jevVIDEO

Jev matches LLM agents on routing but fails invoice math in UiPath test

In tests on four UiPath tasks, Jev matched an LLM agent's routing accuracy at about 7x the speed but approved 3 overbilled invoices, leading to a rule of computing in code, letting Jev decide, and keeping humans on irreversible steps.

SNOK

@SNOKSECArchived 2026.09.29Originally posted 2026.09.28

SNOK @SNOKSECOriginally posted 2026.09.28

Open original on X

We tested Jev on 4 tasks from our UiPath projects. Routing: as accurate as an LLM agent, ~7x faster. Invoice checks: it approved 3 overbilled invoices - it doesn't do math. Rule: extract, compute in code, let Jev decide, humans own irreversible steps. https://snok.ai/en/news/blog/jev-uipath-maestro-trust-layer-delegate/?utm_source=x&utm_medium=social&utm_campaign=jev-uipath&utm_content=snoksec

x.com/snoksec/status/2104483818941870258Archived 2026.09.29 10:06