From Manual to Managed: Setting Up an AI-Powered IT Workflow

An IT workflow is managed when requests arrive in one place, routine questions get answered by software, and every action the AI takes runs inside permissions you set once. Three moves get you there:

• Enable AI tooling for the whole organization, with admin control
• Connect it to the systems where IT work already lives
• Put an AI agent in front of the queue, with guardrails on what it can touch

Verified against Claude Team and Enterprise plans and Jira Service Management Cloud on 2026-08-05.

Prerequisites

• A Claude Team or Enterprise plan, with Owner access for the organization-level steps
• Jira Service Management Cloud with a service project and a linked knowledge base, for the AI front door in Step 5
• Admin access to the tools you plan to connect, such as Slack, Google Workspace, or Jira

Step 1: Pick One Front Door for Requests

Decide where IT requests land: a Slack channel, the service portal, or a shared inbox. Every automation below attaches to that intake point, so consolidating on one keeps the rest of the setup simple. If requests currently arrive by hallway conversation and direct message, this step alone recovers real time.

Step 2: Enable Connectors for Your Organization

Connectors are how Claude reaches your business systems. On Team and Enterprise plans, an Owner turns them on for everyone:

1. Navigate to Organization settings → Connectors.
2. Select Browse connectors at the bottom of the page.
3. Select a connector your workflow needs, such as Slack, Google Drive, or Jira, then select Add to your team.
4. Repeat for each system in the workflow.

Enabling a connector makes it available to the team. It grants nobody access by itself, which is exactly what you want: each person still authenticates individually, and Claude inherits each person's permissions from the connected service. Someone who cannot open a file in the source system cannot reach it through Claude either.

Step 3: Connect and Authenticate Each Tool

Each team member connects the services they will use:

1. In a chat, select the + button in the lower left, or type / to open the menu.
2. Hover over Connectors, then select Manage connectors.
3. Select the + button next to Connectors, choose the service, and select Connect.
4. Follow the authentication prompts to grant access.

Step 4: Set Tool Permissions Before You Automate

This is the step that separates a managed workflow from an experiment. Owners can restrict what every connector is allowed to do, org-wide:

1. Navigate to Customize → Connectors.
2. Select a connector to open Tool permissions.
3. Permissions are grouped by type, such as read-only tools and write or delete tools. Set each category to Always allow, Needs approval, or Blocked.

A sensible starting policy: read actions on Always allow, write actions on Needs approval. The AI can research, summarize, and draft freely, while anything that changes a system waits for a human click. Individual users cannot override these settings, and they are best configured for scale from day one, before the first automation ships rather than after.

Step 5: Put an AI Agent in Front of the Help Desk Queue

With the plumbing in place, add the piece that answers routine requests on its own. In Jira Service Management:

1. Open the Studio app, select Agents, then Create an agent, and select Virtual service agent.
2. Connect the agent to the intake channel you chose in Step 1.
3. Activate AI answers: select Manage in Studio, then Knowledge, turn the toggle on next to AI answers, and select Activate.

AI answers searches your linked knowledge base and resolves the questions it can, in every connected channel, immediately. Requests it cannot answer become tickets for your team, so people spend their time on the work that needs judgment.

Verify the Workflow

• In a fresh Claude chat, ask for something from a connected system, such as a summary of a document in Google Drive. Confirm it returns real data.
• Trigger a write action and confirm it pauses for approval, matching the policy from Step 4.
• Submit a test request your knowledge base covers and confirm the virtual service agent answers it. Submit one it does not cover and confirm a ticket reaches a human.

Troubleshooting

• Team members cannot see a connector you enabled: enabling at the organization level makes a connector available but grants no access. Each person authenticates individually under Customize → Connectors.
• Authentication keeps failing: disconnect and reconnect the service from Customize → Connectors, and confirm the account has an active subscription with the service.
• AI answers stays silent: confirm a knowledge base is linked to the service project and the relevant articles are viewable by portal customers.

Would an IT queue that mostly runs itself free up your week? ScaleIt configures and manages AI-powered IT workflows for startups and SMBs. Book a free call and we will build this around the tools you already use.

Verified against Claude Team and Enterprise connector documentation and Jira Service Management Cloud virtual service agent documentation on 2026-08-05. Vendor docs: support.claude.com/en/articles/11176164, support.atlassian.com/jira-service-management-cloud/docs/set-up-your-virtual-agent-in-jira-service-management.