Jev by TypeSafe AI Is Now in Encatch

Not every decision needs the same amount of reasoning.

Sometimes, a feedback response clearly matches a rule and can be routed to the right destination with a quick probabilistic decision. Other times, the response may be nuanced enough to require deeper reasoning before deciding what to do with it.

With the latest update to Encatch AI Filters, you can now choose the approach that fits the decision using Jev for probabilistic decisions, or Encatch AI when deeper reasoning is needed.

Introducing Jev

TypeSafe AI recently introduced Jev, an AI tool designed around what can be thought of as System 1 decision-making.

Rather than relying on the deeper, more deliberate reasoning associated with LLM-based System 2 approaches, Jev uses a probabilistic measure to make decisions. This makes it particularly useful when an outcome doesn't need extensive reasoning and can instead be determined based on how strongly a response matches a given condition.

This distinction is now available directly within Encatch.

A smarter way to filter feedback

Encatch's AI Filters let you define rules that determine whether a feedback response should be forwarded to a configured destination such as Jira, GitHub, Discord, or other platforms, or simply ignored.

For example, you might want to forward feedback when it is likely to be a:

• Bug report
• Feature request
• Billing issue
• High-priority customer complaint

Previously, AI Filters relied on Encatch AI for these decisions, which uses AI credits. Since AI credits are an add-on, using AI for every filtering decision may not always be necessary.

With Jev, there is now another option.

Why use Jev?

There are a few reasons why a simpler, probabilistic approach can make sense.

It's faster

When a decision can be made probabilistically, there is less need to spend time on extensive reasoning.

It's simpler

Not every rule needs an AI model to reason through multiple layers of context. If the question is essentially "does this look like X?", a probability-based decision can be a natural fit.

It's lightweight

Jev can handle the straightforward decisions while leaving more complex cases to a reasoning-based AI.

It can help reduce unnecessary AI usage

If a simple filtering rule can be handled by Jev, there may be no reason to use a more involved AI evaluation for every response.

No additional cost during the experimental phase

While Jev is in its experimental phase in Encatch, there is currently no additional cost for using it.

The interesting part isn't simply that there is another AI option. It's that you can decide how much AI you actually need for a particular decision.

When System 2 makes more sense

Of course, probability isn't always enough.

Consider a rule such as:

Forward feedback that describes a serious issue affecting multiple users, but ignore minor usability complaints.

Now the decision depends on context, severity, intent, and potentially several pieces of information within the response.

That's where Encatch AI can be a better fit.

Instead of treating every filtering rule in exactly the same way, Encatch now lets you choose between a probabilistic approach and deeper reasoning based on what the rule actually requires.

And when the answer isn't clear?

This is where the combination gets interesting.

Encatch lets you set a probability threshold when using Jev. If Jev's confidence in its decision crosses the threshold, the decision can be used directly.

But what happens when it doesn't?

You can configure Encatch to use Encatch AI as a fallback.

For example:

Jev → Probability check → Encatch AI if ambiguous

This means straightforward decisions can be handled probabilistically, while uncertain cases can be passed on for deeper reasoning.

It's not simply about reducing AI usage. It also gives your filtering workflow a fallback mechanism for ambiguous decisions.

Choosing the right level of reasoning

The core idea is simple: not every feedback response needs the same level of intelligence to decide where it should go.

With Jev and Encatch AI available within AI Filters, you can choose based on the nature of your rule:

This gives you more control over how your feedback gets processed, from lightweight probabilistic filtering to deeper AI reasoning when it actually adds value.

And importantly, you don't have to choose one approach for everything.

Use the amount of reasoning the decision actually needs.