🔍 Read the full analysis: Could Future AI Be Defined By Systemic Functionality Over Sentence Writing? on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
TypeSafe AI has launched Jev, an AI model prioritizing decision-making through structured, typed responses rather than generating free-form text. This shift could redefine how AI is integrated into enterprise systems, emphasizing efficiency and reliability.
On September 15, 2026, TypeSafe AI unveiled Jev, a groundbreaking AI model that departs from traditional text generation by producing typed decisions with probabilities, designed for automation within software systems. This development challenges the prevailing assumption that large language models (LLMs) should primarily generate human-like text, signaling a potential paradigm shift in enterprise AI applications.
Jev is part of TypeSafe’s broader initiative to develop System One Models, a new class of AI optimized for decision-making rather than language generation. Unlike conventional LLMs, Jev accepts structured questions and returns typed responses such as choices, scores, or yes/no probabilities, all accompanied by calibrated confidence levels. It processes multiple questions simultaneously, providing structured outputs that can be directly acted upon by software, eliminating the need for parsing or interpretation of text outputs. The model is claimed to be significantly faster and cheaper—answering in 70 to 500 milliseconds at a cost of approximately $0.042 per million input tokens—making it highly suitable for high-volume enterprise automation tasks. TypeSafe emphasizes that Jev’s outputs are schema-compliant, which reduces errors caused by formatting issues common in text-based models, although it does not eliminate the possibility of incorrect decisions within the allowed options. The launch was supported by $40 million in funding from DCVC, with Diogo Almeida, a co-inventor of RLHF and InstructGPT, leading the development. While initial benchmarks show promising results, independent testing indicates that accuracy varies depending on question complexity and framing, with some overconfidence observed in probability estimates. The company acknowledges that Jev’s effectiveness depends heavily on how well users design their questions and interpret its structured outputs, marking a shift from reliance on language models to decision-focused AI for enterprise use.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
Implications for AI Development and Enterprise Automation
The introduction of Jev and the emphasis on systemic functionality over text generation could transform how AI is integrated into business operations. By focusing on structured, decision-oriented outputs, companies can automate routine judgments more reliably and at lower cost, reducing dependence on human oversight. This shift addresses longstanding issues with LLMs, such as hallucinations and formatting errors, which have limited their use in critical decision-making contexts. If widely adopted, Jev’s approach could lead to a new standard for enterprise AI, prioritizing speed, cost-efficiency, and schema compliance over human-like language capabilities. Furthermore, this development challenges the dominant narrative that AI must excel at language generation, opening pathways for models optimized for decision-making, automation, and integration within existing software systems. As AI becomes more embedded in operational workflows, the shift towards systemic functionality could enhance reliability, reduce costs, and enable more scalable automation across industries.
enterprise decision-making AI software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Evolution of AI Towards Decision-Centric Models
Over the past three years, AI development has been heavily focused on improving language models’ reasoning, context handling, and code generation capabilities. Major launches, including OpenAI’s GPT series and Anthropic’s Claude, have aimed to produce more human-like text. However, critics have pointed out persistent issues such as hallucinations, overconfidence, and the need for human oversight. In response, some innovators, including Diogo Almeida of TypeSafe, argue that the core assumption—that all AI problems should be addressed via language models—may be flawed. Almeida’s team has developed Jev, a model built explicitly for decision-making within software, which produces structured responses instead of text. This approach is rooted in psychological theories of System 1 thinking, emphasizing rapid, intuitive judgments, and is designed to address the limitations of traditional language models. The launch reflects a broader trend of exploring alternative AI architectures that prioritize reliable, schema-conformant decisions over natural language generation, especially for enterprise automation tasks.
“Jev represents a fundamental shift from generating words to producing typed decisions, making AI more like a function call than a colleague.”
— Diogo Almeida, TypeSafe AI
As an affiliate, we earn on qualifying purchases.
Uncertainties Around Model Accuracy and Adoption
While Jev’s structured decision approach shows promise, questions remain about its accuracy, reliability, and scalability across diverse enterprise contexts. Initial benchmarks indicate variability in performance, with some tests showing lower accuracy compared to other frontier models like Claude or GPT-6 Astra, especially on complex or ambiguous questions. The company admits that Jev’s success depends on how well users craft questions and interpret responses, suggesting a learning curve and potential limitations in generalizability. Furthermore, it is not yet clear how quickly and broadly this approach will be adopted by enterprises, or how it will perform in real-world, high-stakes environments where errors could have significant consequences. Ongoing independent testing and real-world trials are needed to evaluate its robustness and practical utility, and whether it can truly replace or supplement existing language models in enterprise workflows.
As an affiliate, we earn on qualifying purchases.
Next Steps for Adoption and Validation
TypeSafe plans to continue refining Jev through ongoing testing and user feedback, aiming to improve accuracy and reliability. The company will likely release more case studies demonstrating its application in various enterprise scenarios, including customer support, decision automation, and operational workflows. Industry observers will watch for independent evaluations and benchmarks to assess how Jev performs outside controlled environments. Additionally, further development may focus on expanding the range of question types and integrating Jev more seamlessly into existing software platforms. As adoption grows, users will need to develop best practices for question design and response interpretation to maximize the benefits of this decision-centric AI approach. Regulatory and safety considerations may also influence how broadly this technology is deployed, especially in high-stakes sectors.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Jev, and how does it differ from traditional AI models?
Jev is a decision-oriented AI model that produces typed, structured responses with probabilities, rather than generating free-form text. It acts more like a function call within software, enabling faster, more reliable automation.
Why is the focus shifting from text generation to systemic functionality?
This shift aims to address issues like hallucinations, formatting errors, and overconfidence in language models, by providing calibrated, schema-compliant decisions that can be directly acted upon by software systems.
What are the potential benefits of decision-focused AI for enterprises?
Benefits include increased speed, lower costs, improved reliability, and easier integration into existing workflows, enabling more scalable and automated decision-making processes.
What challenges remain for Jev’s widespread adoption?
Challenges include ensuring high accuracy across diverse tasks, designing effective questions, interpreting structured responses correctly, and validating performance in real-world, high-stakes environments.
Will this approach replace traditional language models entirely?
It is too early to tell, but initial indications suggest that decision-oriented models like Jev will complement rather than fully replace language models, especially in enterprise automation contexts where reliability is critical.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
