Anthropic's Claude AI Boosts Robotics Efficiency and Handles 95% of Internal Analytics

In August 2025 Anthropic ran Project Fetch, an experiment where employees used a robotic quadruped (a “robodog”) to perform tasks with the assistance of its Claude Opus 4.1 language model. The team equipped with Claude completed tasks substantially faster than the team without it. In a Phase Two repeat, the newer Claude Opus 4.7 operated without human assistance and was up to 20 times faster than the fastest human team, though it still struggled with the precise “fetching” of a beach ball.

Anthropic also reported that Claude now answers roughly 95 % of the company’s internal analytics queries, achieving about 95 % aggregate accuracy. "95% of business analytics queries are automated via Claude, with ~95% accuracy in aggregate," the firm said, noting that encoding repeatable analytical workflows as skills and using a governed semantic layer has freed data‑science staff for more strategic work such as causal modeling and forecasting.