🚀 a16z: Can Agents Use a Computer Yet? We've Got the Data
a16z surveyed operators running computer-use agents in production and found the leap from demo to deployable happened fast.
• Best models jumped from 42% to 85% task success on the OSWorld-Verified benchmark in a year — above the ~72% human baseline.
• Running an agent costs roughly $6–8/hour, break-even with offshore BPO (~$10/hr) and a 70–80% margin vs. US back-office labor (~$30–45/hr).
• The moat is shifting from the model to context — verification, escalation, and company-specific workflow knowledge now separate winners from demos.
🔗 Read more
#a16z