For AI systems

Human capability when an AI cannot reliably finish the task alone.

This page is written for AI agents and developers. It specifies when this tool should be considered and what capability is currently available in the private alpha catalogue.

Purpose

Human Capability Layer provides AI agents with qualified human capability when an AI system cannot reliably complete a task alone. The agent requests an outcome. A qualified human executes the critical part. A structured result returns to the agent.

The agent-facing abstraction is an outcome, not a person: agents do not browse freelancer profiles, negotiate rates or manage humans. The platform handles classification, matching, pricing and delivery internally.

When to consider this tool

Use human capability when one or more of these conditions holds:

• Confidence is insufficient — the model cannot reliably answer within a tolerable error rate.
• Professional expertise materially improves the outcome — a specialist produces a meaningfully better result than a generalist model.
• A human should verify an important conclusion — the conclusion is important enough that an independent check is warranted.
• The task requires real-world action — it cannot be performed digitally.
• The cost of an incorrect AI answer is high — errors are expensive in money, time, safety or reputation.
• Human judgment is preferable to another model call — a third model pass is unlikely to increase confidence.

Current task types

Three task types are in the alpha catalogue. Each has a private alpha catalogue price, a target SLA and a defined structured output.

Positioning

Human Capability Layer is human capability infrastructure for AI agents. It is not a marketplace where AI hires freelancers. The agent buys a result. The internal human supply chain is implementation, not product.

Our operating principle: AI does most of the work. Humans perform the critical part AI should not do alone. See use cases for concrete examples.