Decision Models Surge Across Tech Industry as Microsoft and Startups Chase Customization
Lightweight decision models designed to pick between preset options without generating output tokens are rapidly gaining adoption across enterprise production environments, challenging traditional large language models.
🌐Work & SocietyPhoto: TechCrunch AI
Key points
- Decision models route and evaluate preset choices instead of generating output tokens.
- Microsoft launched Microsoft-Decision-1.1 on Azure to target enterprise efficiency.
- U.K. startup Sqwish raised $3 million to focus on customized post-training tools.
The Rise of Decision Models
Decision models have quickly captured a substantial share of enterprise AI workflows following their sudden emergence. Unlike traditional large language models that generate lengthy text tokens, these specialized systems use their core intelligence to rapidly and affordably select between predefined choices while providing a confidence score. Usage metrics highlight this dramatic shift, with Vercel reporting that more than one in four requests on its AI model router went to decision models in October, despite the technology having been invented just a month earlier.
The market has expanded rapidly from early breakthroughs like TypeSafe's Jev family of models to major hyperscalers entering the space. Microsoft announced the immediate availability of its Microsoft-Decision-1.1 model on Azure, pointing to the intense industry demand for speed, calibration, and cost efficiency in production settings. Meanwhile, early-stage innovators are finding distinct niches. U.K.-based startup Sqwish recently secured a $3 million pre-seed round for its 'kettei fat' model, focusing on built-in tools that allow clients to continuously refine the software using proprietary data.
Customization as a Competitive Edge
Industry leaders emphasize that standard performance benchmarks fail to capture how these tools perform in real-world scenarios. Microsoft engineering vice president Achint Srivastava noted that overly optimized benchmark models frequently struggle under actual enterprise workloads. Instead, developers are increasingly prioritizing tailored accuracy and low latency. By enabling organizations to continually fine-tune decision logic with internal data, startups aim to carve out a sustainable advantage against dominant hyperscalers and frontier labs.
- Vercel reported over 25% of router requests went to decision models in October.
- Microsoft launched its Microsoft-Decision-1.1 model on Azure.
- U.K. startup Sqwish raised $3 million in a pre-seed funding round.
What to watch next: Watch for enterprise adoption rates of Microsoft-Decision-1.1 on Azure and how smaller startups differentiate through proprietary post-training data tools.
Source TechCrunch AI — techcrunch.com/2026/11/17/as-decision-model-use-skyrockets-is-customiza… ↗
Read next
🏢Big TechAWS Expands Forward Deployed Engineering to Help Enterprises Scale AI
Amazon Web Services has expanded its Forward Deployed Engineering program to 15 countries to help businesses transition AI projects from prototypes to production-ready sy…
🤖AI ModelsGoogle Expands Gemini 'Tasks' Feature to Automate Daily Errands
Google is rolling out its 'Tasks' feature to a wider user base, allowing the Gemini AI to perform autonomous actions like booking appointments and making purchases.
🚀Startups & FundingStarSling Launches AI-Native CI Platform to Speed Up GitHub Actions
Y Combinator-backed startup StarSling has introduced an AI-native continuous integration tool designed to accelerate GitHub Actions through faster hardware and automated …
🤖AI ModelsAmazon Introduces SMART Toolkit for Integrating WhatsApp AI Agents via Bedrock AgentCore
Amazon Web Services has released an open-source toolkit designed to simplify connecting AI agents built on Amazon Bedrock AgentCore to WhatsApp.
🤖AI ModelsGoogle Extends Temporary Gemini Chat Retention to Two Months
Google is lengthening the retention period for temporary Gemini chats from 72 hours to approximately two months to boost safety and service reliability.
🚀Startups & FundingHarvard Study Reveals US Hyperscale Data Centers Consume Power Rivaling Entire States
A new Harvard study reveals that America's 403 hyperscale data centers consumed between 68 and 99 terawatt-hours of electricity during the 2024-25 study period, rivaling …