Snowflake has launched a new managed service called the MCP Server, designed to enhance the use of enterprise data for artificial intelligence applications, particularly in the financial services sector. This service allows users to access their own data alongside third-party sources, including market analysis and expert research, facilitating data-driven decision-making and streamlining processes such as customer service and investment analytics. The MCP Server enables integration with various AI applications and tools, significantly improving how financial services can leverage data. The service is currently in preview and reflects Snowflake’s commitment to providing accessible, secure, and efficient data solutions.
IBM has launched Granite 4.0, a new generation of open source large language models designed to deliver high performance while significantly reducing memory and cost requirements. This latest release combines two architectural approaches—transformers and Mamba—allowing it to achieve over a 70 percent reduction in GPU memory consumption compared to traditional models, making it particularly suitable for enterprises handling long documents and multiple requests. The Granite 4.0 models are open sourced under a permissive Apache 2.0 license and have received certification under ISO 42001 for AI governance and transparency. This positions IBM as a competitive player in the growing landscape of large language models, especially against strong offerings from companies like Alibaba and OpenAI. Early access testing by enterprise partners such as Ernst & Young and Lockheed Martin has indicated that Granite 4.0 models are tailored to meet real-world business needs, emphasizing efficiency and trust in AI deployment.
Otter.ai has introduced new agentic artificial intelligence features that enhance integration with third-party applications and streamline project management workflows. The company now offers a public application programming interface, allowing data to flow both ways between its platform and systems like Salesforce and HubSpot, enabling automated task management based on discussions held during meetings. Co-founder and CEO Sam Liang claims that customers could save the equivalent of one full-time employee’s workload for every 20 users, translating to a tenfold return on investment. Otter.ai‘s transcription services have already processed over one billion meetings.
Google has unveiled a new AI model called Gemini 2.5 Computer Use, which is designed to navigate and interact with websites in a way similar to human users. This innovative model utilizes visual understanding and reasoning capabilities, allowing it to execute tasks such as filling out forms directly within user interfaces that were not originally designed for automation. The Gemini model stands out by performing exceptionally well on various web and mobile benchmarks, although it currently only operates within a browser environment and lacks full desktop control. It supports thirteen specific actions, including typing and dragging elements. Demo videos show the model in action, completing tasks like playing a game or browsing trending news. Developers can access this model through Google AI Studio and Vertex AI, with a public demo available on Browserbase.
Why do we care?
Snowflake’s got a new managed service, IBM’s pushing a leaner open-source model, Otter’s promising to replace a worker per 20 users, and Google’s latest AI can literally use your browser.
Here’s what that really means — AI’s stopped being a lab experiment. It’s now crawling into real workflows. Snowflake’s turning data management into an AI pipeline, IBM’s saying “you can run this yourself,” Otter’s making your meetings do the work for you, and Google’s bots are actually clicking around the web.
I’m skeptical of Otter’s ROI math — you’re not firing staff yet — but directionally? It’s clear. The MSP job is shifting from maintaining systems to connecting them. You’ll win if you can bridge AI tools, manage compliance, and make them deliver outcomes clients can see.
AI isn’t a product — it’s infrastructure. And whoever manages that intelligence layer? That’s your next managed service.