For the last two years, most conversations about AI in commercial real estate have been about demos. Someone types a prompt into a chat window, gets a plausible-looking paragraph back, and everyone nods. Interesting. Not something an asset manager builds a hold-period strategy around.
That phase is ending. AI in CRE is moving out of the demo and into the work that actually moves money: offering memorandum review, rent roll checks, valuation, feasibility, deal screening, and transaction support. This is a quieter shift than the hype cycle, and it matters far more, because it changes where the advantage lives.
The question for owners and asset managers is no longer whether AI is useful. It is whether you are positioned to capture the advantage when your competitors are pointing the same tools at the same deals.
The demo era is closing and the workflow era is opening
Start with a document every asset manager knows well: the offering memorandum. An OM is a sales document. The rent roll and the T-12 show what the property is actually doing. The gaps between the OM and those operating documents are exactly what cost you money if you miss them during diligence.
That is not a new insight. What is new is that AI can now find those gaps at speed. Thesis Driven is running a workshop teaching operators to abstract and audit OMs with two agents, one reading the OM and one reading the lease data, surfacing the exact points where the documents disagree. That is not a chat toy. That is underwriting labor compressed into minutes.
The same pattern is showing up in valuation. Green Street reported that one beta client turned a 100-hour, five-market site-selection exercise into a 15-minute prompt using its data delivered directly into the AI tools the team already uses. Same data. A fraction of the time.
This is the real signal. AI is being pointed at specific, repeatable, high-value operating tasks, not open-ended conversation. And the adoption base is still early enough that positioning now matters. Census Bureau data charted by the Economic Innovation Group shows just under 22 percent of U.S. businesses used AI in any business function as of July 2026, up from about 17 percent the previous November. We are early. That is precisely why the groundwork you lay this year decides who compounds an advantage and who plays catch-up.
Where AI helps and where it hurts
Here is the part most vendor pitches skip. AI does not deliver uniform value across a workflow. It delivers enormous value at the front of a process and gets riskier the closer you get to an irreversible commitment.
Thesis Driven made this point cleanly in the context of design and construction. A tool that reads floor plans and counts materials with 98 percent accuracy is genuinely useful when you are evaluating a site and need a rough cost. But once you are ordering materials, 98 percent accuracy is a twelve-week delay, because someone ordered 98 windows and the job needed 100.
The firms getting real value are the ones that can tell those two situations apart. Early-stage screening, feasibility, and anomaly detection tolerate imperfect accuracy because a human reviews the output before anyone commits capital. Late-stage execution does not.
For an asset manager, the translation is direct. Use AI to widen the top of the funnel, screen more deals, catch more OM-to-rent-roll discrepancies, run more feasibility passes. Keep human judgment on the decisions that reprice the asset or trigger a wire transfer. The value is real, and so is the discipline required to capture it without introducing new risk.
Think about what this means over a full acquisition cycle. An analyst who can screen three times the deal flow without sacrificing rigor changes what your pipeline can hold. The constraint stops being human hours and starts being the quality and reach of the data those hours are pointed at. That is the shift worth planning around.
The model is the commodity. Your data is not.
Now the strategic core. If AI is moving into your underwriting and valuation workflows, ask a simple question: what actually separates your firm's AI output from the firm bidding against you on the same deal?
It is not the model. Model performance is converging fast, and the cost to hit a given benchmark is falling several times over each year. When capability converges, model selection stops being strategy and becomes procurement. You pick the engine the way you pick a payroll processor.
What differentiates one portfolio from another is four things only the owner can build. Proprietary operating data. Repeatable workflows. An orchestration layer that lets your team apply any decision engine under your own rules. And institutional knowledge encoded into systems rather than trapped in a departing analyst's head.
Every one of those four depends on something upstream that most owners have never actually controlled: the data itself. And here is where the workflow shift gets uncomfortable. The AI tools reading your OMs, your rent rolls, your operating history, and your building systems are only as good as the data they can reach. If that data lives inside a vendor's platform, in a format you cannot export, with a history you cannot retain, then the intelligence AI produces from it becomes the vendor's asset, not yours.
That is the whole game. If you don't own your data & digital infrastructure, your vendors do.
What owner-controlled AI readiness actually requires
AI readiness is not a software purchase. It is a data & digital infrastructure question, and it has a specific shape. This is the work OpticWise organizes through the PPP 5C™ plan from Peak Property Performance®, and each step maps directly to what the workflow era demands.
Clarify. Before you point any AI at your portfolio, you need to know what data you actually have, where it lives, who controls it, and whether it is trustworthy and portable. A PPP Audit™ maps ownership and identifies where operating data leaks into vendor platforms you cannot reach. You cannot underwrite with AI on data you cannot get to.
Connect. The operating data that feeds valuation and diligence comes from building systems, meters, access control, and property management systems. If those run on fragmented, vendor-controlled connectivity, your data arrives inconsistent and incomplete. Owner-controlled connectivity, repeatable property to property, gives you a clean feed rather than a patchwork.
Collect. AI is only as reliable as the data model behind it. Capturing and normalizing operating data into one consistent, reusable format is what lets you run the same OM-versus-rent-roll check across every asset instead of rebuilding the workflow at each address.
Coordinate. This is governance, and it is what separates responsible AI from automation without guardrails. Identity, access, privacy, data lineage, retention, and rules of use. When an AI agent reads your lease data, you need to know what it can touch, what it can retain, and under whose permission. Property Brain™ governs that.
Control. Now you can plug in any decision engine, any vendor platform, any large language model, and let it act under your permissions on data you own. Standardize that across the portfolio and Property Brain™ becomes Portfolio Brain™, so the intelligence compounds across buildings instead of restarting at every deal.
That sequence is the difference between renting AI capability and building a durable capability of your own. It is also the difference between an AI initiative that survives a vendor change and one that collapses the moment a contract lapses.
Why this reaches NOI and valuation
An asset manager does not fund a data project for its own sake. So connect the workflow shift to the scoreboard.
Better, faster OM and rent roll review means you catch discrepancies that would otherwise show up as a repricing surprise late in diligence, or worse, after close. Every discrepancy caught early is a dollar of value protected before the price moves against you. Faster, cheaper feasibility and deal screening means your team evaluates more opportunities per analyst, which improves selection and acquisition quality across the hold.
On the operating side, the same owner-controlled data that feeds acquisition AI also feeds expense control, tenant consumption allocation, and the operating narrative that supports a property tax appeal or a refinancing package. Recoverable NOI in multi-tenant office runs roughly 0.60 to 0.90 dollars per rentable square foot per year, and in multifamily roughly 500 to 600 dollars per door per year. Capitalize that recovered NOI at prevailing cap rates and every dollar becomes roughly fifteen to twenty-five dollars of asset value at refinance and exit.
That capitalization math is why this is an asset-management question and not an IT question. A one-time efficiency gain is nice. A durable, owner-controlled data foundation that compounds across the hold and shows up as a cleaner operating story at refinance is a different order of value entirely.
AI accelerates the discovery and the workflow. Owner-controlled data & digital infrastructure is what turns that acceleration into value you keep rather than value your vendor captures.
The window is open now
The firms integrating AI into real underwriting and valuation workflows are not waiting for the technology to mature further. They are moving now, while adoption is still early and the advantage is still available to build. Deep-dive coverage of operators like Metropolis shows how quickly this space is professionalizing around specific operating use cases rather than generic promises.
The owners who compound an advantage from this shift will be the ones who did the unglamorous work first: mapped their data, secured their connectivity, normalized their operating model, and governed their permissions. Then, and only then, they plugged in the AI. When the model converges to a commodity, and it will, they still own the four things that matter.
Start with one property. Run a PPP Audit™ to see what data you actually control and what is quietly leaking into vendor platforms. Establish Property Brain™ as the governed foundation. Prove you can apply a decision engine to your own data under your own rules. Then scale the standard across the portfolio.
The AI moving into your workflows this year is powerful. The question is whether it will be building intelligence you own, or intelligence someone else rents back to you.
Own your data & digital infrastructure. Operate with strategic foresight. Build for the long game.
