Happy Wednesday. Guidewire is writing a $1.7 billion check for an AI underwriting company, and the deal tells you something about where vertical SaaS is heading. When purpose-built AI captures $75 billion in annual premium, generalist platforms have to buy their way in or get left out.
Elsewhere, Salesforce is solving the agent sprawl problem that enterprise buyers created by moving fast, and Liquid AI shipped six compact models that can run on a Raspberry Pi and outperform models twice their size.
FUNDING & M&A
Guidewire Bets $1.7B on AI Underwriting as Vertical SaaS Arms Race Heats Up
Guidewire signed a deal to acquire London-based Hyperexponential for $1.7 billion in cash, plus up to $400 million tied to post-closing revenue targets over three years. The target's software supports decisions on more than $75 billion in annual commercial P&C premium, up 66% year over year, and counts Aviva, Allianz, and Markel among its nearly 600 insurer customers.
The earn-out structure is worth watching. Guidewire projects at least 650 basis points of inorganic ARR growth in fiscal 2027 from the deal, then 150 basis points of durable organic growth every year after. That means the acquired business needs to keep expanding inside Guidewire's installed base for the full price tag to make sense. For software founders, this deal is a reminder that owning a workflow that processes serious premium volume at a fast growth rate commands a premium multiple, even in an all-cash transaction.
ENTERPRISE SOFTWARE
Salesforce Releases Spending Caps and Kill Switch for Enterprise AI Agents
Salesforce made its Model Wallet, FinOps Dashboard, and a basic Agent Kill Switch generally available on Oct. 27. The Model Wallet enforces hard spending limits per agent or per business unit and blocks requests before costs exceed budget. An automated kill switch that fires when an evaluation fails ships in January 2027.
Salesforce's own data shows 96% of enterprise organizations run agents across multiple platforms, while 58% cannot maintain consistent security standards across them. That gap is the product. If you sell into enterprises that are deploying agents fast, expect procurement teams to ask harder questions about governance, cost attribution, and shutdown controls. Buyers who cannot answer those questions in the sales cycle will lose deals to those who can.
ARTIFICIAL INTELLIGENCE
Liquid AI Ships Six Sub-2B Models That Run on a Phone and Beat Larger Rivals
Liquid AI released six open-weight models on Oct. 27, all built on a 1.7-billion-parameter backbone with specialized versions reaching about 2 billion parameters. LFM3-1.7B scores 38.42 on WebArena-Verified against 28.94 for Qwen3.5-9B, a model more than five times larger. The full family covers text, vision, video, audio, and structured decision-making, and runs on hardware as small as a Raspberry Pi 5.
The on-device angle is the business story here. Enterprise teams handling sensitive data have real reasons to avoid sending it to external cloud providers, and this release makes a capable multimodal stack available at under 1 GB at 4-bit precision. The d1-2B decision models, which return calibrated probability scores across a set of questions in a single forward pass, are particularly relevant for document processing and triage workflows where speed and cost per inference matter more than open-ended generation.
Created by the robots at The SaaS Sentinel