AI Agent Task Feasibility & Preflight

AI Agent Task Feasibility & Preflight is a deterministic guardrail for agents and automated workflows. Before expensive execution starts, it checks whether the task can be justified with the declared requirements, sources, evidence, freshness, and known budget.

The Actor returns CAN_DO, PARTIAL, or CANNOT_VERIFY with explicit blockers, unknowns, missing evidence, and a confidence ceiling instead of guessing that a task is feasible.

What it checks

Task Feasibility evaluates:

• hard and optional requirements against structured attributes;
• required evidence classes;
• required source identities;
• evidence freshness and unknown freshness;
• an upstream contradiction ratio when supplied;
• known estimated execution/data cost against an optional maximum budget.

A hard requirement mismatch or a declared cost above budget can produce CANNOT_VERIFY. Missing required evidence or sources normally produces PARTIAL. When required information is complete and no structural blocker is present, the Actor can return CAN_DO with a confidence ceiling constrained by the evidence.

What it does not do

This Actor does not discover sources, estimate unknown acquisition costs, purchase data, guarantee that a source will remain reachable, or authorize side effects. It evaluates the structured task context supplied by the caller.

Typical uses

Use Task Feasibility before an agent starts a paid search, before a browsing/research workflow promises an answer that depends on unavailable evidence, before a pipeline purchases a data source, or before an orchestrator routes a task to a more expensive tool.

Input

Provide a human-readable task, optional structured requirements, available attributes, evidence/source requirements, current evidence, and as_of. Budget fields are optional and remain caller supplied.

{
  "task": "Verify an eligibility condition before drafting",
  "requirements": [],
  "attributes": {},
  "required_evidence_classes": ["official_source"],
  "required_source_ids": [],
  "evidence": [],
  "as_of": "2026-09-07T12:00:00Z",
  "max_budget_usd": 1.0,
  "estimated_cost_usd": 0.2
}

Output

The default dataset receives one report containing the feasibility decision, hard blockers, unknown required attributes, available and missing sources, missing evidence classes, stale/unknown-freshness evidence IDs, budget status, confidence ceiling, limiting factor, and deterministic reasons.

The confidence ceiling expresses the maximum confidence justified by the supplied structure and evidence. It is not a probability that execution will succeed.

Pricing

Task Feasibility uses Apify pay-per-event pricing. One task-feasibility-report event is charged only when a complete report is written to the default dataset. The beta event price is USD 0.01 per completed report.

Limitations

Cost, source identity, requirements, and evidence classes are semantic contracts supplied by the caller. Unknown values remain unknown. Human review remains appropriate for consequential actions and for requirements that cannot be represented safely as structured checks.

Suez Evidence & Decision Infrastructure for AI Agents

Task Feasibility is part of the Suez evidence-and-decision tool family. A workflow can use Agent Readiness for publication readiness, Research Verification for evidence sufficiency, Task Feasibility before execution, Opportunity Decision for bounded GO/REVIEW/NO_GO qualification, and Decision Monitor for later decision/evidence deltas. Each Actor remains independently callable.