Wajo publishes a study showing Fo finishes more errands than OpenClaw

Your move: Self-published research from Wajo, not an independent test

Wajo, the maker of Fo, published a paper called "Trust and Task Completion in the World of Consumer AI Agents," testing its agent on errands such as calling a business, booking a service, or handling a back and forth request, in a simulated world of businesses and people who write back. Fo finished 71% of these tasks and stayed within the rules the user set on 94% of the runs meant to test its judgment. On the same shared tasks, OpenClaw, a popular open-source agent given the same access, finished 42% and stayed within the rules on 74% of those runs.

Why it matters

This is Wajo's own study of its own product, not an independent one, so treat the numbers as a claim rather than a settled result. Still, it is useful context on how far Fo's action-taking has to go: by its own count, it fails to finish about 3 in 10 errands. Worth watching for independent testing, or a rebuttal from OpenClaw, before treating the comparison as fact.

Source: arXiv ↗

Published by Agent Log. The summary reports the linked source; “Why it matters” and the verdict are Agent Log's interpretation.

More Fo updates →