Complete AI Training
Start for Free

Your job's AI kit

Get your AI kit

Tell us who you are and what you do. We show you your kit right away and email you the link: skills, prompts, AI agents, MCP servers and courses for your job.

500+ jobs ready, and we make a kit for any other job. No payment needed to look.

Share

AI news ·

Epoch AI Finds No Model Meets Its Employee Standard

Frontier AI models failed to handle complex workplace tasks end-to-end, though one matched human contractors on a single narrow research assignment.

Epoch AI tested frontier models against its internal benchmark for complex workplace tasks and found that none met the standard required of a full employee. The evaluation covered multi-step assignments that demand sustained reasoning and domain judgment, and every system fell short of end-to-end ownership.

One partial exception emerged in a bounded research setting. Claude Fable 5.2 matched the performance of external contractors on a data-center research task, demonstrating that a frontier system can be competitive when the scope is tightly defined. The finding does not imply the model replaced expert oversight; it performed at parity with human contractors only within that specific, constrained assignment.

The results indicate that current frontier systems can contribute to narrow technical workstreams but do not yet replicate the breadth of an employee operating across varied, complex tasks. For organizations evaluating integration, the data suggests a role for models as targeted assistants on bounded projects rather than as substitutes for expert decision-making and accountability.

Share