“If you think AI is just here to help, you haven’t been paying attention to how fast it’s replacing the routine.”
Part 2: The Disruption
In Part 1, I laid out how AI is transforming productivity, scaling expertise, and expanding what’s possible, especially for small and nimble commercial real estate firms. But that’s only half the story.
Now we turn to the uncomfortable part: displacement. Because while AI might be the greatest force multiplier of our time, it’s also one of the most destabilizing. Not because it’s evil, but because it’s ruthlessly efficient. And in a business like CRE, where execution still depends on experience, networks, and nuance, we’d be foolish to ignore what’s coming.
Welcome to the “Jagged Frontier”
There’s a concept propagated by Stanford’s AI research: the “jagged frontier”. It means AI isn’t advancing in a smooth, predictable curve. It’s lumpy. Erratic. It can outperform humans in some areas (like data summarization) while completely failing in others (like understanding social nuance). This uneven progress makes it hard to see what’s safe and what’s not, until the wave hits.
We’re already seeing it in the job market:
• Tech layoffs linked to AI adoption totaled over 42,000 in 2024.
• Entry-level tech jobs shrank 25% last year and are down over 50% since 2019.
• Mid-tier roles like schedulers, analysts, and operations staff are quietly being replaced by AI agents trained to handle “glue work.”
CRE analogy: That assistant who used to compile your debt stack, or the intern who used to prep your lease audit? Their next job might be a prompt.
CRE Is Not Immune
If you think commercial real estate is too relationship-driven or judgment-heavy to be affected, think again.
AI is already capable of:
• Qualifying leads and writing prospecting emails.
• Scraping and summarizing offering memoranda.
• Populating base-level underwriting models.
• Analyzing foot traffic, income data, and rent comps.
• Generating full investment summaries based on a synthesis of various data sources.
Even middle management isn’t safe. Microsoft’s internal reorgs have shown that AI is starting to take on the responsibilities of project managers, schedulers, and even coders. In CRE, that could mean asset managers who rely on templated reporting or brokers who still rely on static marketing decks.
The Rise of the Agent-as-Producer
Here’s where it gets existential. We’re entering what some are calling the “Human-led, Agent-operated” era. In this model, AI agents don’t just assist. They produce. Humans simply assign goals, and agents execute, with humans stepping in only when judgment, escalation, or emotional nuance is required.
Yuval Harari takes this even further. He argues that if AGI (Artificial General Intelligence) reaches human-level cognition, we may be “hackable animals”, outmaneuvered by algorithms that understand our behavior better than we do.
That’s not a dystopian future. That’s a warning: the line between augmentation and replacement is thin, and it’s moving fast.
What AI Still Can’t Do (Yet)
Despite all this, AI has critical blind spots. And in CRE, those gaps still matter.
AI struggles with:
• Weighing tradeoffs with limited context.
• Prioritizing stakeholder needs.
• Understanding nuance in interpersonal dynamics.
• Making decisions with incomplete or unstructured data.
• Interpreting zoning edge cases, city council politics, and environmental review nuance.
It’s also prone to hallucinations. That means your agent may summarize a rent roll, but mis-categorize escalations. It might draft an LOI but skip a critical clause on tenant improvements. AI makes easy things easier. But it can make the hard things worse, unless paired with human oversight, and domain expertise.
Beware the Slopware
One of the great ironies of AI is that it makes it easier to build, and easier to build garbage. Some are referring to this as the “slopware problem.” Inexperienced users can now generate decks, models, and memos in minutes. But they often lack the judgment to know when the output is flawed. This is especially dangerous in CRE, where a single bad assumption can destroy a deal.
Think of underwriting:
• The IRR looks good, but are the lease rollover risks modeled cogently?
• The cap rate looks right, but are the market comps for properties with important differences?
• The DSCR checks out, but were the T12 repairs normalized?
Without critical review, i.e. domain expertise, AI becomes a risk amplifier, not a solution.
The Real Bottleneck Isn’t AI, It’s Us
Ironically, the biggest barrier to safe, productive AI use isn’t the technology. It’s human:
• Most teams are overwhelmed by the pace of change.
• Few know how to integrate AI into daily workflows.
• Pricing models (per user, per month) punish experimentation.
• Security fears block adoption.
• And cultural resistance, especially in CRE, slows everything down.
We don’t have a tech problem. We have a human enablement problem.
Up Next: The Co-Evolution
In Part 3, we’ll stop thinking in binaries – replace vs. enhance – and move toward a smarter paradigm:
What does it mean to lead an AI-enabled team?
How do we build workflows where agents and humans collaborate?
And how do small shops in CRE use this moment to leap ahead of larger firms?
The answer isn’t fear. It’s design.