Build Workflows with Admin Assist

Admin Assist in Atomicwork allows you to create and edit workflow automations using plain English. You describe what you want, and the Universal Agent, Atom, builds, validates, and deploys the workflow for you without any manual drag-and-drop configuration.

How Admin Assist works

When you create a new workflow, you will see a canvas with options to add a manual trigger or build with AI.

1. Click Build with AI on the workflow canvas to open the chat sidebar.

2. Describe your workflow in natural language when prompted.

3. Review the structured plan Atom creates, which includes a trigger and action steps. You can confirm, cancel, or ask for changes.

4. Answer follow-up questions during the data collection phase to fill in specific attributes like tags, and groups.

5. Wait for Atom to compile the plan into a deployable workflow graph. Structural and pre-deployment validation catches errors before anything goes live to avoid runtime issues.

6. Once deployed, you will see a success message in the chat.

Admin Assist appears in more than one workflow-building context. You can use Build with AI to construct complete workflows, or use the embedded assistant in the Execute Code action to generate JavaScript or TypeScript for a single workflow step.

For detailed setup and testing instructions for code generation, see the Execute Code action guide.

Sample usecases

Here are some usecases that our customers have built using Admin Assist:

1. Analyse the request to see if the requester is frustrated and if they are frustrated, set the priority to high. Else set priority as medium.

2. If the last reply to the request is a thank you message, don't reopen the ticket. Add a private note to let other agents know to ignore it because it's a thank you. If it's not a thank you, set the status of the ticket to open.

3. Check if request type is an incident. If so and if the description or subject mention it to be a wifi issue, set priority to high and assign ticket to a teammate.

4. Schedule a workflow at 9am to check if any high priority tickets are close to violating the Resolution SLA. If so, escalate it to a supervisor through email or Slack / Teams.

Editing existing workflows

Once a workflow is deployed, you can edit it via chat. Edits go through a confirm, cancel, or modify loop where Atom shows you what will change. Supported edit operations include:

• Rename a step.

• Update configuration details like field values or message text.

• Change conditions on a filter.

• Insert a new step after an existing one.

• Delete a step.

• Modify the workflow trigger.

• Rebuild the workflow from scratch.

Troubleshooting common issues

Deployment failed

This indicates the workflow passed internal validation but the platform rejected it. Try rephrasing your prompt with more specific values that match your actual workspace options.

Validation failed

This means structural validation caught an issue, such as an unreachable node or a broken connection. Simplify your prompt and break complex workflows into smaller pieces.

Configuration missing

If a node shows missing configuration after building, click on the misconfigured node in the canvas and manually fill in the missing attributes before publishing.

Build seems stuck

The pipeline has checkpoints at every phase. Use the Stop button on the overlay to cancel, then try again.

Tips for success

• Be specific in your prompt. Instead of asking to notify someone, specify the exact role and priority level.

• Review the plan carefully before confirming deployment.

• Mention scheduled explicitly if you want a workflow that runs periodically.

• Make one edit at a time when modifying an existing workflow.

• Remember that you can always click any node to inspect or tweak its configuration manually.