How to setup AvoAI
A guide on Avochato’s AI agent functionality
Welcome to the AvoAI product guide! This guide provides step-by-step instructions, helpful tips, and best practices to help you set up AvoAI and maximize the effectiveness of your AI agent. Whether you’re just getting started or looking to optimize your workflows, you’ll find everything you need to get the most out of AvoAI.
AvoAI Agent
Follow the steps below to create and set up your AI agent. You can update your selections at any time, both before and after activating your agent.
Note: Even when your AI agent is turned ON, it will not automatically engage with contacts. The agent will only participate in conversations if you manually assign it to a conversation or include it in Avochato’s auto-assignment and skill-based routing features.
STEP 1: Configure Your Agent
1. Name your Agent
What do you want your AI agent to be addressed as? When AvoAI is asked its name, this is what it will reply with.
2. Choose a Personality
How do you want your AI agent to interact with your audience/contacts? We have a variety of different combinations to suit a wide range of personalities.
3. Greeting Behavior
How do you want your bot to address your audience/contacts? When AvoAI replies to your audience/contacts, this will be the preferred name it will call them by. “None” would mean the bot will not refer to your audience/contacts by any specified name.
4. Response Delay (Seconds)
How long of a delay do you want your AI Agent’s messages to send? AvoAI messages are set to respond after this delay. This will allow your team to manually edit or cancel pending messages, prior to that outbound sending.
- Set a longer delay to give your team time to review your bot’s responses before they send. However, please note that messages will send even while an agent is editing a message once the time delay is up.
- Set a shorter delay to respond nearly instantly to each inbound message. However, please note that if your team plans to review messages beforehand, this will give them less time to do so before the message sends.
5. Use Contact Fields in Responses
This option lets your bot have access to your contact’s data. Selecting “Yes” means that Avochato contact fields, including all custom fields, will be used to help the assistant provide more specific responses to each contact that engages with the assistant. This data is NOT used to train third party large language models or shared across different contacts or inboxes, and is only used when replying to that contact. The contact fields used are the following:
6. Hours of Operation
Control when your AI agent will reply based on the time of day and your inbox’s Office Hours. By default, AI agents will respond during all hours.
Reminder: if an inbox has no Office Hours configured, any time of day is considered to be “during office hours” for that inbox. Office hours can be configured per day of the week, per inbox.
7. Escalate to a Teammate
Automatically transfer conversations to a teammate when your AvoAI Agent detects that human assistance is needed. Conversations are reassigned based on your inbox's routing rules, considering skills, availability, and office hours. If no teammates are available, your AvoAI Agent will notify your team via an in-app notification.
8. When Escalating, Skip Responding
Skip sending a response when escalating to a human. By default, AI Agents respond to contacts when escalating tickets.
9. When Escalating, Trigger Insights
When a conversation is escalated to a teammate, AvoAI will automatically generate Insights to give your team member helpful context as they join the conversation. This makes it easier for teammates, especially during broadcasts, to quickly assess which escalated conversations are high or low priority and respond appropriately. Click here to know more about Insights AI (Sentiments).
10. How should the AI Agent decide which messages to look at?
This setting determines how the AI agent references previous messages in the current conversation when generating a response. You can select whether the AI Agent considers a specific number of past messages, or only messages sent within a chosen time frame (such as the past day or hour). The AI agent only reviews messages from the ongoing conversation with the contact. Messages from other conversations with the same contact are not included. Additionally, failed or unsent messages will be ignored.
11. How many past messages should the bot consider?
This setting works in conjunction with the previous option #10 if “By number of messages” is selected. It specifies how many past messages the AI agent will review in the current conversation when generating a response to a new incoming message.
12. How far back in time should the agent look?
This setting works in conjunction with the previous option #10 if “By time window” is selected. This determines how far back in time the AI agent will review previous messages in the conversation when generating a response to a new incoming message.
13. Enable the AI agent to add these tags to contacts
Enter the list of tags that the AI agent is permitted to apply to contacts. The Agent will only add a tag if you give specific instructions on when to use it (continue below to “STEP 2: Instruct Your Agent” for more info).
You can search for existing tags or create new ones by typing in the input box. After adding your tags, don’t forget to click Update or Next to confirm your changes. Tags are case-sensitive.
After you’ve configured your AI agent, you’re ready to activate it. Simply go to the Overview page and toggle the status to ON.
For the best results and more accurate responses, we highly recommend training your AI agent with detailed instructions before activation. Providing specific guidance will help your AI agent respond more precisely to inbound messages. Continue on below for guidance in training for AI agent with instructions.
STEP 2: Instruct Your Agent
Add New Instructions via “Instruct Agent”
To train your AI agent on how to better interact with your audience/contacts, click on the purple “Instruct Agent” button on the right to show the new instructions window.
- Enter an Instruction Title
Create a descriptive name for this instruction. This name will help you organize and find this instruction if you later need to edit or disable it. Note that this title will not be used by the AI agent in its responses, only the instruction content itself will be used.
- Content
The content you enter in this instruction will be used by the AI agent. This section is fully customizable—you can add as much detail as needed to guide your AI agent's behavior (there is no character limit).
Below is a sample content of an instruction
“You are a customer support agent at [insert company name]. You will provide details about our product and services, and instruct contacts to do the following steps [insert relevant details and instructions here based on your use case]. You will refer to the online documentation located at [insert link(s) here].”
Specific, public URLs must be used, not just a single domain link. In the future, functionality to scrape the contents of an entire public domain may be available.
Add New Instructions via “Recipes”
If you're unsure how to craft instructions for your AI Agent, we've prepared templates called "recipes" to help you get started. Click on any recipe to open an instruction window with pre-populated content that you can customize.
Organize Existing Instructions
Instructions you've created will show in a table below recipes. You can either disable, enable, edit or delete them as you wish. Please note that deleting instructions will completely remove it from the system and cannot be recovered.
Tips and Tricks for Maximizing Your Instructions
- Multiple Instructions
You can create multiple instructions for your AI agent to learn from. While creating multiple instructions isn't required, it can be helpful for users who want to organize their training data while updating other instructions.
As an example, Avochato's AI agent could be trained with instructions on both sales and support information. However, if Avochato was in the process of updating the contents of the support instructions, we could mark that specific instruction as inactive so as not to disrupt the AI agent from answering sales questions.
- Create “If This, Then That” (IFTTT) Statements
Use IFTTT statements in your instructions to guide your AI agent’s behavior by specifying clear triggers and responses. Start by defining the type of inbound message that should prompt a response, and then provide the exact reply you want the bot to send.
Example IFTTT statement:
If a user asks about the ABC topic, the AI agent will respond with: “Thanks for your question. ABC can be used for …”
This approach helps ensure your AI Agent replies consistently and accurately to specific inquiries.
Activate your AI Agent
Once you’ve finished building and training your AI agent, navigate to the Overview page to activate it. Simply click the Status toggle so that it displays as Active.
Your AI agent is now ready to interact with your contacts and can work alongside your team as part of your auto-assignment roster.
Note: The AI agent will only engage in conversations when you assign it to a contact. It will not automatically join conversations unless specifically assigned.
You can manually assign your AI agent to any conversation in your inbox:
- Go to All Messages and select the conversation you’d like to assign
- In the Assignee dropdown, choose your AI agent
Once assigned, your AI agent will begin responding to the conversation after the next incoming message.
To remove the AI agent, simply unassign it from the conversation at any time to stop it from responding. If you want to disable the agent completely, go to the Overview tab and toggle the Status to Inactive.
AvoAI Fees
AvoAI fees are calculated based on your actual usage of Avochato’s AI features. These charges are deducted from your account balance. To view details about your AI usage and associated fees, go to your Billing page and click on the “Balance and Usage” tab.
