Agenda Outputs

After a user completes an agenda, the real value often comes from what’s captured - like conversation summaries, key moments, and performance scores. That’s where agenda outputs come into play. You can send agenda outputs to external systems like your CRM, check them out in your dashboard, or get them delivered straight to your inbox!

Types of Outputs

1. Asked Outputs: These are direct questions posed to users during a conversation. For example, collecting a user’s name or email address.
2. Inferred Outputs: These are derived from the conversation context without explicitly asking the user. For instance, you could check a users written English or determine if they are a corporate customer by analysing their responses to other related questions.

Asked Outputs

Asked outputs are specific questions asked to the user, and their responses are recorded as outputs. For example, you might ask the user for their name, email, or phone number. Only the Gather Item  can have asked outputs.

Note: If information—such as the user’s email—is already collected through a web form at the start of the chat, it can automatically pre-populate the corresponding asked output. In such cases, the question will be skipped during the conversation.

• Output Name: The name of the output field. For example: email, name, location, check-in date, etc.
• Gather Logic (Optional): Define your rule for collecting the output data. For example:
  • Check-in date: "User check-in date, format as YYYY-MM-DD"
  • Location: "User current location, format as full address: street, city, state, zip code"
  • Phone number: "User phone number, must be a 10-digit number starting with 0"
• Condition Logic (Optional): Control which output field should be asked or skipped. For example:
  • "ask if {{Phone}} is not provided" (if you have an earlier asked output asking for phone number and the user declined to give it then this output will be asked)
  • "skip if {{Number of Children}} is none" (If you have aksed user ealier if they have children and they say no then you might want to skip a question asking their children ages)
• Pre-Populate Logic (Optional): At the start of the agenda item, the AI will attempt to pre-populate the output field. If successful, the field will be skipped in the conversation.
• Data Type: Select the type of data being output, such as string, boolean, etc.

Asked Output with Format Constraints

By default, the output field will follow the output data type (see more details in section below). However, you can also specify a format constraint to ensure the output value is in a specific format. The format will follow regular expression syntax (regex ). Just put the your regex enquoted with << and >> and put it at the end of the output gather logic prompt.

Example 1::

"User phone number, must be a 10-digit number starting with 0" <<^0\d{10}$>>

Example 2:

"User email, must be a valid email address" <<^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}$>>

Note: Regex is too complicated? You can ask our Persona to generate a regex for you. Just ask “Can you generate a regex to verify email endding with .com for me?” and Persona will generate a regex for you.

Some common regex patterns:

• Email: ^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}$
• Phone number (10 digits starting with 0): ^0\d{10}$
• Zip code (5 digits): ^\d{5}$
• Date (YYYY-MM-DD): ^\d{4}-\d{2}-\d{2}$
• Time (HH:MM): ^\d{2}:\d{2}$
• URL: ^https?://(www\.)?[-a-zA-Z0-9@:%._\+~#=]{2,256}\.[a-z]{2,4}\b([-a-zA-Z0-9@:%_\+.~#?&//=]*)$

Output Data Type

Define the data type you want to collect. This is useful when you want to accurately extract specific information from the user.

For example, when collecting the number of team members, define the data type as integer so that the system collects values like 1, 2, 3 rather than a text string like "one", "two", "three".

Supported data types:

• string
• boolean (true/false)
• integer (whole number)
• float (decimal number)

Inferred Outputs

Inferred outputs are pieces of data extracted from the conversation without directly asking the user.

• Output Name: The name of the output field. For example: user’s English skills, user’s communication style, summary of user’s messages, etc.
• Gather Logic (Optional): Define your rule for collecting the inferred data, similar to Asked Outputs.
• Data Type: Select the type of data being output, such as string, boolean, etc.

Testing Inferred Outputs

You can test and reset inferred outputs by clicking on the buttons below:

The Reset Output option restores the output description to its original state, discarding any modifications made. This is useful when you want to revert back to the default description (which is a summary) without applying changes.

The Testing Output feature allows you to select a previous conversation, input a new output description, and observe how the system processes it. This provides a way to preview how the updated description affects the response. Once satisfied with the results, you can apply the new description to update future interactions accordingly. Step by step:

1. Hit the test button of your desired inferred output to test
2. Select an agenda conversation in the past to test
3. Change the output description to describe how to extract your target information
4. Hit the Test button, new output value and reasoning will show up.
5. Verify the new output value if it match your needs. Then hit the Apply button to update the output description. Note: don’t forget to hit Publish agenda to save the changes.