A custom agent is ONE doing a job that you define. It starts when something happens in Factorial IT or on a schedule, follows your instructions and only uses the knowledge and connectors you hand it.

Feature/Functionality

Overview

You can build an agent from a blank page or take one of the agent templates and adjust it. Each agent has triggers (what starts it), instructions (what it does), knowledge and connectors (what it can use) and a model tier. You can test it before it runs for real and follow every run afterwards.

Before you start

• To create, edit, enable or delete agents, you need the Manage AI agents permission. View AI agents lets you see agents and their runs.
• Publish knowledge base items for every policy or procedure the agent must follow.
• Set up connectors for every system outside Factorial IT the agent must reach.

How to use

Go to AI Agents > Agents and click Browse templates. In the Templates catalogue, choose a template to adapt, or click Start from scratch.

Name and describe the agent:

• Name: up to 80 characters. This is how the agent shows in the list and on its runs.
• Icon: click Shuffle colour to give the agent a different colour.
• Description: one or two sentences on what the agent is for.

Click Add trigger and pick a type:

• Event: the agent runs each time the chosen Factorial IT event happens. Search the list and select the event.
• Schedule: the agent runs on a repeating pattern. Set Starting (a date), every n hours, days, weeks or months, at a time, and a time zone. Weekly schedules let you choose the days; monthly ones let you choose the day of the month.

Write the Instructions. The agent gets them on every run: what to check, what to do, what to leave to your team and how to report back. See "Write good instructions" below.

To have ONE draft them, click the assistant icon beside Instructions. It reads the name, description and triggers on screen, then either asks you questions or writes a draft from that setup. The draft goes into the field. Nothing is saved until you click Create.

Under Knowledge, select the published knowledge base items the agent may read. The agent reads these items and no others. Unpublished items are not listed.

Choose what the agent can reach:

• Web access: lets the agent search and read the web.
• Connectors: switch on each connector the agent may use, including the Factorial IT connector for your Factorial IT data. Expand a connector to set a policy for each tool.

Open Advanced settings and choose the Model tier used on every run of this agent:

Click Create. The agent is created already enabled and runs the next time a trigger fires.

Write good instructions

The agent follows its instructions word for word on every run. Cover four points:

• The job. What the agent is for and what it reviews on each run.
• The actions. What it may do alone, and what it should suggest rather than do.
• The limits. What it must never do, and when it should stop and hand over to your team.
• The output. What it delivers (a ticket, a comment, a report) and who receives it.

Name the people, policies and thresholds it should use. If a rule already lives in a knowledge base item, select that item under Knowledge instead of pasting the rule into the instructions.

Test an agent with Dry run or Run now

You can see what an agent does without waiting for a trigger. Open the agent: its page offers Dry run and Run now.

Dry run works on your real data but changes nothing. Any email, ticket, comment or device action it would have made is logged instead of carried out. The run details include a What it would have done section ahead of the report. Tools set to Always refuse are left out of a dry run, just as on a real run.

Run now runs the agent for real once you confirm. Use it when a dry run shows the result you expect.

Both use the saved version of the agent, so save your changes first. An agent triggered only by events cannot be started this way, because it has nothing to work on. A disabled agent can still do a dry run, but you must click Enable agent before Run now.

Runs you start show as Running in the Runs tab until they finish.

Follow its runs

Open an agent and go to the Runs tab to see each run, the trigger behind it and the result. Click a run for its details: a summary of what the agent did, its report, any files it produced and, if it failed, the error.

To see runs from all your agents together, go to AI Agents > Runs.

Manage your agents

The Agents list shows each agent with its type (Factorial IT or Custom), its triggers and status. For the last 30 days, it also shows its runs, the runs that need attention and the AI credits used. Narrow the list with the tabs (All agents, Enabled, Disabled, Failing), search and filters, and use Save as to keep a view.

From the Actions (…) menu of an agent:

• Disable agent: stops every trigger. The agent keeps its setup and run history, and Enable agent switches it back on. A disabled agent carries on with tickets and tasks that already have it as their AI agent. It cannot be added to new tickets or tasks: it shows greyed out in the AI agent picker. To stop it on a ticket, set the ticket's AI agent to None.
• Delete agent: removes the agent. Its past runs stay available.

Changes take effect from the next run. While there are unsaved changes, the agent shows Pending changes.

Find your former workflows

On 12 and 13 October 2026, Factorial IT turned the workflows you had configured into custom agents. Each agent keeps the schedule of its workflow and its instructions exactly as written, plus a short note on how to send its reports and emails. Its report is now the run report of the agent, in HTML, rather than a PDF or CSV file. Past workflow runs were not carried over.

Some workflows became disabled agents, for instance those that had not run in the previous 35 days. They are in the Disabled tab of AI Agents > Agents and do nothing until you click Enable agent in their … menu.

A migrated agent can use the tools of the Factorial IT connector without asking, whether it arrived enabled or disabled. Review the connectors of each one and set read-write tools to Always refuse for any change it should not make alone.

Workflows that were inactive or had no instructions were not migrated and no longer exist.

Troubleshooting and FAQ

Troubleshooting

• Run now is not available. The agent may only have event triggers, so it has nothing to work on. If it is disabled, click Enable agent first.
• Your latest changes did not apply to a run. Dry run and Run now use the saved agent. Save, then run again. Edits only affect the next run.
• The agent cannot use a tool. The tool is set to Always refuse for this agent, or the connector is switched off. Expand the connector in the agent to check.
• A disabled agent is still working on a ticket. That is expected. Set the ticket's AI agent to None to stop it.
• You cannot find a migrated workflow. Look in the Disabled tab of AI Agents > Agents. Workflows that were inactive or had no instructions were not migrated.
• The agent is not in the AI agent picker. It is disabled. Enable it to add it to new tickets or tasks.

FAQ

1. Can an agent ask me before it runs a tool?
   No. Nobody is there to answer during a run, so only Always allow and Always refuse exist for agents.
    
2. Does a dry run change anything?
   No. It records every email, ticket, comment or device action instead of doing it.
    
3. Can an agent have more than one trigger?
   Yes. It can combine several events and schedules.
    
4. What happens to past runs when I delete an agent?
   They stay available.
    
5. Can the agent read every knowledge base item?
   No. It reads only the published items you select under Knowledge.