AI News This Week
Anthropic Launches Claude Managed Agents for Enterprise - Anthropic shipped managed agents that handle sandboxing, permissions, and state management out of the box - cutting enterprise deployment timelines from months to weeks. Notion, Asana, and Rakuten are already live.
Google Drops Gemini 3.1 Ultra with a 2M Token Context Window - Google's new flagship model can process 2 million tokens across text, image, audio, and video in a single request.
Microsoft Launches Its Own AI Models - Microsoft released three in-house foundational models (the MAI series) on Azure this week, directly competing with OpenAI and Google. This isn't a side project - it's a strategic declaration that Microsoft is no longer content being OpenAI's distribution layer. It wants its own model stack, on its own terms.
79% of Organizations Are Struggling with AI - New data from Writer's 2026 Enterprise AI report paints a stark picture: 54% of executives say AI adoption is straining their organizations, and only 16% of companies have actually redesigned jobs or workflows around AI capabilities. Most firms are buying tools, plugging them into existing processes, and wondering why the ROI isn't there.
Magma Proptech Opens First U.S. Office in Miami - European proptech disruptor Magma — which uses digital twins, AI, and blockchain for ownership records and permitting — just planted its flag in the U.S.
Topher dropped a new skill in the Prompt Library this week that retail leasing brokers are going to love.
The Retailer Map Builder is a Claude Cowork skill that generates interactive retailer maps on demand. Give it a property address, answer a few questions about what you're looking for, and it builds a visual map showing every nearby retailer by category (e.g. grocery, dining, banking, services, fitness) with color-coded markers and a legend. You can customize the radius, the marker sizes, and the categories. The whole thing takes about 2 minutes.
If you've ever spent an hour manually pulling competitive retailer data for a site selection package or landlord presentation, this is the kind of workflow automation that saves real time. No subscriptions. No GIS software. Just an address and a question.
The best part? It's a reusable skill so you can run it once, run it a hundred times, and it produces the same clean output every time. That's the power of building AI into your workflow instead of using it ad hoc.
Not sure when a skill beats a one-off automation? Do You Need a Skill or an Automation? breaks it down.
You can see how it works in the video below. Find the fully built Claude skill in the CRE AI Studio and try it on your next retail deal.
Site selection is one half of the picture. For sourcing the deals themselves, see Turn Claude Cowork Into Your Prospecting Pipeline.
Some of the biggest names in CRE are handing their teams a free AI tool and calling it a strategy. We know that free plans for any of the major tools (ChatGPT, Claude, Co-Pilot, etc.) will result in little output and even less adoption.
We’ve been a bit shocked to see how little prepared some of these firms are and how even less willing they are to invest in tools that can materially change how their workers perform.
It is, however, a sign of the times that we’re still early in this new world we’re living in. The bigger firms have so much more at stake than the smaller, more nimble ones. A lot of these bigger companies are public and simply giving everyone a $20 subscription to Claude isn’t a smart strategy.
The enterprise plans, however, should be considered strongly. With their extremely rigid data protection protocols, companies should feel secure.
And the next logical step should be to create their own AI policy that mandates what can and cannot be done using these tools.
That’s a lot more red tape than what the smaller players are facing but this strategy would be so much better than what we’re seeing now.
For example, the free tier of Copilot is a taste, not a toolkit. It caps out on the features that actually matter for CRE workflows: deep document analysis, advanced reasoning over financial models, long-context processing of lease abstracts and T12s, agentic automation that's starting to reshape how deals get underwritten and assets get managed.
When your analysts can summarize a paragraph but can't feed a 25-page OM into AI because of limits, they’ve probably just wasted more time than if they just did it the old fashioned way.
This pattern isn't new. Before AI, firms would say they'd "adopted technology" for lease drafting because they had Word templates and a shared drive. Meanwhile, the teams that actually invested in purpose-built tools were cutting lease review time by 80%. The gap between "we have something" and "we use something that works" was enormous. AI is following the exact same trajectory.
The real cost here isn't the $0/month license. It's the opportunity cost. Every month a team spends tinkering with a free tool is a month they're not building muscle memory on the workflows that actually move the needle: automated variance analysis, AI-assisted lease abstraction, intelligent rent roll screening, deal pipeline triage. The firms pulling ahead aren't the ones with the biggest IT budgets. They're the ones that invested in the right tier of the right tools and paired them with actual process redesign.
And doesn’t everyone want to get ahead?
This is a leadership problem, not a technology problem. When a CRO or Head of Asset Management tells their team "use the free Copilot," the implicit message is clear: AI isn't important enough to invest in. The team hears that loud and clear. They tinker for a week, hit the limits, and go back to doing things the old way. Meanwhile, the 50-person shop down the street with Claude Pro and a handful of custom workflows is running circles around them.
We’re seeing this firsthand.
Here's the contrarian take: giving your team free AI tools isn't progressive. It's the new way of saying "we're not ready." If you wouldn't hand your analysts a free version of Excel and expect institutional-quality models, why would you do it with AI?
We're 18 months into the enterprise AI wave. The early adopters have separated from the "we're exploring it" crowd. The firms that invested are seeing meaningful reductions in operating costs. The free-tier crowd is still writing marketing emails.
The question isn't whether your firm uses AI. It's whether your firm takes it seriously enough to actually pay for it.
And not taking AI seriously is…a serious problem.