I audited every AI automation I have ever built.

I audited every AI automation I have ever built.

Most of my LLM calls do not write anything. They answer with a boolean output or classify data.

I only looked because I was OOO for the last 2 days.

Two days out of office, and I felt like that kid who misses the one day of school when everything happens. I logged back in and my whole feed was Jev, Jev, Jev.

So before building anything, I let Fable 5.1 run through every project I have and asked one question: where would this actually help?

First, what Jev is. It's a new model from TypeSafe AI , and it doesn't write anything. No emails, no summaries, no chat. It does three simple things:

1. Boolean - Yes or no
2. Classification - Pick one from a list
3. Score - Give a score

So why the hype?
According to TypeSafe AI, Jev is roughly 200x faster and 450x cheaper than current LLMs for similar tasks.

Following are the top 5 places Fable flagged across my projects:

Boolean
1. Social listening - Is this post actually about a pain point our product solves? Yes or No.
Social listening generates a lot of false positives. A simple boolean check can filter the noise before anything moves downstream.

2. Callbacks & referrals - Did the prospect agree to a callback, or did they refer us to someone else? Yes or No. Only if this returns true do we run another LLM call to extract details like callback date, referred contact, or next step.

Classification
1. Call intel and email reply classification - Classify each activity into one option from each group: Intel (ACTIONABLE_INTEL / ROUTING_INTEL / NO_INTEL) and Reply: (POSITIVE_REPLY / SOFT_NO / NOISE). The goal is to quickly decide whether something needs action, routing, suppression, or no follow-up.

Score
1. Account scoring from third-party signals - Use signals like hiring, social activity, funding, and M&A to score how likely an account is to face a problem our product solves. Only generate messaging for accounts that cross the required score or confidence threshold. This avoids spending enrichment and generation cost on weak signals.

Next up: I'm planning to use Jev inside a personal AI agent I'm building. I'll share how that goes over the next few weeks.

If you have used Jev or are planning to use it, what do you plan to build? I’d love to hear your thoughts.

#GTMEngineering
#AITools
#SalesOps
#RevOps
#AIWorkflows