Support workflows

Give your AI support agents an outcome to carry through: resolve the ticket, restore the service, fulfil the request, or escalate with the evidence needed to finish. In Studio, digital workers run the configured cloud workflow, coordinate human decisions, and produce a record of what happened. Your team uses the desktop workspace for investigation and infrastructure work that needs an operator. These examples show the full support loop across MSP, ISP, and voice operations. A server fault starts with a customer’s ticket and ends with verified recovery and a customer update. An upstream outage includes carrier coordination and subscriber communication. A voice complaint includes a working test call before the ticket is resolved. Org Brain gives that work a memory: eligible prior knowledge informs the investigation, and reviewed lessons improve the next response.

​

Build around the support outcome

An agent’s assignment should name all seven parts. “Summarize this ticket” is one step; “investigate this ticket, carry out permitted remediation, and return a verified customer resolution or a complete escalation” defines a support outcome.

​

Choose a workflow

​

Agent work and operator work fit together

The scenario walkthroughs include both cloud agent work and technical recipes for your team. Configure the supported portions as a worker workflow; use the desktop for steps that need local access, a live terminal, a browser session, or an engineer’s judgment. The run should make each handoff explicit so someone owns the next action and knows what evidence to return.

• Digital workers carry the configured work across triggers, service connections, workflow stages, human tasks, and outcome review.
• Org Brain supplies reusable organizational knowledge. Worker knowledge search currently retrieves eligible organization-visible sources; channel or private material is not automatically available to a worker.
• The desktop workspace and Copilot help operators investigate live systems, run procedures, inspect browser sessions, and perform authorized infrastructure changes.
• Channels, work items, and artifacts keep the request, decisions, findings, and result accessible to the people handling the case.

​

Check the systems each workflow needs

These are example operating designs, not installed agent templates or a list of bundled integrations. A named ticketing, monitoring, vendor, or messaging system identifies a dependency to configure and verify. A desktop connector or MCP tool does not automatically become a cloud worker tool, and workers do not inherit the desktop’s SSH sessions, browser sessions, RDP access, or Key Chain credentials. Use worker Setup to check available tools and configure service connections, allowed endpoints, and workflow grants. Production connector calls reach public HTTPS endpoints; private-network targets are rejected. The agent can execute a permitted change on your or your customer’s service through that API, after an independent authorized manager approves the exact request. This approval also applies to ticket replies, customer email, and ticket closure. If a required action is unavailable, make it a human task with a clear result to return. Native Slack routing supports worker questions, approval cards, and outcomes where configured. Teams messages and customer email can use separately configured Microsoft Graph, Outlook, Gmail, or help-desk connections, with the required worker account, scopes, grants, and write approval. Native Teams approval routing remains off. Native inbound-email triggers are not enabled in production, and Studio’s SES member-notification service has separate rollout restrictions. See digital workers for current availability and connectors and MCP for desktop integrations.

​

Start with a complete, measurable loop

Choose a recurring request with a clear success condition and a bounded set of actions. Begin with human review at consequential decisions, then use run evidence to refine the workflow.

1. Customer service account change: read the ticket, execute an approved API change, verify the account state, send the approved reply, and resolve the case.
2. Dell server disk failure: receive the ticket, confirm the fault, coordinate replacement, verify storage health, and update the customer.
3. Microsoft 365 outage: identify the shared incident, keep affected customers informed, verify recovery, and resolve only the tickets actually covered by it.
4. Upstream provider outage: correlate alarms, open a carrier case, coordinate remediation, and close every communication path after recovery.
5. One-way audio: collect call evidence, arrange the approved infrastructure fix, prove two-way audio, and complete the customer response.

For each, measure whether the request reached a verified resolution or a useful escalation. Include the action result and customer communication in the outcome.