AI Agents

Introduction

AI Agents in Genio automate processing and decision-making based on structured data. They extract relevant information, generate insights, and store results in mapped fields within the system.

Creating an Agent

AI Agents can be created in the Genio interface:

  • Navigate to Processes > AI Agents.
  • Click on Insert and define its properties.
  • Configure prompts, data mappings, and execution settings.

Defining an Agent

Each AI Agent consists of the following components:

General Properties

  • Id: A unique alphanumeric identifier (underscores allowed).
  • System Prompt: Defines the AI’s processing logic and response structure.
  • User Prompt: Dynamically structured data passed to the AI.
  • Base Area: The primary table providing context for the agent.

Mapping Table

Defines how the AI response is processed and stored.

Field Description
Destination Field The field where the AI response will be stored.
Mapping ID The property name in the AI request schema. Each mapped field becomes an entry in the AI request.
Mapping Type Specifies whether the mapping is a Formula or AI Input Mapping.
Expression If Formula, contains the Genio formula to compute the value. If AI Input Mapping, it can be left blank.

Expressions in the User Prompt

Genio expressions allow dynamic values to be injected into the User Prompt using {{ }} syntax. Supported expressions:

1. Field References

  • Example:
    {{ [TICKET->TITLE] }}
    Inserts the TITLE field from the TICKET table.

2. Function Calls

  • Example: {{ GetCurrentYear() }}
    Returns the current year.

3. Conditional Expressions

  • Example: {{ iif([TICKET->PRIORITY] = 'High', 'Urgent', 'Normal') }}
    Returns "Urgent" if TICKET->PRIORITY is "High", otherwise returns "Normal"

Example: Ticket Validator Agent

This agent reviews tickets and suggests improvements before submission.

Field Value
Id TicketValidator
System Prompt Review the provided ticket and identify if any changes are necessary. Organize them clearly and mark correct sections with ✅.
User Prompt Ticket Information:
- Category: {{ [TICKET->CATEGORY] }}
- Program: {{ [TICKET->PROGRAM] }}
- Priority: {{ [TICKET->PRIORITY] }}
- Title: {{ [TICKET->TITLE] }}
- Description: {{ [TICKET->DESCRIPTION] }}
Base Area TICKET (Support Tickets Table)
Mappings 1. AIReview → AI Input Mapping → ReviewComments
2. CheckedByAI → AI Input Mapping → Verified

Execution Flow

  1. The AI receives ticket details in the User Prompt.
  2. It checks for missing information, categorization errors, and clarity.
  3. The response is stored in the mapped fields.
Example Input
CATEGORY PROGRAM PRIORITY TITLE DESCRIPTION
Incident CRM System High Login Failure User cannot log in after update.
AI Request
Ticket Information:
- Category: Incident
- Program: CRM System
- Priority: High
- Title: Login Failure
- Description: User cannot log in after update.
AI Response
ReviewComments Verified
Title could be more specific. Describe what happens when trying to log in. 1

Calling an AI Agent

Button to AI Agent

This component allows users to trigger an agent directly from a form, with two execution modes:

  • Persist Mode:

    • Executes the agent and immediately saves the response to the record.

  • Client-Side Mode:

    • Executes the agent and updates the fields only on the client.
    • The user must manually save the record afterward.
    • ⚠️ All fields involved in the AI response must be present on the form, or an error will occur.

To use it, add the Button to AI Agent to a form and configure the agent to be called.

Using a trigger

Alternatively, agents can be called via Triggers:

  • Go to the Triggers menu.
  • Create a new trigger and set its conditions.
  • Choose the action Call an AI Agent and select the desired agent.

This allow more flexibility in the way that the agents are called

Trackchanges:

  • 365.04: Agent definition and trigger action to call agents
  • 366.11: Button to Agent
  • 368.20: Fixed problem when there is more than one response parameter of type 'Agent'