UiPath AgentHack · Maestro Case

AgentOps Case Control Room

Turn AI-agent work into a UiPath-governed case with evidence, risk, approval, robot work items, and human handoff.

Case room screenshot
Live demo: https://daideguchi.github.io/agentops-case-control-room/
Problem

AI agents are becoming operational workers, but real work is not a chat transcript.

When an agent touches production systems, teams need to know what happened, what evidence was used, what risk was found, who approved it, and how work can be resumed safely.

?

What did the agent do?

!

Was a risky action blocked?

✓

Who approved the final decision?

Business impact: safer AI operations for release, support, IT, and automation teams.
Solution

A UiPath-governed case for human-AI operations.

  • Human opens an exception case.
  • AI agent creates a read-only investigation plan.
  • Robot gathers ticket, pull request, and owner evidence.
  • Policy gateway blocks risky production action.
  • Action Center-style tasks route approval to humans.
Action Center demo
Core thesis: agents can move fast, but UiPath makes the work governable.
Architecture

Maestro Case

Case stages: intake, agent investigation, robot/API evidence, risk review, approval, handoff.

Robot Work Items

Evidence collection is modeled as queue-ready robot work items.

Action Center

Human approval tasks preserve evidence, risk, and decision context.

Implementation artifacts include a BPMN-style process blueprint, case data model, Action Center form schema, robot queue model, state machine, transaction logs, and simulated case run.

uipath-package/maestro-process.bpmn

runtime/case-state-machine.json

runtime/orchestrator-transaction-log.jsonl

Claim boundary: verified local prototype and UiPath-ready implementation package.
Working Proof

Repeatable local verification is included.

  • 26 shared AgentOps events
  • 3 cases
  • 5 approval gates
  • 1 blocked production action
  • 2 robot work items
  • 3 Action Center tasks
  • 6 Orchestrator transaction events
  • 9 Action Center decision events
  • Final state: ClosedRejected

cd uipath-agenthack

bash scripts/run_uipath_local_checks.sh

verify_ok

uipath_verify_ok

Public GitHub: https://github.com/daideguchi/agentops-case-control-room
Why Now

Enterprise agents need an operating model, not just more autonomy.

AgentOps Case Control Room shows how human, AI, robot, and API work can be reviewed as one case. That is the bridge between a prototype on a laptop and agentic automation that enterprises can trust.

Shared dashboard
Next: import the generated process, queue, and approval models into UiPath Automation Cloud / Maestro and verify a live run.