Skip to main content

Review and Approve Agent Work

Use this workflow when PoolPM agents prepare work that may affect jobs, customers, schedule, procurement, finance, or leadership review. The userbase is a builder team that wants AI leverage without losing human control.

Workflow outcome

At the end of this workflow, every agent-prepared item should be approved, rejected, held for more information, or routed to the right source workflow with a clear owner.

Agent Board showing AI agent tasks, approvals, status, meetings, and transcripts
Agent Board is the approval and audit surface for AI-prepared work. Use it before letting agent output affect real operations.

Step 1: Open Agent Board

Open Agent Board when DispatchIQ, PulseIQ, C-Suite, or a workflow page shows agent-prepared work.

Check:

  1. Which agent prepared the work.
  2. Which workflow or record it affects.
  3. Current status.
  4. Required approval.
  5. Supporting transcript, meeting, or task context.
  6. Whether the action is reversible.

Step 2: Read the task in business terms

Do not approve a task because the agent sounds confident. Read it as an operator:

  • What is the agent trying to change?
  • Which customer, job, vendor, subcontractor, or payment does it affect?
  • Does the source record support the recommendation?
  • Is any risk missing?
  • Is the next action clear to a human?

Step 3: Verify source context

Open the source workflow before approving anything important.

Agent work typeSource check
Schedule recommendationJob detail, CascadeIQ, Build Calendar, procurement readiness.
Subcontractor packageSubcontractor record, job scope, SOW package, assignment.
Purchase orderProcurement, vendor, price book, PO lifecycle.
Payment actionFinancials, source transaction, job completion, approvals.
Customer responseCustomer service, callback, meeting notes, ticket history.
Leadership summaryDispatchIQ, PulseIQ, C-Suite, source records behind the summary.

Step 4: Choose approve, reject, or hold

Use the action that matches the evidence:

DecisionUse when
ApproveSource context matches, risk is understood, and the action should proceed.
RejectThe recommendation is wrong, unsafe, unclear, or based on missing context.
HoldThe recommendation may be right but needs another record, person, or document.
RouteThe work belongs in a source workflow before it can be decided.

Always include a reason for rejection or hold. That reason becomes training context for the team and audit context for the company.

Step 5: Confirm downstream effect

After approval, verify the related workflow changed as expected:

  1. Job note, schedule, package, PO, ticket, payment, or callback reflects the action.
  2. The owner can see the result.
  3. The homeowner, subcontractor, or vendor is not notified prematurely.
  4. DispatchIQ no longer shows the item as unresolved unless follow-up is still needed.

Step 6: Escalate uncertain decisions

Escalate instead of approving when:

  • Financial or legal risk is unclear.
  • Homeowner-facing wording could create a commitment.
  • A schedule action depends on missing materials.
  • A subcontractor assignment lacks scope or compliance context.
  • Agent output conflicts with the source record.
Portal areaUse it for
Agent BoardAgent task review, approvals, meetings, transcripts, and audit context.
DispatchIQFinding agent work during daily triage.
PulseIQ / C-SuiteLeadership-level summaries and escalations.
Jobs / Procurement / Financials / Customer ServiceSource workflows that prove whether agent work is safe.