Luke Tatman on AI and Commercial Real Estate

Luke Tatman on AI, Commercial Real Estate, and the Value Beyond Information

When information becomes cheaper to find and assemble, professional value does not disappear. It becomes easier to see.

That is the central argument in episode 81 of Venture Step. Dalton Anderson and Luke Tatman use commercial real estate, insurance, law, finance, and consulting to test a common assumption: established firms are protected because they know things that clients cannot easily know.

AI weakens the narrowest version of that advantage. A client or smaller firm can produce a first-pass market summary, read a large document set, draft a model, or create a polished presentation with fewer people and less time.

The conversation then asks the harder question. If access to information is no longer enough, what is the professional actually being paid for?

The answer that emerges is more demanding than "relationships." It includes better data, source verification, judgment, integration, accountability, service, sales, negotiation, and the ability to turn evidence into a decision.

Luke brings an operator's view of commercial real estate

The preserved transcript introduces Luke as a corporate strategist at a large privately held real estate firm. It does not name the company.

Luke describes his work as helping steer company strategy through technology implementation, resource allocation, pattern recognition, and data analysis. Commercial real estate gives him a useful vantage point because firms observe how clients across many industries make decisions about workplaces, facilities, investment, and growth.

Current public evidence belongs to a different time boundary. As of July 27, 2026, Luke's LinkedIn profile lists Avison Young as his employer and describes him as a Senior Associate helping companies align real estate strategy with business goals. California's Department of Real Estate salesperson list names Luke Jamison Tatman under license 02217368 among the salespeople affiliated with Avison Young - Southern California, Ltd.

Those records establish a current public relationship. They do not establish that Avison Young was the unnamed episode-era employer. Luke's exact title and preferred public introduction still require his approval.

The full dated source boundary appears in the [[Luke Tatman on AI Strategy and Commercial Real Estate|Luke Tatman guest background]].

The first threat is to a task, not a job title

The conversation begins with white-collar work that once looked protected.

Luke focuses on jobs built from repeated steps: create the marketing brochure, fill out the model, prepare the research, and move the package to the next person. Those tasks can demand care and experience while still being structured enough for software to compress.

Dalton pushes the argument toward lower barriers to experimentation. A person who once needed a developer, designer, data team, or large budget can now test a prototype with general tools and a modest amount of time.

The episode sometimes speaks in broad terms about jobs and industries. Current research supports a more precise task-level view.

A 2026 peer-reviewed Organization Science experiment studied 758 Boston Consulting Group consultants. On tasks inside the tested model's capability frontier, AI users completed more work, worked faster, and produced higher-quality results. On a complex task outside that frontier, AI users were less likely to reach the correct answer.

That is not a commercial real estate study. It does explain why job-level predictions are unreliable. Two tasks that look equally difficult to a person can sit on opposite sides of a model's current capability.

The operating response is to map the work, measure a bounded task, and keep the professional close to the evidence.

Information access was only one layer of the moat

Luke describes the traditional commercial real estate value proposition as a combination of information and professional barriers to entry.

A firm could invest in databases, researchers, analysts, market relationships, and internal software. It could gather comparable transactions, demographic evidence, rents, property records, and local context that a client could not easily assemble.

AI does not make every source public or correct. It reduces the effort required to search, organize, compare, and explain the material a user can access.

The distinction matters. Information access and information quality are not the same thing.

An early screen may tolerate a range. A large warehouse, office lease, investment sale, financing, or development decision may turn on a small change in rent, occupancy, capital expenditure, timing, or financing. The closer the decision gets to capital and commitment, the less useful a plausible approximation becomes.

flowchart LR
    A["Access and retrieval"] --> B["Assembly and synthesis"]
    B --> C["Proprietary context"]
    C --> D["Judgment and verification"]
    D --> E["Workflow integration"]
    E --> F["Accountability and execution"]
    F --> G["Client outcome and trust"]

AI puts immediate pressure on the first two layers. It can also assist with the later ones. The later layers remain harder to separate from a firm, professional, engagement, and decision.

The Venture Step article [[When Information Stops Being the Moat]] tests each layer against substitution, accountability, and client outcome.

Relationships matter until the service gap becomes too large

Dalton raises the obvious defense. A professional who has served a client for 20 years has trust, history, and access that a new tool does not have.

Luke agrees that people continue to do business with people they like. Commercial real estate is still driven by deal origination, persuasion, timing, and the ability to bring a buyer and seller to a decision.

The disagreement is about how much underperformance the relationship can carry.

A trusted advisor may survive a modest difference in price, speed, or capability. The relationship becomes less protective if another firm provides a materially better answer, broader service, faster turnaround, or lower cost.

That makes "we have relationships" an incomplete strategy. The relationship must improve the work. It should create better context, faster coordination, more honest communication, stronger negotiation, lower execution risk, or greater confidence when the evidence is uncertain.

The same reasoning applies to technology. "We use AI" is not a strategy either. The client cares whether the service became more useful and whether the firm remains accountable when the output is wrong.

The middle-market squeeze is a hypothesis, not a result

Luke proposes a market structure with two strong positions.

Large firms can use data, specialization, distribution, and scale. Small firms can compete through focus and close relationships. An undifferentiated middle can be stuck with more coordination cost than a boutique and less capability than a platform.

That is one of the episode's sharpest ideas. It is also the claim that needs the most discipline.

Current evidence shows an adoption gap. The OECD reported that in 2025, 52 percent of large firms used AI compared with 17.4 percent of small firms across countries with available data.

Eurostat's current enterprise AI report found 40.43 percent adoption in professional, scientific, and technical services and 24.82 percent in real estate activities across covered EU enterprises. Large firms adopted more often. Lack of expertise, uncertainty about legal consequences, and privacy concerns were common barriers among firms that had considered adoption.

Those statistics do not show that AI has caused middle-market firms to lose market share, margins, clients, or employment.

The counterargument is strong. General models and industry software lower the fixed cost of experimentation for smaller firms. A mid-sized firm may have enough volume to build a governed workflow and fewer legacy systems than a global competitor. It may also combine specialists under one roof without losing partner attention.

The full analysis, [[Why AI May Squeeze the Middle Market|Why AI May Squeeze the Middle of Professional Services]], defines the hypothesis, gives the middle a strategic rather than purely statistical meaning, and identifies the evidence that would confirm or reject it.

Commercial real estate shows where production can compress

Offering memorandums provide the episode's most concrete industry example.

An offering memorandum combines property facts, financial information, market context, images, comparables, maps, and an investment narrative. The work contains repeated structures that a system can help assemble.

Henry's current Compass Commercial case study says one broker reduced a basic broker opinion of value from roughly two hours and a full offering memorandum from as much as seven hours to about 20 minutes.

That is a vendor-published customer result. It is not an independent benchmark or an industry average.

The credible takeaway is the use case, not the number. Document extraction, research, financial tables, maps, narrative, and layout can move through a connected system. The broker and client still need to verify the property facts, assumptions, financial figures, comparables, claims, images, and disclosures.

Current professional guidance reinforces the boundary. The RICS standard for responsible AI use requires covered firms to address confidential data, system appropriateness, supplier due diligence, risk registers, professional judgment, output assurance, and client communication for material use.

The Venture Step guide [[How AI Changes Commercial Real Estate Work]] maps those controls from client instruction and document intake through underwriting, marketing, negotiation, diligence, and close.

The value moves from gathering data to telling the right story

Luke argues that the industry can move from labor-intensive gathering toward the harder work of deciding what the evidence means.

He describes a pitch team in which the broker brings the relationship and sales responsibility while analysts help build the factual story. As more firms gain access to similar tools, the differentiator becomes the quality of the data, the logic connecting it, and the decision the client can make from it.

"Story" can sound like decoration. In professional work, it should mean disciplined explanation.

A strong story identifies the decision, establishes the evidence, exposes the assumptions, explains the uncertainty, compares realistic alternatives, and shows why one course of action follows. It does not use fluent language to conceal a weak source.

That is where AI can help and mislead at the same time. It can turn evidence into readable prose. It can also make unsupported reasoning look complete.

The professional advantage belongs to the team that can distinguish the two.

Entrepreneurship does not require leaving the company

Luke's practical advice is aimed at the person whose current task is becoming easier to automate.

The answer is not to protect the manual process because it once required skill. It is to identify where the process fails, learn what the new tools can do, and propose a better way to deliver the work.

Dalton clarifies that this is entrepreneurship inside the company. A person can build a new workflow, service, analysis, or internal product without quitting to launch a startup.

The responsible version is a controlled experiment. It starts with one recurring problem, a baseline, authorized data, a known reviewer, and a decision about what counts as correct. It measures total workflow effort and downstream errors, not only the speed of the first draft.

The guide [[How to Practice Entrepreneurship Inside a Company]] turns that advice into a pilot charter. Scale is treated as a new authorization, not an automatic reward for a promising demo.

Continue the thread

Readers interested in workflow design should continue with E097 and [[Should This Task Be a Workflow, an Agent, or Manual]]. It separates predictable orchestration from open-ended judgment.

E103 and [[From Selling Hours to Selling Outcomes]] examine what happens when AI changes the labor required to produce professional work but the client still buys an outcome.

E120 and [[How to Manage an AI Coworker]] move the discussion from task automation toward ongoing delegated responsibility.

Together, the episodes trace a progression from cheaper production to redesigned workflows, pricing, authority, and management.

Source and release note

This story is based on the preserved E081 transcript. The canonical public episode URL, feed record, audio, video, and recording date have not been recovered into the package. The legacy note lists September 8, 2025 as a publication date, but that date remains provisional until it can be matched to a public record.

Direct quotations are omitted because the transcript contains recognition errors and the audio is missing. Luke's current title, preferred introduction, and portrait require approval.

The article can complete editorial review. Publication remains held until the episode identity and guest approval boundaries are resolved.

AI assisted with research organization and drafting. Dalton Anderson remains responsible for the analysis, source boundaries, and publication decision.