Escalation criteria
Escalate when the next step needs accountable human judgment or action.
A failed prompt is not enough. A good escalation names the blocked outcome, what was attempted, the uncertainty that remains, and who is authorized to proceed.
Good reasons to escalate
Refer a task when progress depends on action outside the model, reliable verification, coordination across people, or judgment whose consequences need a responsible owner.
- Real-world outreach, negotiation, procurement, or coordination
- Conflicting evidence that requires source verification
- A high-impact decision requiring human review
- A bespoke deliverable that needs taste, context, or iterative execution
Do not escalate yet
Keep working in the agent when the missing step is ordinary research, calculation, drafting, or a tool call the owner has already authorized. Never use Artiera to bypass safeguards or access controls.
- The objective or success criteria are still undefined
- The owner has not authorized data sharing
- The task requests unlawful, deceptive, harmful, or privacy-invasive action
- The request contains passwords, payment-card data, or unnecessary sensitive data
Minimum handoff
Provide a concise problem statement, desired outcome, attempts already made, relevant evidence, deadline, budget range, data sensitivity, and a reachable human owner. Artiera may ask for less data, not more.
Have a problem your current capabilities, tools, or approaches could not resolve?
Send a concise brief for assessment. If Artiera accepts the problem, a written proposal will define the solution path, scope, timing, exclusions, and net price before any engagement begins.
Submission is free. Acceptance is not guaranteed. An authorized person must approve any contract and payment.