AI models & tools
MCP Is Turning Compliance Software Into Infrastructure for Business Agents
Process Street has launched a hosted Model Context Protocol server that lets AI assistants access live compliance workflows, tasks and operational datasets while inheriting existing user permissions. The server works with ChatGPT, Claude, Microsoft Copilot Studio and other MCP-compatible applications. For smaller SaaS founders, this signals an important product shift: business software increasingly needs to become usable by agents, not only humans.
AI Agent Firewalls Create a New Security Product for Smaller Software Providers
San Francisco startup Operant AI has launched a Semantic Firewall designed to evaluate what AI agents are attempting to do and block, permit or redact actions in real time. For smaller cybersecurity firms and SaaS builders, the opening is emerging one layer above the model: companies deploying autonomous agents increasingly need controls over tools, data and execution, not simply protection around prompts.
Smallest.ai raises 13 million dollars: voice AI startup counts RingCentral and Truecaller as customers
Voice AI startup Smallest.ai has closed a 13 million dollar Series A led by Seligman Ventures, bringing total funding above 21 million dollars. The San Francisco company, founded in late 2024, sells real-time voice models to enterprises including RingCentral and Truecaller, and is betting that an asynchronous architecture that listens, thinks and speaks simultaneously beats larger rivals on latency.
iFLYTEK launches GuideX: public service agent completes responses in 0.42 seconds
China based iFLYTEK has launched GuideX at the 2026 World Artificial Intelligence Conference in Shanghai. The company presents the system as an interaction agent that can move beyond answering questions to completing public service tasks. Its reported response chain of 0.42 seconds and deployment across several regions make the launch relevant to founders building customer service, workflow and vertical AI products.
OpenAI proposes AI ROI scorecard: successful tasks replace token costs
OpenAI has proposed a four-part framework for measuring enterprise AI value around useful work, total cost per successful task, dependability and economic improvement at scale. The approach challenges procurement decisions based mainly on token prices or seat adoption and gives SaaS leaders a practical way to compare models, workflows and human review costs against completed business outcomes.
SAP completes Prior Labs acquisition: more than 1 billion euros backs structured data AI
SAP has completed its acquisition of German artificial intelligence company Prior Labs and committed more than 1 billion euros over four years to expand the business as an independent frontier AI laboratory. The deal gives SAP a specialist team focused on tabular foundation models, a category designed for structured business data such as payments, suppliers, sales opportunities and customer retention.
The 95 percent problem: why most enterprise AI pilots still produce no profit
Behind this month's biggest AI announcements sits one recurring statistic: roughly 95 percent of enterprise AI pilots deliver no measurable profit impact. The gap between capability and value has become the industry's defining commercial problem.
Microsoft commits $2.5 billion and 6,000 engineers to enterprise AI’s deployment problem
Microsoft is putting $2.5 billion and a 6,000-engineer unit behind on-site enterprise AI implementation, an acknowledgement of the industry's uncomfortable statistic: the vast majority of enterprise AI pilots never produce measurable profit impact.
AI agents startup 8090 Solutions raises $135 million to replace enterprise software seats
8090 Solutions, the AI agents platform co-founded by investor Chamath Palihapitiya, has raised $135 million on the thesis that agents will replace seats, not just assist them. The wager divides the software industry.