Skip to main content
Agents combine user-defined instructions, knowledge, and Actions to create tailored AI experiences. Think of them as specialized AI teammates designed for specific tasks. ASSIST AI Agents can be used individually or shared across users within the platform.

​
When to Use an Agent

Agents are ideal for repeatable workflows, especially those that can be shared across a team. Common use cases include:
  • Onboarding Assistant
  • AI Helpdesk
  • Engineering Copilot
  • Legal Reviewer
These agents are usually focused on well-defined tasks, such as reviewing documents for policy compliance or completing RFPs based on past submissions.

​
Creating an Agent

Instructions (Prompting)
The Instruction field is infinitely flexible, so it’s important to try out different variations and validate the Agent’s performance before sharing it more broadly.
Below is an example of the default instruction used in the ASSIST AI :
You are a helpful assistant for ASSIST AI. Your role is to assist employees and clients by answering questions, providing information, and supporting day-to-day tasks related to the organization’s work. Always respond in a professional, courteous, and accurate manner. If you are unsure about something, say so — never guess or fabricate information. Escalate sensitive or unresolved issues to the appropriate team or contact.Keep responses clear and concise. Use Markdown formatting where helpful. The current date is [[CURRENT_DATETIME]].
One of the key advantages of creating an Agent is the ability to precisely define its behavior and goals using natural language. For example:
  • Always present results in a table
  • Prefer direct quotes from documents instead of paraphrasing
  • Highlight information from documents older than three months

​
Knowledge

Give the agent only what it needs to know. Fewer, more focused documents lead to better, more consistent results.
When building an Agent, you can attach knowledge from Connectors or upload files directly. Connector-based knowledge stays automatically updated as source documents change.

​
Actions

Actions (also known as tools) enable Agents to interact with external systems through APIs. Some Actions enhance the quality of responses, while others allow Agents to perform tasks beyond answering questions. Examples include:
  • Updating ticket statuses based on conversation outcomes
  • Checking service health to provide real-time updates
  • Advancing deals in a CRM based on call transcripts or user inputs