Vasion Intelligence

Last Updated: October 25, 2026

Build and run agents with Vasion Intelligence. You configure an agent with instructions, skills, and an AI model. Use agents for tasks that need judgment, such as reviewing a claim or summarizing a request, which standard workflow steps cannot do.

This feature is in beta. Capabilities, configuration, and availability may change as development continues. Contact your Vasion representative with feedback to help shape this feature.

An agent is an AI assistant that follows instructions to read documents and other information and return answers or files.

In Vasion Automate, "agent" means an AI agent in Vasion Intelligence. It does not mean the PrinterLogic Agent software that you install on devices.

Key Points

Review the following key points:

  • You write agent instructions in plain language, or you describe the agent and let Vasion Intelligence generate the first version. You do not need to write code.
  • Agents work in a chat session or as a step in a Business Process workflow, and return a text response, a file, or both.
  • The Vasion Platform provides ready-to-use agents, and you can create your own.
  • Permissions control who can open, run, create, and edit agents.
  • Guardrails block unsafe content and mask sensitive information, such as passwords and payment card numbers.

How Vasion Intelligence Works

An agent combines the following components:

  • Instructions: Describe what the agent does and how it responds.
  • Files: Provide up to five reference files, 256 MB each, for the agent to use.
  • Skills: Define reusable instructions that the agent follows for specific tasks.
  • Tools: Add predefined capabilities that the agent can use, such as running code in a secure environment.
  • Model: Sets the AI model that powers the agent. The models that you can select depend on your region.

You use an agent in one of two ways.

Chat Sessions

Start a chat session with an agent to ask questions, upload files, and review the response. A chat session works well for one-time or exploratory tasks. Refer to Agents to learn more.

Workflow Steps

Add the Agent step to a Business Process workflow to run an agent automatically. When the workflow reaches the step, the agent does the following:

  • Receives information from earlier steps, such as form data and documents. You can select up to five documents.
  • Follows the instructions that you enter in the step.
  • Returns one text response or one file.

Later steps in the workflow use the response through the agentResult variable. To get a generated file, add an Email step and attach the agent output. The file format comes from the instructions that you write for the agent. Agents can create several file types, such as PDF, DOCX, XLSX, and CSV. To generate a file, the agent must have the Code Interpreter tool. For all supported file types and limits refer to Agent File Output Reference.

Agent Types

Vasion-Built Agents

Vasion creates and maintains these agents for specific uses, such as claims analysis for healthcare. Vasion protects the core instructions of these agents, so you cannot edit them. You can clone a Vasion-built agent to create your own copy.

Custom Agents

You create these agents to fit your own processes. You set the instructions, skills, and model. After you create a custom agent, you can select it in any workflow that uses the Agent step.

Guardrails

Vasion applies guardrails to every agent. Guardrails screen the information that you send to an agent and the response that the agent returns. You cannot configure guardrails.

Guardrails do the following:

  • Block content in these categories: hate, insults, misconduct, sexual content, violence, and prompt attacks. A prompt attack is an attempt to override the agent's instructions. Vasion Intelligence checks for prompt attacks in your input only.
  • Mask sensitive information, such as passwords, payment card details, bank account numbers, and government-issued ID numbers. Vasion Intelligence replaces the value with a placeholder, and the agent continues.

When a guardrail blocks your input, the agent displays "Your request was blocked by a content safety filter." For a blocked response, the message is "The response was blocked by a content safety filter." The message does not identify which filter applied.

Guardrails do not mask general contact details, such as names, email addresses, and phone numbers, and they do not replace agent permissions or access levels.

Requirements

Review the following requirements:

  • Permission to access Agents.
  • Use access or higher on the agent that you want to run.
  • Creator permission.

Agent Step Requirements

In addition to the general requirements, the Agent step requires the following:

Next Steps

Refer to the following: