UiPath adds TypeSafe connector to run Jev decision models in workflows

Small decisions, big costs

UiPath argues that every automated process is full of small judgment calls: which team should handle a ticket, whether an invoice is disputed or just late, or whether an extracted value can be trusted. Today, the company says, each of these calls is sent to a large language model (LLM) or an agent, and at scale that costs time and money in the form of latency and higher prices.

The company points out that LLMs and agents suit reasoning, planning and open-ended work, but many workflow steps are bounded decisions: pick one of a few options, place something on a scale, or answer yes or no. In enterprise processes these decisions are often needed for thousands of records. Some are clear enough to automate, while others need human context and guidance.

UiPath is partnering with TypeSafe to bring its Jev models to UiPath developers and enterprise customers as a response to this problem.

What the connector offers

The TypeSafe connector is available in preview in UiPath Integration Service. Users connect it with their own TypeSafe API key, after which Jev activities appear in their workflows. UiPath says they can be tried in UiPath Maestro Automate, Maestro Orchestrate and any UiPath surface that supports connectors. Searching for "Jev," "TypeSafe," or "System One" finds them.

Because the connector uses the customer's own TypeSafe key, requests are governed by the customer's direct agreement with TypeSafe and are not routed through the UiPath AI Trust Layer, according to UiPath.

Jev returns a typed answer, the probability of each option, and its confidence in the decision. Developers can set confidence thresholds for automated action. When confidence falls below the threshold for a task, a UiPath workflow can bring in a human to review the information, add missing context and guide the decision before the process continues. UiPath describes this as a "two-brain-system" implementation.

How Jev works

With Jev, a developer passes a state, which is the JSON context for the decision, and a set of questions. Each question is one of three types:

• Choice: which of these options applies?
• Score: which level on an ordered scale?
• Noul: how true is this statement, from 0 to 1?

Every answer is constrained to the options the developer defines and comes back with per-option probabilities and a confidence score. Several questions about the same state can be answered in parallel in a single request. UiPath frames Jev as fast thinking for bounded decisions, while LLMs and agents continue to handle deliberate, multi-step reasoning. The UiPath Platform can orchestrate both in series or in parallel.

Erik Gafni, CTO of TypeSafe, said most decisions inside a business process "don't need a model to write a paragraph." In his words, they need "the right answer, in the right shape, with an honest signal of how sure the model is."

Early internal results

UiPath says it tested Jev and other decision models on real workloads, including grading agent output, extracting fields from documents, classification and re-ranking search results. The company describes these as early internal experiments and reports four findings:

• Jev showed 94% agreement with an LLM-as-a-judge system in evaluation datasets.
• Confidence thresholds let teams escalate tougher decisions to reasoning agents or humans while keeping fast throughput in most cases.
• UiPath's internal evaluation showed similar decision making at 7.7x lower cost and 4x-20x faster than the equivalent LLM judge.
• Document extraction can be treated as retrieval and choice rather than generation: one Jev call finds which part of a document holds each field, and a second picks the exact value.

These figures come from UiPath's own testing and have not been independently verified in the source.

What it means for developers

For teams building automations on UiPath, the connector offers a way to hand routine classification and scoring steps to a specialized model instead of a general LLM, with a built-in route to human review when confidence is low. Since it is a preview and relies on a customer's own TypeSafe tenant and key, users need an existing agreement with TypeSafe. UiPath points developers to the TypeSafe documentation for details on Choice, Score and Noul questions, and invites them to share their work on the UiPath Community Forum.