When Agentic AI Meets the Physical Economy

I'm Tony Uphoff, and this is the Uphoff on Media podcast. I spent 35 years running business-to-business information, marketing, and technology companies. Now, agentic AI is rewriting the rules for all 3. I help operators, investors, and executives stay ahead of it. New episodes drop twice a week on Apple and Spotify, drawn from my Substack, Uphoff on Media. Okay, let's get to work. This episode is titled When Agentic AI Meets the Physical Economy: AI's Impact on Commercial Real Estate. I was recently asked to speak to the Southern California chapter of the Councilors of Real Estate about the impact AI is having on their industry. The irony hit me the moment I started preparing. The Councilors of Real Estate is one of the most selective professional organizations in the country. Membership requires a minimum of 10 years in the commercial real estate business, a 31-page application, and 4 interviews. Members are expected to be at or near the top of their field. The roughly 900 members nationally and internationally include major commercial property owners, developers, architects, contractors, appraisers, agents, and consultants. Getting admitted is genuinely difficult. The talk drew sustained engagement, tons of pointed questions and requests for follow-up sessions from a room of about 20 of these professionals. That is not a room that nods along out of politeness. Here's what I told them and what the response confirmed. Before we get started though, a note. If you know someone in commercial real estate, please share this podcast with them. The impact AI is having on the physical economy carries real lessons for those of us who operate in the digital economy too. So even if CRE is not your world, continue to listen and pass it along. Worlds colliding. To open the discussion, I acknowledge the irony. We were going to talk about the impact of an intangible, artificial intelligence, on the ultimate tangible asset. Commercial real estate is land, concrete, steel, and glass. You can walk through it. You can touch it. It generates cash flow. You can underwrite and model. It sits at the most physical end of the business spectrum. And it is being reshaped by a technology that exists entirely as math. That irony is not unique to CRE. It is the defining tension of the next decade of business. Agentic AI is colliding with the physical economy. Industries organized around tangible assets, physical workflows, and on-the-ground operations are encountering a technology that restructures the knowledge work wrapped around those assets faster than most operators in those industries currently realize. This is a theme I'll return to on Uphoff on Media when agentic AI meets the physical economy. The industries I see most impacted include commercial real estate, logistics and warehousing, and manufacturing. Not in any fixed sequence when the date of the moment or a conversation like the one I had with the Councilors of Real Estate makes the case for the next one, I'll write it. The thesis underneath all of it, the conventional wisdom about which industries are protected because they deal in physical assets, is wrong. Not because physical assets do not matter, because the knowledge work organized around these assets is being substituted, not enhanced. And that's the distinction that changes everything. Let's start with CRE. First, the AI distinction that actually matters. There's an enormous amount of coverage of AI in the general media right now, most of it focused on generative AI tools like ChatGPT. That technology is useful. I use it every day. But the impact of generative AI will be dwarfed by what comes next. Generative AI is reactive. You prompt it, it produces something. You ask it to summarize a lease abstract, draft a tenant response, produce a market overview. It waits for you at every step. It does not plan, it does not pursue goals, it does not act without you directing it. Agentic AI is goal-directed. You give it an objective and it figures out how to achieve it. It can browse the web, query databases, analyze documents, send communications, make decisions within defined parameters, and loop back to refine its own work, all without a human directing every step. The practical difference is not incremental. Wired into the right data sources, an agentic AI system can be told, pull every industrial property in a 50-mile radius of Phoenix that has traded in the last 18 months from our CoStar and county recorder feeds, flag the assets with the steepest price moves, and draft a ranked memo on each one with the supporting comps attached. That is achievable today for a firm with the right data access wired in. The next layer, having the agent independently inform buyer identity and capital source where it is not disclosed and reason about institutional capital flows behind a transaction is harder and less reliable right now. It's also coming fast. Either way, the work runs overnight, the output is waiting in the morning. This is not an enhancement to the analyst function. For many firms, it replaces significant portions of it. These systems exist today and are being deployed in financial services, legal, logistics, and healthcare. They are coming to commercial real estate. In some forward-leaning firms, they're already here. The frame that will mislead you. Before getting into the CRE data, there is a conceptual point worth establishing because how you think about this technology will determine how you respond to it. Almost every technology you have adopted in your career has been additive. It added a capability you did not have before without replacing what you were already doing. CRM systems, email, Excel, CoStar, DocuSign, Zoom. In every case, the broker still brokered, the deal still closed because a human being with judgment, relationships, and accountability made it happen. Technology made that person faster, better informed, more organized. It did not replace them. That is the frame most operators bring to every new technology. When something new shows up, the instinct is to ask What does this add? Where does it fit in my existing workflow? That instinct will mislead you with agentic AI. Substitutive technologies do not add to the existing workflow. They replace significant portions of it. The printing press did not improve the scribal system. It made scribes obsolete. The automobile did not augment horse-drawn carriages. It eliminated the carriage industry within a generation. The internet did not enhance physical retail, classified advertising, and travel agencies. It dismantled them. Across every technology disruption I have observed and led businesses through, intelligent and experienced professionals consistently underestimate substitutive change because their mental model is calibrated to a world where the core activity survives every technology wave. They look for the additive angle. They optimize the existing workflow. And by the time the substitutive reality becomes undeniable, The window for strategic advantage has already closed. 2 stories running at the same time. The AI story in commercial real estate is not one story. It is 2 very different stories running simultaneously in different asset classes on different timelines. Story 1: AI as a massive demand driver. Data centers are the hottest asset class in commercial real estate right now, and AI is the primary engine. Jones Lang LaSalle projects that roughly 100 gigawatts of new data center capacity will come online between 2026 and 2030. That equates to $1.2 trillion in real estate asset value creation. CBRE's 2026 Data Center Outlook puts it in sharper relief. Pre-releasing rates on capacity now under construction are running in the mid-70% range, against a historical norm of 40 to 50%. That is not developers betting on demand. It is demand that is already under contract before the building exists. This sector is not demand constrained, it's power constrained. The ability to deliver 300-megawatt-plus power capacity in under 36 months has eclipsed fiber connectivity as the dominant site selection factor. Power is the new location variable. Utility relationships, behind-the-meter solutions, and powered land are driving investment decisions in ways that would have been unrecognizable to this industry 5 years ago. Data center construction has now surpassed office construction in the United States. If you own or manage industrial, flex, or large footprint suburban assets, the conversion about data center conversion is a strategic question, not a speculative one. Story 2, AI as a potential demand destroyer. In Q1 2026, AI companies accounted for 22.7% of all office leasing across major US technology markets, according to CBRE. That is 11.5 million square feet of demand from a tenant category that barely existed 5 years ago. In San Francisco, a market many wrote off, AI tenants have absorbed large blocks of Class A space that sat dark through the post-pandemic years. And yet the same technology driving that demand is automating the entry-level workforce whose companies hired into that space. As AI takes on the analytical coordination and processing work done by associates and junior staff, the companies currently expanding their footprint may need significantly less space per employees going forward. That signal is already visible. Meta eliminated 8,000 employees while simultaneously raising its 2026 data center budget to between $125 billion to $145 billion. Hear that again. That was not a typo. It says $125 billion to $145 billion. That is the shape of the AI economy. Fewer people, more infrastructure. The office opportunity is real but concentrated. The best building— buildings in the strongest markets with genuine amenity differentiation will capture AI company tenants. The rest of the market faces the same structural headwinds it has navigated since 2020. AI may accelerate those headwinds rather than relieve them. What AI is doing inside CRE operations. The productivity gains in underwriting and analytics are already measurable. Work that took analyst teams days is being completed in minutes. Rent roll analysis, lease abstract review, and market comp screening are being automated at speed across the industry's leading firms. Morgan Stanley research found that 37% of tasks currently performed by REITs and commercial real estate firms can be automated. That represents a potential efficiency gain of $34 billion by 2030. A self-storage operator has already cut on-property labor hours by 30% through AI-powered staffing optimization. Morgan Stanley's research goes further. Brokerage and service firms have the highest automation potential of any CRE subsector, with a possible 34% increase in operating cash flow precisely because they're the furthest along in adopting AI tools at scale. The capital markets priced risk directly. In February 2026, shares of CBRE and JLL fell 12% on AI disruption concerns. Cushman Wakefield dropped 14%, the largest single-day drops for those firms since COVID Investors are asking a specific question about how much of what the major brokerages do can be automated. The correct answer is not all of it. The relationship-intensive, judgment-dependent, contextually complex work of CRE is genuinely resistant to AI substitution. Complex transactions require human accountability. Significant deals require human trust. That does not go away. But the analytical and administrative infrastructure supporting those transactions is a different story. And that infrastructure is significant cost and significant Headcount. The agentic frontier adds another layer. Leading firms are already planning for AI systems to handle routine late lease monitoring, portfolio reporting, tenant communication workflows, and initial screening. The broker of the future is a judgment and relationship layer on top of AI infrastructure, not an information gatherer. 8 recommendations. Number 1, get your data house in order first. Agentic AI is only as good as the data it can access. Fragmented, inconsistent, or siloed information will limit effective deployment. Run an audit of where your critical portfolio and operational data actually lives, whether it is accessible, and whether it is structured. This is the unglamorous prerequisite. Most firms are simply not ready. Start with underwriting and analytics. This is the highest ROI entry point for most CRE operations. Pilot AI tools in due diligence and underwriting workflows before touching anything client-facing or relationship-intensive. Build confidence in the output quality and develop institutional knowledge before expanding. Redesign roles, not just workflows. The instinct is to bolt AI onto existing job descriptions. The strategic move is to ask what each role is actually for. Analysts become AI systems directors. Property managers become exception handlers and relationship escalation points. Leasing teams become closers rather than initial contact points. This requires deliberate organizational design, not just a software subscription. Evaluate your portfolio for data center conversion potential. If you own or manage industrial, flex, or large footprint suburban office assets, Run a rigorous feasibility assessment focused on power access and fiber infrastructure. Not every asset qualifies. Missing it where it does is a significant opportunity cost. The positioning window is open now. Number 5, identify and protect your high-value human functions. Relationship-intensive, judgment-dependent, contextually complex work is genuinely resistant to AI substitution. Name it explicitly. Invest in it. Do not let efficiency pressure lead you to automate or understaff the functions that AI cannot replace. These are your competitive moat. Number 6, assign someone to track this full-time. AI in CRE is moving fast enough that passive monitoring is not sufficient. Someone in your organization or a trusted external advisor needs to be continuously tracking developments, evaluating tools, and connecting AI capabilities to your specific business model. This is a core operational function now, not a side project. Number 7, think in the 18 to 36 month window. The early majority of agentic AI adoption in CRE arrives in the next 18 to 36 months. Pilot programs launched this year become institutional knowledge and workflow advantage by 2027. Waiting until the technology is fully mature means starting from zero when the competitive landscape has already shifted. And number 8, do not let perfect be the enemy of deployed. Imperfect early adoption builds organizational learning. Waiting for certainty means starting from scratch at the moment of maximum competitive pressure. And here are 5 trends to keep on your radar. Power infrastructure as the new location variable. Sites capable of delivering 300 megawatt plus in under 36 months are being repriced. Land near utility substations deserves a serious second look. AI-driven lease negotiation and management agents— early deployments are underway. Know what your counterparties are using before it appears across the table for you— from you. Automated property management— AI-managed maintenance scheduling, tenant communications, and facilities optimization are compressing operating costs. Underwriting assumptions built on today's expense ratios need revisiting. Synthetic market intelligence. AI-generated market data is beginning to compete with traditional data providers on speed and accessibility. The quality gap is closing faster than most operators in this space currently realize. And the AI immunity trade in capital markets. Institutional capital is distinguishing between AI-vulnerable and AI-resistant asset classes. Physical, power-intensive, infrastructure-grade assets are being repriced as scarce in the AI economy. This is a structural tailwind for the right CRE positions if you understand which ones qualify. The bigger pattern— CRE is where I'm starting because the irony is sharpest here. The most tangible asset class in the economy being reshaped by the most intangible technology. But the pattern underneath is not unique to real Across the physical economy, the same dynamic is playing out. Knowledge work automation is hitting industries organized around physical operations. The analytical layers, the data processing, the document workflows, the coordination functions— these are being substituted by agentic AI regardless of whether the underlying business deals in steel, logistics, farmland, or square footage. The professionals who navigate this best will not be the ones who waited for the story to be fully written. They will be the ones who made a clear-eyed assessment of what their work actually consists of, identified what is genuinely substitutable and what is not, and repositioned their value toward the judgment, relationships, and contextual intelligence that no AI system can replicate. That assessment is hard to make from inside an industry. It requires the right frame. I'll be returning to this theme. The next industry I anticipate covering will likely be logistics and warehousing, a sector that is simultaneously the infrastructure layer for the AI economy and a primary target of its most aggressive automation. The tension is different from CRE. The urgency is higher. But I'll get there when the moment is right. If there is an industry in the physical economy you would like to see covered, please drop me a comment and let me know. Well, that's it for this episode of Uphoff on Media podcast. If this was useful, please rate the podcast, and if so inclined, write a quick review. It helps other executives and operators find the show, and I read every one of them. Thanks. Back at it next week.