Workflow Automation CoE > Playbooks > AI Agent as an Activity
Foreword
This article assumes a basic understanding of Playbooks in Workflow Studio, including how stages and activities are configured. Familiarity with Now Assist and AI Agents on the ServiceNow AI Platform will also be helpful. If you are new to Playbooks, check out the Playbooks documentation to get started.
Introduction
Modern business processes increasingly rely on AI to accelerate decision-making and reduce manual work. With the 'Use an AI Agent' activity in Playbooks, you can embed a Now Assist AI agent directly into your workflow as a first-class activity, right alongside your existing interactive and non-interactive steps.
Instead of switching to the Now Assist panel or running the agent separately, agents and fulfillers can interact with the AI directly within the playbook experience. The AI agent gathers context, performs work, and produces structured outputs that downstream activities can consume, all without ever leaving the record.
In this article, I will walk through what the feature does, how to configure it, and share a practical example to help you get started.
What is the 'Use an AI Agent' Activity?
The 'Use an AI Agent' activity is a new activity type available under Common Activities in the Playbooks design environment. When added to a stage, it lets you:
- Select an existing Now Assist AI agent
- Provide the agent with a plain-text objective (supporting dynamic data pills)
- Define the expected outputs the agent should produce
- Choose whether the agent runs as the active user or the user who triggered the playbook
The activity bridges human-in-the-loop decision-making and autonomous AI action. The agent does the research and proposes actions, while the human confirms. This means your AI is not acting unilaterally; it is working collaboratively within the guardrails of your defined process.
Before You Begin
Make sure the following are in place before adding this activity to your playbook:
- Role required: playbook.admin or pd_author
- The Now Assist for Platform plugin is installed in your instance
- Required apps: Process automation designer 28.4, Playbook Experience 28.4
- You have at least one AI agent created and available
Configuration and an example
A detailed step-by-step process and example for utilizing this activity can be found on our documentation page. I strongly encourage reviewing it if you are planning to use this feature.
Why This Matters
Many teams today treat AI as an add-on, something you open separately after you’ve already started your task. The 'Use an AI Agent' activity flips that model. The AI is a native step in the workflow, not an afterthought.
This unlocks several practical benefits:
- Consistency: Every case goes through the same AI-assisted research step, no more variability based on whether an agent ‘remembered’ to check similar cases.
- Speed: Agents spend less time searching and summarizing, and more time solving.
- Context: The agent has full access to the playbook's data pills, so its objective is always grounded in the specific record it is working on.
- Traceability: Outputs from the AI agent can flow into subsequent activities as structured data, making the AI contribution a visible, auditable part of the process.
Tips for Getting the Most Out of This Feature
- Write descriptive objectives. The more context you give the agent in the Objective field, the better its output. Use data pills to personalize the instruction to the specific record.
- Define your expected outputs carefully. Outputs from this activity can be referenced by downstream activities. Take a moment to define them in a way that makes them easy to consume later in the flow.
- Choose ‘Run as’ intentionally. If your AI agent needs to act on behalf of the person actively reviewing the record, use ‘Active user’. If the agent should run automatically when the playbook starts, choose ‘The user triggering the playbook’.
Closing Thoughts
The 'Use an AI Agent' activity represents a meaningful step toward truly AI-native process automation. Rather than bolting AI onto existing workflows, this feature lets you design AI participation directly into the fabric of your business process, with the right guardrails, context, and human oversight built in.
If you are already using Playbooks and have Now Assist AI agents available in your instance, this is worth exploring right away. Start with a high-volume, research-heavy activity in an existing playbook and see how much time the agent can save your team.
As always, I hope this was helpful and if you have built something interesting with this feature, share it in the comments!
Further Reading
https://sn.works/HybridWorkflows Learn about combining deterministic and probabilistic workflows for optimized outcomes
https://sn.works/CoE/AgenticPlaybooks Learn to use AI agents to complete playbook activities
https://sn.works/CoE/StartPlaybooks Getting started with Playbooks
