Anthropic Researcher: AI to Automate 95% of Computer Jobs by 2028

A senior Anthropic researcher forecasts that by 2028, frontier models will automate 95% of computer-facing labor, with residual human oversight persisting into the next decade. This is a capability timeline, not a market prediction, but it anchors planning cycles for anyone deploying AI in knowledge-work pipelines.

For operators, this compresses the depreciation window for human-centric workflow design. Any system architecture built today around human-in-the-loop review should assume that review layer becomes a bottleneck or liability within three years. The strategic shift is toward autonomy-grade tooling: models that not only execute tasks but validate their own outputs against business rules. Infrastructure spend should prioritize observability and rollback mechanisms over staffing for QA. Teams that currently monetize “AI-assisted” services will find that margin erodes as the baseline capability becomes commoditized; the durable edge becomes owning the data and the evaluation harnesses that verify autonomous execution. Expect a second-order effect: the scarce human role shifts from task execution to exception handling and system definition, which changes hiring criteria and internal training programs accordingly.