IBM and Microsoft Launched Imagine Enterprise for Industry Agents

IBM and Microsoft launched Imagine Enterprise for industry agents

IBM and Microsoft launched a platform on Tuesday for the part of AI work that slows companies down. Microsoft said Imagine Enterprise helps companies build, run, and govern agents that carry out industry-specific processes. IBM Consulting brings the industry side, with more than 6,000 prebuilt agents and assets. Microsoft brings the foundation: Foundry, Copilot Studio, and GitHub Copilot to build and coordinate agents, with Fabric, Microsoft 365, Entra, Purview, Defender, and Azure underneath for data, identity, governance, security, and cloud.

The first industry frameworks cover financial services, insurance, and retail, the regulated end of the market, where the checks matter most. More industries are planned after those. The companies did not share pricing or a general availability date, so this is a launch to evaluate, not a product a team can budget for yet.

Most teams can build an agent by now. Fewer can answer the rollout questions. Who owns it, what data may it touch, and how does anyone check its work? This platform starts from those answers instead of adding them later, which is the honest order. The headline count needs care, though. The 6,000 mixes finished agents with other assets, so the useful question is how many fit your process as they are.

Judge it the way you would judge any rollout tool. Take one process and count the hours it takes today. Then ask whether the governed version finishes it faster, with a record of the agent's work that a compliance lead would accept. If it cannot show that on one process, a bigger library will not save it. Governance is not the brake on agents. Done right, it is what lets a leader say yes.

Anthropic put Claude inside the meeting

Anthropic put Claude inside the meeting on Tuesday. TechCrunch reported Claude Meetings adds Claude to a video call as a visible participant. While people talk, Claude drafts the plan, maps a timeline on a whiteboard, and answers questions in the chat. It can also pull facts from connected work apps, like last year's figures.

The controls are part of the launch. Claude names itself in the chat and posts a notice that it is there. It will not join a call with outside guests unless an admin allows it, and it leaves if an outside guest joins. Meetings is in beta on Team and Enterprise plans, and it starts with Google Meet. After the call, the notes, action items, and materials sit in claude.ai.

Two side tools came with it. Claude Whiteboards lets a team build a board together and turn it into a doc or a deck. Claude Sheets builds working spreadsheets, like a campaign tracker or a hiring budget, and shows each change it makes. Both are in beta on paid plans. Claude Data is now generally available, and it answers questions from a data warehouse or customer system with dashboards and written stories. Put together, this is Anthropic moving onto the ground Microsoft holds with Teams and Office: the meeting, the spreadsheet, and the data behind both.

The problem here is follow-through, and every team knows it. Meetings end, and the plan sits in scattered notes until someone rebuilds it. Work that starts during the call removes that rebuild step. The measure is simple. Count the action items created before people hang up, then check how many are done a week later. Set the rule early for which calls Claude may join, because a guest in the room changes what people say.

Google released Gemini 4 Carbon to everyone, with a swarm preview

Google made its strongest model public on Tuesday. Google said Gemini 4 Carbon is now open to developers, enterprises, scientists, and consumers. Two weeks ago, only a set of cybersecurity defenders could use it, through the Fairwind Program. Carbon is built for long, hard jobs, like multi-step coding and open-ended scientific research.

The bigger change is how the work gets organized. Google is previewing /swarm in Antigravity, its tool for building with agents. One prompt becomes a team of agents that you can track as they work. Carbon can take the deep roles, like design, coordination, and auditing. Lighter models do parallel work in separate workspaces, and verifier agents test every change.

Google says Carbon ranks first on three outside tests of real work, at more than twice lower cost per task than the next best model. Those tests are CursorBench, Perplexity's WANDR, and Zapier's AutomationBench. The rankings are Google's report of those tests, so treat them as the company's numbers. Carbon starts at $2 for a million input tokens and $10 for a million output tokens. Simon Willison noted the price is set to double on December 1.

A team should judge this model on cost per finished task, not cost per token. A swarm runs several agents at once, and each one spends while it works. Give the strong model the coordination and checking roles, and give the volume to lighter models. Then measure tasks finished and time saved, and set a spending limit before the swarm starts. That is the difference between a busy demo and a service a budget owner can approve.

Also today, New York opened registration under its RAISE Act, a safety law for large AI developers. Companies with more than $500 million in yearly revenue must register by the start of next year, publish safety frameworks, and report serious incidents within 72 hours. OpenAI's Business Hub moved down the list on a busy day: brands can now claim a verified profile in ChatGPT and launch an agent that answers questions about what they sell, in beta for eligible United States businesses. Meta put its Muse agent on Apple Watch and Wear OS, where people can use voice, texts, or dictation. And Google added two chips, the TPU 8tx, built for training models on sensitive data, and the TPU 8i Max, with double the memory of the chip it builds on.

A governed platform, an AI in the meeting, and agent teams on demand. The race is no longer about the smartest model. It is about who runs agents a business can trust, at a cost it can explain.

About this brief: This brief was researched and drafted by my personal AI assistant and approved by me before publishing. Product details change fast; check primary sources before you build on them.

Want more with the Tech Simplified crew? Subscribe on Substack.

Get the week's briefs in one email. The Briefcase packs Monday to Friday's AI Briefs into one Sunday email.

Follow me on LinkedIn for project updates, ideas for using AI, and lessons you can bring to your own work.