A.CRE
, ,

Episode 11 of Multipliers: Most People and Companies Are Ill-Prepared for the AI Wave

The re-skilling wave is not coming. It is already here. And most people and most organizations in commercial real estate are ill-prepared for the AI Wave.

That is not a criticism. It is an observation from two people who are in this every day, who track the model releases, build the tools, teach the workflows, and still find themselves overwhelmed by the pace of change. If it feels like too much to keep up with, that feeling is accurate. The question is what you do with it.

Spencer Burton and Sam Carlson spend this episode working through that question from three angles: what autonomous agents actually are and why the use cases in CRE are still thin, how AI skills have evolved from a single markdown file into something closer to a full software program, and what the AI re-skilling and retooling requirement actually means for the average real estate professional who just wants to do deals. They also get personal, in a way the podcast rarely does, about the fears that sit underneath all of it and the psychological framework that determines whether those fears become limiters or fuel.

In this episode of the Multipliers podcast, Spencer and Sam cover the current state of autonomous agents, the complexity of modern AI skill building, and the honest conversation about what it takes to stay ahead when the tools are moving faster than anyone can track.

Episode 11 of Multipliers: Most People and Companies Are Ill-Prepared for the AI Wave

Michael Belasco is out visiting family this week, which gives Spencer and Sam an unusual amount of runway to go deep on topics that have been building across several episodes. The conversation opens with Sam’s observation that AI is now present in every ten minutes of his working day, and that he has started treating the different models in his workflow almost like teammates with distinct personalities. That thread pulls Spencer into the autonomous agent question, which then opens into the re-skilling and retooling conversation he has been thinking through at the gym between sets. Spencer is running CRE Agents and teaching at UNC Kenan-Flagler, and he is preparing a training session for Friday on skill building, which gives the technical sections of this episode a live, in-progress quality. The conversation connects back to Episode 10, where Spencer demoed the first AI skill built for a real estate financial model. This episode zooms out from that demo and asks the harder question: what does it actually take for an entire industry to absorb a change of this magnitude?

Listen on: Apple Podcasts | Spotify

Why This Episode, Why Now

The timing of this episode is May 2026, the morning after Google I/O, which released Gemini 3.5 Flash, Project Astra, and a desktop agent called Spark that competes directly with Claude Cowork and Perplexity Computer. The release landed on top of GPT-5 a month prior, on top of Opus 4.7 before that, and the cumulative weight of that pace is something Spencer names directly: it is overwhelming even for people who track this for a living.

Sam’s opening observation sets the tone. He has reached a point where AI is present in every segment of his working day, and the different models have become something like teammates, each with distinct strengths he routes to deliberately. ChatGPT holds the longest context window of his personal history and can answer questions about his audience without him having to re-explain the context. Claude is his prompt engineer and writing partner. Manus is his builder. Gemini is the one he is watching most closely right now. That kind of multi-model workflow was not how most people used AI a year ago. It is increasingly the norm for anyone operating at the frontier.

The deeper context is a conversation Spencer had been having internally about the re-skilling and retooling requirement. Re-skilling means learning to work differently: incorporating AI into workflows, understanding what the tools can and cannot do, building the judgment to evaluate outputs. Retooling means replacing or rebuilding the systems, software, and processes that organizations have built up over years, much of which was never designed to work with AI. Both are necessary. Both are hard. And for the average commercial real estate professional who came up in a world where the skills were stable and the tools changed slowly, both are happening simultaneously at a pace that has no precedent.

Spencer is also prepping a Friday training session on skill building, which surfaces something important: even the curriculum for teaching AI skills has to be rebuilt constantly. What he taught in a skill drop in December is already behind what is possible now. The world where a single skill.md file was sufficient is in the rear view mirror. That rate of change is not slowing down.

Episode Highlights

Here are the themes that stood out.

1. Autonomous Agents: What They Are and Why the Use Cases Are Still Thin

Spencer’s definition of an autonomous agent is worth anchoring to, because the term gets used loosely. An autonomous agent is a system you give an objective to and then set free. It wakes up on a schedule or in response to an event, works toward its objective within the context window it has available, and then either sleeps or continues depending on what it encounters. The creator of OpenClaw shared publicly that his team had spent $1.3 million in tokens running a hundred of these agents continuously. That number is the use case filter: autonomous agents are expensive, and most of the cost comes from the fact that they run through an API at usage-based pricing rather than a flat subscription.

Spencer has experimented with them, including a skill drop for AI.Edge members on spinning up an OpenClaw instance on an AWS EC2 server. His honest assessment: he has not yet found a use case in CRE where the value justifies the cost. The exception that caught his attention came from a Gary Vee interview on My First Million, recorded the morning of this episode. Vee uses an autonomous agent to manage what he calls a relationship graph: a system that ingests transcripts from podcasts and calls, photos from meetings, and any other relational data, and maintains a living log of every connection he has and how. His vision is that in two years, his agent will automatically send a congratulations message to someone when they hit a milestone, because the system understands the context of that relationship.

Spencer’s reaction is the right one: for a networker at Vee’s scale, that is a legitimate use case. For most CRE professionals, it is not yet there. The co-pilot model, using AI to assist and amplify work you are actively doing rather than delegating objectives to an autonomous system, still delivers more value at lower cost for most people in the industry. That will change. But right now, the autonomous agent conversation is more about what is coming than what is ready.

2. The Re-skilling and Retooling Wave

Spencer has been thinking about re-skilling and retooling from two angles simultaneously: as someone who delivers training to major investment managers and as someone who is constantly rebuilding his own workflows to keep pace with what the tools can do. The distinction between the two terms matters and is worth spelling out.

Re-skilling is about people. It means learning to work differently, incorporating AI into daily workflows, developing the judgment to evaluate AI outputs, and building the technical proficiency to direct the tools effectively. This is happening at every level of the industry, from the analyst who needs to understand how to use an AI-powered underwriting skill to the managing director who needs to understand what their team is actually producing when they say they used AI.

Retooling is about systems and organizations. It means looking at the software, processes, and infrastructure that a firm has built up over years and asking what needs to be rebuilt to work with AI. In commercial real estate, that includes everything from the legacy software that manages deal flow to the Excel models that underpin every underwriting. Argus is the example Spencer raises explicitly: a dominant tool in certain modeling scenarios, designed for a pre-AI world, and now facing a genuine question about how long it remains the default.

The complexity is that both are happening at the same time, faster than most organizations can absorb, in an industry that has historically been slow to adopt new technology. Sam’s framing is useful: the move is not to wait for the dust to settle. The move is to start tinkering now, get something working, and let the learning compound. The people who are already in the workflow will adapt faster to each new wave because they have a frame of reference. The people who are waiting for stability will find the gap has widened considerably by the time they start.

3. How AI Skills Have Evolved

Spencer’s screen share of the A.CRE Accelerator direct cap skill is the most technically detailed section of the episode, and it surfaces something important: the skill he is showing is described as fairly simple, and it already looks like a software program.

The skill lives in a private GitHub repository and is delivered to Accelerator members through the A.CRE Intelligence Hub, available to AI.Edge members. It is a folder with a primary skill.md file, reference files covering modeling conventions, inputs, outputs, and recalculation instructions, code snippets the AI uses to handle specific technical challenges in Excel, a YAML schema file that makes the model structure readable by the AI, and different mode files for tutor mode, reviewer mode, and watch-me-build mode. The Excel model itself is embedded inside the skill folder.

Five months ago, the state of the art for a skill was a single markdown file describing what you wanted the AI to do. That world is gone. What Spencer is now preparing to teach on Friday is meaningfully more complex, and the gap between what exists today and what a person with no engineering background can reasonably build on their own has widened.

The watch-me-build mode is worth highlighting specifically. When an Accelerator member activates it, the AI builds the model the same way Spencer would, step by step, so the member can watch the methodology before doing it themselves. The sequence is deliberate: watch Spencer build it, do it yourself by hand, then fire up the AI and let it help. The long division principle, in practice.

4. The Skill Builder as the Next Critical CRE Role

Spencer made an observation that connects the technical detail of the skill walkthrough to a broader organizational reality: every task in the industry needs a skill built for it. That is not hyperbole. It is the logical endpoint of a world where AI is the operating layer through which most analytical and operational work gets done.

The challenge is that building a skill that actually works requires the person who built the underlying process to encode the methodology. You cannot outsource this to someone who does not understand the work. The person who created the template Excel model is the natural person to create the skill for it, because they are the one who knows why every formula does what it does and what the output should look like when it is correct.

Most firms do not yet have someone in this role. The options are: develop someone internally who has both the domain knowledge and the technical proficiency to build skills, hire for it, or rely on generalist skills built by platforms like CRE Agents for the tasks that are common enough to warrant it. The bespoke tasks, the ones that are specific to how a particular firm underwrites or manages assets, require someone inside the organization who can build and maintain them.

Sam’s observation is the optimistic counterpoint: the people who are capable of this exist inside most organizations already. Arturo on the A.CRE team is an example. The barrier is not talent. It is knowing where to go to learn it, which is exactly what AI.Edge exists to provide.

5. What Scares You Most: The 3am Fear and the GFC Shadow

The second half of the episode takes a turn that is unusual for the Multipliers format, and more valuable for it. Spencer asked Sam the question Gary Vee was asked on My First Million: what scares you most? The conversation that followed was honest in a way that most professional development content is not.

Spencer goes there first. About once a month, he wakes up at 3am in a negative loop: none of this is going to work. He lies there, identifies the thought as irrational, tells himself to go back to sleep, and is up by 5:36am regardless. He connects it to PTSD from the great financial crisis, and specifically to the experience of being in the CRE industry when deal flow effectively stopped for years. He and Sam were both 38 at the time, building careers in a market that had no opportunity. The people who came up after them, Sam’s younger brother included, never had that experience. For Spencer and Sam, the felt memory of it sits underneath everything they build, as a residual fear that a decade of good decisions can be erased by forces you did not cause and cannot control.

The reason Spencer raised the question is worth noting. He is not sharing this as a confession. He is sharing it because he believes fear is the greatest limiter of professional growth and that naming it honestly is the prerequisite for managing it. The support and resistance band analogy Sam uses later is exactly right: the limits on what you can earn, build, or become are largely psychological, not external. The fear is real, but it is also a resistance line that can be crossed.

6. The Psychological Operating System and the Signal vs. Noise Problem

Sam’s response to the fear question is one of the more substantive things he has said on the podcast. He describes a realization he had recently about support and resistance lines in trading: those lines are based on psychology, not fundamentals. And if that is true in markets, it is also true in personal earning and growth. The ceiling on what you believe you can make is a mental construct. The version of you that made $100,000 thought $100,000 was a lot of money. The version of you that makes $500,000 looks back and realizes $100,000 would not cover the bills. The operating system upgraded. The external variables changed. But the thing that actually drove the change was internal.

Sam’s conclusion: the version of your operating system you are running today, with everything you have learned and experienced, will recalibrate quickly to new circumstances even if the environment shifts dramatically. You do not need to fear the next disruption the way you feared the last one, because you are not the same person who faced the last one.

The signal vs. noise problem is the practical application of this. Sam describes something that he and Spencer both know they should be doing for the business, something boring, high-effort, and with a long payoff horizon. It keeps getting crowded out by the more immediate, more exciting, more urgent noise. The AI exercise they ran with their custom MCP server pointed directly to that thing as the highest-leverage move available. Elon Musk, as Sam frames it, is extraordinarily successful in large part because he filters to signal almost exclusively. Most people do not.

7. Optimism as the Operating State

Sam closes the episode with something that earns its place precisely because it follows everything that preceded it. He is more optimistic about AI right now than he has been at any point since the beginning. Not because the complexity has reduced. It has not. Not because the fear has gone away. It has not. But because he has been using it daily long enough to see what it is actually going to do, and the hyperbolic worst-case predictions do not match the reality he experiences.

His formulation is the right one for anyone sitting on the sidelines: AI is the musical instrument of our time. You are not going to learn to play it in a day. But after three, four, five months of daily use, it becomes natural. The MCP server that sounds like jargon today is something you will spin up in thirty minutes once you have done it once. The fear of the unknown is real, but it is also temporary, and the only way through it is to start.

Spencer’s closing parallel is the one to carry: every person had to learn to type when the personal computer arrived. Every person had to learn to use a mouse. Nobody opted out of the personal computer era and remained competitive. This is the same transition, compressed into a shorter timeline, with higher stakes for the people who wait.

The Bigger Idea

The title of this episode is a provocation, not a criticism. Most people and most companies in commercial real estate are ill-prepared for the AI wave. That is true. It is also not a permanent condition. The point of saying it is not to generate anxiety but to be honest about the starting line, because the people who acknowledge where they are starting from can actually plan the path forward. The ones who tell themselves they are fine because they opened ChatGPT twice are not planning anything.

Re-skilling and retooling are two distinct requirements that are easy to conflate. Re-skilling is personal: it requires you to change how you work, what you know, and how you evaluate the outputs your tools produce. Retooling is organizational: it requires you to look at everything your firm does and ask what needs to be rebuilt. Both take time. Both require investment. And both are mandatory, not optional, for anyone who wants to compete in the version of CRE that is taking shape right now.

The skill builder observation deserves to be taken seriously as a career signal. Spencer is direct: every task in the industry needs a skill built for it. The person who can build those skills, the one who understands both the domain and the tooling deeply enough to encode the methodology correctly, is in a position of structural advantage right now. That advantage has a window. The window is probably two to three years. After that, the capability will be more widely distributed and the competitive edge will compress. The time to build is now.

Sam’s psychological operating system framework is the piece that ties the personal and professional threads together. The fear of disruption is real, and for anyone who lived through the GFC in a CRE career, it is particularly acute. But the operating system you are running today is not the one you were running in 2008. It has been updated by everything you have learned and built since. The next disruption will require recalibration, not starting over. And the people who have been using the tools, building the skills, and staying in the workflow will recalibrate faster than the ones who were waiting.

The A.CRE Accelerator builds the technical foundation. AI.Edge keeps you current as the tools keep evolving. And platforms like CRE Agents encode the methodology into tools that produce outputs you can trust and act on. The wave is here. The question is whether you are building toward it or waiting for it to pass.

Frequently Asked Questions about Episode 11 of Multipliers: Most People and Companies Are Ill-Prepared for the AI Wave

An autonomous agent is a system you give an objective to and then set free. It wakes up on a schedule or in response to an event, works toward its objective using the context available to it, and operates without requiring a human to prompt each step. Standard AI tools like Claude or ChatGPT require you to be present and directing the conversation. Autonomous agents run in the background. The tradeoff is cost: they consume tokens through API usage at a rate that adds up quickly, which is why the creator of OpenClaw reported spending $1.3 million in tokens running a hundred agents continuously.

Spencer has tested autonomous agents and finds that for most CRE applications, the token cost outweighs the value produced. The co-pilot model, using AI to assist and accelerate work you are actively doing, still delivers more value at lower cost for most professionals in the industry. The Gary Vee relationship graph example is the closest thing to a compelling CRE use case he has encountered: an autonomous system that manages relationship context at scale for someone who has thousands of professional connections. For most practitioners, that level of network complexity does not yet justify the cost.

Re-skilling is about people: learning to work differently, incorporating AI into daily workflows, and developing the judgment to evaluate AI outputs. Retooling is about systems and organizations: replacing or rebuilding the software, processes, and infrastructure that were designed for a pre-AI world. Both are necessary and both are happening simultaneously in commercial real estate. Re-skilling is what an individual does. Retooling is what a firm does. Spencer noted that legacy software like Argus faces a genuine question about its long-term role as AI-native alternatives emerge.

Five months before this episode, a skill was essentially a single markdown file describing what you wanted the AI to do. That world is gone. Spencer walked through the A.CRE Accelerator direct cap skill on screen, and it includes a primary skill.md file, multiple reference files covering modeling conventions and inputs, code snippets the AI uses to handle technical Excel challenges, a YAML schema file, mode files for tutor, reviewer, and watch-me-build functions, and the Excel model itself embedded inside the folder. Spencer described it as a fairly simple skill. It looks like a software program.

Watch-me-build is a mode within the A.CRE Accelerator skill that lets a member ask the AI to build the model the same way Spencer would, step by step, so the member can observe the methodology before doing it themselves. The intended sequence is: watch it built, do it by hand yourself, then use the AI to assist. This is the long division principle applied to skill-based learning: you understand the process before you use the calculator. Members who skip the by-hand step end up with outputs they cannot evaluate.

Spencer argued that every task in commercial real estate needs a skill built for it. The person best positioned to build that skill is the one who created the underlying process, because they understand the methodology deeply enough to encode it correctly. Most firms do not yet have someone dedicated to this. The options are developing someone internally, hiring for it, or relying on generalist skills from platforms like CRE Agents for tasks common enough to warrant it. Bespoke tasks specific to how a firm underwrites or manages assets require someone inside the organization who can build and maintain them.

Sam drew a parallel between support and resistance lines in trading, which are based on psychology rather than fundamentals, and the mental ceilings people place on their own earning and growth potential. His argument is that the version of your operating system you are running today, updated by everything you have learned and experienced, will recalibrate quickly to new circumstances even if the environment shifts dramatically. The person who made $100,000 thought that was a lot of money. The person making $500,000 now looks back and knows it would not cover the bills. The external numbers changed, but the internal operating system drove the shift.

Sam described a high-value initiative he and the A.CRE team have known about for years: boring, high-effort, long payoff horizon. It keeps getting crowded out by more immediate and more exciting tasks. When they ran an AI analysis on their business asking what would double revenue in twelve months, it pointed directly to that initiative as the top priority. The signal was always there. The noise kept winning. Sam used Elon Musk as an example of someone who filters almost exclusively to signal, and connected the ability to act on signal rather than noise to breaking through the resistance lines that cap most people’s growth.

Spencer used the personal computer as a historical parallel for a transition that felt overwhelming at the time but turned out to be non-optional. Every professional had to learn to type, use a mouse, and get comfortable staring at a screen rather than working with pen and paper. Nobody who opted out of that transition remained competitive. AI is the same kind of transition, compressed into a shorter timeline with higher stakes for the people who wait. The complexity is real, but so is the precedent: the people who started early adapted first, and the gap between them and late adopters compounded over time.

Sam’s closing is the answer: AI is the musical instrument of our time. You are not going to learn it in a day, but after three to six months of daily use, it becomes natural. The MCP server that sounds like jargon today is something you will spin up in thirty minutes once you have done it once. The move is not to wait for the tools to stabilize. It is to start somewhere, get it into your workflow, and let the learning compound. The A.CRE Accelerator builds the foundation. AI.Edge keeps you current. And platforms like CRE Agents give you the tools to put it to work right now.

by