🔍 Read the full analysis: Can 'System One' AI Revolutionize How We Use Artificial Intelligence? on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
TypeSafe AI announced Jev, a new ‘System One’ AI model designed for decision automation within software. It produces structured, calibrated decisions instead of text, promising faster, cheaper, and more reliable enterprise automation. Its impact and effectiveness are still being evaluated.
TypeSafe AI has introduced Jev, a new class of AI model designed explicitly for decision-making within enterprise software systems. Unlike traditional large language models (LLMs), Jev produces structured, typed decisions with associated probabilities, aiming to replace text generation with faster, more reliable automation. This development marks a significant shift in how AI can be integrated into business processes, potentially reducing costs and increasing decision accuracy.
Jev is built on the concept of ‘System One’ thinking, inspired by psychologist Daniel Kahneman, emphasizing quick, intuitive judgments rather than slow, deliberate reasoning. Developed by Diogo Almeida, a co-inventor of RLHF techniques used in models like InstructGPT, Jev is designed to produce decisions in a schema-conformant, structured format, such as ‘team: billing, confidence: 0.94’, rather than generating free-form text.
The model operates by answering three types of questions—choices, scores, and yes/no probabilities—in a single parallel pass, returning calibrated probabilities that can be directly acted upon by software. Its creators claim Jev can answer in 70 to 500 milliseconds at a cost of approximately $0.042 per million tokens, claiming it is over 193 times faster and 445 times cheaper than traditional LLM workflows for similar tasks.
TypeSafe asserts that Jev eliminates errors stemming from output formatting or hallucinations common in LLMs, although it does not guarantee correctness of decisions, only schema compliance. Benchmark tests show that Jev’s accuracy varies depending on the task, with some tests indicating a 62.6% agreement rate compared to frontier models, and others showing higher performance with narrow questions. The company admits that accuracy depends heavily on how judgments are broken into smaller questions.
Jev vs. LLMs: who should make the call?
Jev, from TypeSafe AI, is a “System One” model. It doesn’t write text. It returns a typed decision with a confidence score that your software can act on directly.
Same support ticket, two kinds of answer
“This ticket appears most likely related to billing, although it could also concern account settings or a recent plan change. I would suggest reviewing the invoice history before…”
A person reads it, or code has to parse the prose.
team: "billing"Software reads it and acts. Nothing to parse.
How they differ
| LLM | Jev | |
|---|---|---|
| Output | Text written for people | A choice, a score or a yes/no probability |
| Speed | Seconds per call | 70–500 ms* |
| Price | Input and (pricier) output tokens | $0.042 per million input tokens, output free* |
| Knows when it’s unsure | Often sounds confident when wrong | Confidence score on every answer |
| Explains its answer | Yes | No, which matters for audits |
| Best at | Reasoning, writing, open questions | Routing, tagging, scoring, duplicate checks |
* Vendor-reported. TypeSafe also claims up to 194× faster and 445× cheaper on its own selected workflows.
Accuracy is something you build
Jev is far cheaper and faster, but not more accurate than frontier models. How you phrase the question matters a lot.
TypeSafe’s benchmark scores agreement with two frontier models rather than verified ground truth. The five-question result used weights fitted on 1,000 labelled examples.
The real idea: a confidence dial you control
“duplicate listing”, confidence 0.62
Raise the threshold for fewer mistakes and more manual review. Lower it for more automation and more risk.
Only use Jev when all four hold
Good fits
- Routing tens of thousands of support tickets a day
- Flagging duplicate listings in a product catalogue
- Replacing a keyword filter that mis-tags half its matches
Poor fits
- Drafting customer emails or release notes
- Reviewing a few high-stakes contracts a month
- Anything that needs a written explanation
Potential Impact on Enterprise Automation
If proven effective at scale, Jev could significantly alter enterprise automation by enabling faster, cheaper decision-making processes that require less human oversight. Its ability to produce structured, calibrated decisions could reduce reliance on costly reasoning models and improve reliability in tasks like support ticket routing, fraud detection, or compliance checks. However, the current benchmark results suggest that accuracy and reliability are still developing, and widespread adoption will depend on how well Jev performs in real-world scenarios.
enterprise decision automation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on AI Decision Models and Industry Shift
Over the past three years, the AI industry has focused heavily on improving large language models, emphasizing capabilities like reasoning, long context handling, and code generation. Companies like OpenAI and Anthropic have led the charge, but concerns about hallucinations, overconfidence, and high costs persist. Meanwhile, enterprise users increasingly seek automation solutions that are faster, cheaper, and more reliable.
TypeSafe’s approach with Jev represents a departure from the dominant paradigm, emphasizing decision accuracy over text generation. The company argues that the same reinforcement learning techniques used to train chatbots are poorly suited for automation, advocating instead for a method called Reinforcement Learning for Calibrated Decisions (RLCD). This shift aligns with broader industry trends toward structured decision-making and formalized AI outputs.
Since its launch with $40 million in funding, Jev has undergone initial testing, with some independent evaluations indicating moderate accuracy and highlighting the importance of task framing. The industry is watching closely to see if decision-focused models like Jev can outperform traditional LLMs in enterprise settings.
“Jev is designed to produce decisions that software can act on directly, not just generate words for humans to interpret.”
— Diogo Almeida, co-inventor of RLHF
structured decision-making AI tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Jev’s Effectiveness
It remains unclear how Jev will perform across a broad range of real-world enterprise tasks, especially in complex or ambiguous scenarios. Benchmark tests show moderate accuracy, but real-world environments often present unpredictable challenges. Questions about long-term reliability, integration complexity, and how well Jev handles edge cases are still open. Additionally, the impact of overconfidence in probabilities and how that influences automated decision pipelines needs further investigation.
As an affiliate, we earn on qualifying purchases.
Next Steps for Adoption and Evaluation
TypeSafe plans to expand testing of Jev in various enterprise settings, including support automation, fraud detection, and compliance monitoring. Industry observers will be watching for independent validation of its accuracy and reliability in live environments. The company is also likely to refine its training techniques and improve calibration methods. Widespread adoption will depend on demonstrable improvements over existing models and clear integration pathways into enterprise workflows.
decision automation software for enterprises
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How does Jev differ from traditional language models?
Jev produces structured, typed decisions with probabilities, rather than generating free-form text, making it more suitable for automated decision-making within software systems.
What are the main advantages of Jev?
It offers faster response times (70-500 ms), lower costs (around $0.042 per million tokens), and reduces errors related to output formatting and hallucinations.
What are the current limitations of Jev?
Its accuracy varies depending on task design, and it does not guarantee correct decisions—only schema adherence. Its performance in complex, real-world scenarios remains to be fully validated.
Will Jev replace existing AI models in enterprises?
It is too early to tell. While promising, Jev’s adoption will depend on its proven reliability and integration ease compared to traditional models.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
