1. All Collections >
  2. 产品 >
  3. 人工智能代理 >
  4. AI Agent Guides >
  5. How to Use AI Agent with Workflow Menu Options

How to Use AI Agent with Workflow Menu Options

Avatar
Joshua Lim
4 分钟

Menu buttons in a Workflow guide customers at the start of a conversation, while an AI Agent handles open-ended replies. Pairing the two narrows the customer's intent first, then lets the AI Agent continue where it helps most: answering questions, collecting booking details, or handling customers who type a message instead of tapping a button.

This suits businesses that want the control of Workflow menus but still want natural conversations once the intent is clear.

When to use this

Use menu options with an AI Agent when the business wants to:

  • Guide customers with menu buttons first

  • Let the AI Agent handle replies that do not match a menu option

  • Let the AI Agent answer questions after a menu option is selected

  • Let the AI Agent collect details before handoff to a human or team

Within the Workflow, an AI Agent fits in four places:

  • The Failure: Invalid branch of the Ask a Question step, when the customer types instead of selecting a button

  • Under a menu option that needs FAQ answering

  • Under a menu option that collects information before human handoff

  • The default fallback branch

Tip: Keep to one or two menu steps where possible. Multiple menu steps make the conversation feel long and cause customers to drop off partway through.

Before you begin

Make sure you have:

  • A respond.io account on the Growth plan or higher — AI Agents and Workflows are available on Growth and above

  • A connected messaging channel

  • At least one AI Agent — start from a Sales, Support, or Receptionist template

  • Knowledge Sources prepared for any menu option where the AI Agent answers questions

Note: You do not need to build every AI Agent at once. Start with one AI Agent, confirm it works well, then add more to other menu branches later.

Methods

Three setups pair menu options with an AI Agent:

  1. AI Agent as a menu fallback

  2. Guided starting point, then AI Agent takes over

  3. A different AI Agent for each menu option

Scenario 1: AI Agent as a menu fallback

Use this method when the menu stays simple, but the AI Agent should handle customers who type a message instead of tapping a button. The Workflow handles the expected button replies while the AI Agent handles everything else.

Method objectives

  • Keep a fixed menu for common intents

  • Catch and understand replies that do not match a button

  1. Create an AI Agent that identifies the customer's intent when they type instead of selecting an option. Start from a template, then add its Instructions. Under Context, describe the fallback role and list the menu options:

     # CONTEXT
     - You handle customers who did not select a menu button or replied in their own words.
     - Menu options are: Service FAQs, Book Appointment, Talk to Sales.

    Under Top-level flow, tell the AI Agent how to handle each intent:

    1. Understand customer intent: Match what the customer needs to the closest menu option. If unsure, ask one clarifying question, then continue.

    2. Route the conversation: If the customer asks about Service FAQs, answer from the Knowledge Source. If the customer wants to book an appointment, assign to @[Booking AI Agent]. If the customer wants anything else, assign to @[Sales Team].

  2. In the Workflow, use the Conversation Opened trigger, then add an Ask a Question step set to Multiple Choice. Add your menu options.

  3. Under the Failure: Invalid branch of the Ask a Question step, add an Assign To step and select the AI Agent created earlier.

  4. Add a Branch step, create a branch for each menu option, and connect each branch to the right Workflow action.

Important: The Workflow stops once the AI Agent takes over. Add Instructions and a Knowledge Source so the AI Agent can handle the scenarios you route to it.

Scenario 2: Guided starting point, then AI Agent takes over

Use this method when the menu only captures the customer's topic. After the customer selects an option, the AI Agent takes over and continues based on that topic.

Method objectives

  • Use the menu to capture intent quickly

  • Hand the full conversation to one AI Agent after selection

  1. Create an AI Agent that treats the selected menu option as the starting context. Under Context, describe how to use the selection:

     # CONTEXT
     - The customer selected a menu option before reaching you. Use it as the starting context.
     - Menu options are: Service FAQs, Book Appointment, Talk to Sales.
     - If the customer typed something else, identify what they need and continue.

    Under Top-level flow, tell the AI Agent how to continue:

    1. Understand the selected option: Use the selected menu option as the context for the conversation. If the intent is unclear, ask one clarifying question before continuing.

    2. Continue based on the option: If the customer selected Service FAQs, answer from the Knowledge Source. If the customer selected Book Appointment, collect their preferred date, time, name and phone number, then assign to @[Sales Team]. If the customer selected Talk to Sales, assign to @[Sales Team]. `

  2. In the Workflow, use the Conversation Opened trigger, then add an Ask a Question step set to Menu. Add your menu options.

  3. After the menu step, add an Assign To step and select the AI Agent. The AI Agent takes over once the customer selects an option.

Scenario 3: A different AI Agent for each menu option

Use this method when each menu option needs a different AI Agent role. The Workflow checks what the customer selected, then assigns the conversation to the right AI Agent or team.

Method objectives

  • Give each menu option its own dedicated AI Agent

  • Route unclear intents to a general AI Agent

  1. Create one AI Agent for each task. For a menu with Services, Book Appointment, and Check Appointment, a common mapping is:

    • Services → Support AI Agent

    • Book Appointment → Sales AI Agent

    • Check Appointment → a human or team

    • Unknown intent → Receptionist AI Agent

    Give each AI Agent clear Instructions and the right Knowledge Source. Without a valid Knowledge Source, the AI Agent cannot answer questions. For the Receptionist AI Agent that catches unknown intents, add Instructions like this under Context:

    Then, under Top-level flow:

    1. Understand customer intent: Match what the customer needs to the closest menu option. If unsure, ask one clarifying question, then continue.

    2. Route the conversation: If the customer asks about services, assign to @[Support AI Agent]. If the customer wants to book an appointment, assign to @[Sales AI Agent]. If the customer wants to check an appointment, assign to @[Support Team]. If the customer wants anything else, assign to @[Sales Team]. `

  2. In the Workflow, use the Conversation Opened trigger, then add an Ask a Question step set to Menu. Add your menu options.

  3. Add a Branch step and create a branch for each menu option. Connect each branch to the right AI Agent or team. For unknown intents, use the Failure: Invalid branch to assign customers to the Receptionist AI Agent. For example:

    • Services → assign to Support AI Agent

    • Book Appointment → assign to Sales AI Agent

    • Check Appointment → assign to a human or team

    • Failure: Invalid → assign to Receptionist AI Agent

What's next?

分享这篇文章
Telegram
Facebook
Linkedin
Twitter

找不到您正在寻找的东西? 🔎