> For the complete documentation index, see [llms.txt](https://docs.siit.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.siit.io/ai/it-agent.md).

# IT Agent

The IT Agent resolves operational work autonomously. It runs your playbooks across tools like Okta and Jamf, collects approvals in Slack or Teams, and escalates to humans when needed

Siit’s AI‑powered IT Support Agent — an intelligent teammate that can autonomously run playbooks, take approved actions in your tools, and escalate to humans when needed.

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fz9xmM8V3atukfr2XaElR%2Fuploads%2FjIoi6ywicR6RgVXYvmwg%2FSIit%20AI%20IT%20Agent%20Video.mp4?alt=media&token=4f51dec9-afa3-484e-993a-cd5e7ac67227>" %}

### What it can do

* Proactive resolution: watch for issues and fix them before they spread (e.g., MFA lockout resets with approval).
* Automated actions: run your IT playbooks across tools without manual steps.
* Personalized support: understands your org (people, apps, devices) to tailor responses and actions.
* Effortless approvals: collects approvals from the right approvers in Slack/Teams and proceeds instantly when granted.

<figure><img src="https://451675063-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fz9xmM8V3atukfr2XaElR%2Fuploads%2FA9THsYHw0oVKYVx3JNV9%2Fimage.png?alt=media&amp;token=a17444f7-cf0f-42fa-b834-a357b156d6eb" alt=""><figcaption></figcaption></figure>

### Glossary

* **Agent**: an entity that uses LLMs to follow playbooks, decide next steps, and act.
* **Action**: an operation the agent can execute (Siit‑native, integration, or custom webhook).
* **Playbook**: a user‑defined procedure written in natural language that tells the agent what to do and when.

### Creating a Playbook

How it works You define a playbook in plain language. The agent follows it step‑by‑step, asking for missing information, running actions, and respecting your approval rules.

1. **Condition (the trigger) Describe when the playbook should start.**

* Example: “When an employee asks for new equipment.”
* You can scope by channel, tag, or service to target where it runs.

2. **Instructions Write the steps the agent should follow. Insert actions inline using /.**

* Example structure:
  * Step 1: Understand the reason for the request
  * Step 2: Determine hardware requirements
  * Step 3: /create request for hardware
  * Step 4: /send slack message to employee and their manager with the Request ID<br>

3. **Available actions Use actions inside your steps to let the agent do real work.**

**Core Siit**

* /create request
* /submit request button
* /suggest article
* /send email
* /send slack message via Siit app
* /send teams message via Siit app

**Okta**

* /okta reset multi‑factor (Approval available)
* /okta reset password (Approval available)
* /okta add to group (Approval available)
* /okta add applications (Approval available)

**JumpCloud**

* /jumpcloud add to group
* /jumpcloud reset user password
* /jumpcloud reset user mfa

**Google**

* /google add to group
* /google remove from group
* /google set a temporary password (Approval available)
* /google issue a backup verification code (Approval available)
* /google invalidate backup verification codes (Approval available)

**Microsoft Entra**

* /issue a Microsoft Entra Temporary Access Pass (Approval available)
* /set a temporary Microsoft Entra password (Approval available)

The temporary password, the Temporary Access Pass and the backup verification code are [secret codes](https://docs.siit.io/request-management/secret-codes.md). The agent never reads the value: it places a reference in its reply and Siit shows the real value to that employee only, in a private channel. When the conversation is public, the agent sends the code to the employee's direct message and says so in the thread.

**Slack**

* /slack add user to channel
* /slack make channel private/public
* /slack send message to channel
* /slack archive

**Ticketing and webhooks**

* /zendesk create ticket (Approval available)
* /jira create issue (Approval available)
* /webhook (Approval available)

**Custom actions Extend the agent with your own webhooks.**

* Name and description: make intent clear.
* URL: the endpoint to call.
* HTTP method: POST recommended for sending data.
* Headers/body: add headers or use raw JSON; insert variables from the playbook context (requester, app, device, etc.).

**Approvals Gate sensitive steps with approvals.**

* Choose approver: Manager (resolved dynamically), or Specific employees.
* Condition: First to approve or All users must approve.
* Running mode per action: Auto‑run or Approval required. The agent requests approval in Slack/Teams and continues as soon as it’s granted.

1. Mentions @mention people, teams, or inboxes to keep the right stakeholders informed or to escalate.
2. Articles Attach relevant knowledge to the playbook so the agent can include accurate instructions in replies.

**Governance and safety**

* Audience: restrict where the agent listens and acts (channels, teams, services).
* Sandboxed actions: only whitelisted actions are available to the agent.
* Approvals: mandatory for third‑party changes unless explicitly set to auto‑run.
* Audit: every step (prompt, decision, action, approval) is logged; you can replay runs.
* Gradual rollout: publish playbooks as Draft → Pilot audience → Live. Use “simulate” mode to validate before allowing actions.

**Great use cases**

* Access requests: approval → Okta add to group → notify requester.
* Google Group provisioning: approval → Google add to group → notify requester.
* MFA/Password resets with manager approval.
* New laptop requests: collect specs → create request → inform manager → tag and route.
* Slack hygiene (coming soon): archive channels with approvals; manage membership.

### Attachment understanding

The Triage Agent can now read the files an employee shares during intake not just their text. When someone attaches a screenshot, image, document or spreadsheet while describing a problem, the agent reads and analyses its contents and uses them as extra context to triage the request.

**What this unlocks**

* **Better service matching.** A screenshot of an error screen or an app helps the agent route to the right service, even when the written message is vague.
* **Fuller requests.** Details visible in the attachment (error messages, device info, order numbers) can be used to pre-fill form fields and the description.
* **Fewer back-and-forths.** The employee doesn’t have to re-type what’s already visible in the image or document they shared.

**Example**

An employee posts “this keeps happening” with a screenshot of a VPN error. The Triage Agent reads the error in the image, matches the request to the *Network / VPN access* service, and pre-fills the description with the error text all before an admin sees the request.

Attachment understanding works within the same conversational intake flow, on top of everything the agent already reads (knowledge base, services, form fields, and user data).<br>

### Rollout checklist ✅

* Start with [Triage Agent](https://docs.siit.io/ai/triage-agent.md)
* Define 2–3 high‑confidence IT playbooks (password reset, app access, laptop request).
* Require approvals on any action that changes identity, security, or data.
* Limit agent scope to a few channels and services; expand once KPIs look good.
* Track tags like “AI‑assisted” and dashboards to quantify time saved.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.siit.io/ai/it-agent.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
