ROUTEBusiness workflows
Jev Business Opportunity Screening
Jev was used as a preliminary screening tool for business opportunities, retaining all cases regardless of their initial 'worth keeping' or 'reject' labels.
I tried Jev as a cheap first pass for screening business opportunities. It kept every case labelled worth keeping. It also kept every case labelled reject.
I built a benchmark from 50 historical cases: 35 for development and 15 for validation. I froze the routing policy before the final check, though those 15 cases had already been used in an earlier version, so this wasn't a fresh holdout.
The policy routed four of the 15 cases to a cheaper check. It never chose human review: seven of the eight cases needing a person went to deeper AI review, and one to the cheap check. Only eight of nine cases needing either deep AI or human review reached one of those routes, below my predefined 90% threshold.
I'm setting aside this business-screening setup. I'd like to try Jev next on narrower data-quality checks and routing tasks where I can define the decisions more clearly.
typesafe.ai/blog/introducing-system-one-mo