and unambiguous use explicit conditional phrasing for branching test with real world queries reuse well tested ai actions across flows limitations ambiguous instructions may cause incomplete or incorrect execution requires pre configured ai actions