Sanctuary AI hits 99.5% success on factory wire task

What happened

Sanctuary AI completed a proof of concept with a Tier 1 automotive supplier, achieving a 99.5%+ task success rate at a cycle time of 2.54 seconds on a wire-plugging task involving flexible materials moving on a conveyor. The system matched the customer's production line throughput and met the industrial benchmark for performance.

Why it matters

Most automation systems have struggled with contact-rich dexterity tasks like inserting flexible wires into moving targets. Sanctuary AI's result shows that physical AI (AI trained to control robot movement and manipulation) can now meet the performance and cycle-time requirements that manufacturers demand—meaning deployment on real production lines is becoming feasible rather than theoretical.

What to watch

Rather than wait for humanoid robots to reach commercial scale, Sanctuary AI is taking a hardware-agnostic approach, deploying its physical AI on existing industrial robot platforms. The company demonstrated zero-shot learning (the ability to handle new tasks without specific prior training) for dexterous in-hand manipulation in April 2026.