Every asset manager has a handful of jobs that come back on a schedule. The quarterly investor report. The market read that supports a hold, sell, or refinance recommendation. The budget variance explanation. The lease abstract that has to be right before a renewal conversation. The data package a lender asks for when a loan comes due.
Those recurring jobs are where AI in commercial real estate is heading. The early hype cycle was about chat windows and clever one-off answers. What the market is now organizing around is narrower and more useful: building AI into the research, reporting, and decision work that repeats every month and every quarter.
That shift matters to owners for a reason that has little to do with which model is smartest. When AI is embedded in a recurring workflow, the workflow starts to accumulate value: corrected outputs, approved assumptions, source documents, and the reasoning behind decisions. Whoever controls that accumulation controls the advantage.
What the market is signaling
A few recent signals point the same direction, and it is worth being precise about what each one does and does not show.
Thesis Driven is running a live workshop where its CTO builds a GP's market research system in 90 minutes. The framing is direct: a solo operator can't out-hire a national platform's research team, but with the right AI research system, one can out-see it. That is a claim about method, not a reported financial result. What it tells us is that practitioners are treating market research as a system to be built once and run repeatedly, rather than a task re-done from scratch for each deal.
Realcomm's upcoming webinar, AI in CRE: Early Adoption, Real Examples, and Practical Steps, describes early adopters putting AI to work across analysis and reporting, knowledge capture, and workflow automation. Again, this is a program description. It establishes where the industry conversation is going. It does not establish returns.
Outside our industry, Insurance Journal reported on Accenture's view that insurers will get the most from their AI investments by adopting an enterprise-wide strategy rather than focusing on select use cases. That is insurance, not commercial real estate, and I would not treat it as CRE proof. It is useful context because it describes the same pattern: AI value tends to come from how it is organized across the business, not from isolated experiments.
And JLL's recent analysis of physical AI describes AI moving robots from executing predetermined, repetitive tasks to understanding their environment and learning new ones. That piece is about robotics in buildings, which is a different subject. I mention it because the underlying idea carries over to knowledge work: AI becomes valuable when it can learn from accumulated context. In an asset management office, that context is your data and your decision history.
Why the workflow, and not the model, holds the value
Model capability is converging, and the cost of reaching a given level of performance keeps falling. Most firms will end up using several models over the next few years, often switching as prices and capabilities change. When that happens, picking a model starts to look like procurement.
What does not convert so easily is the material a workflow builds up over time. Take a quarterly asset report. The first time you automate it, the AI drafts commentary from rent rolls, operating statements, and market data. Your team corrects it. They tell it that a particular expense spike was a one-time roof repair, that a tenant's slow payment is a known timing issue, that the comparable set for this submarket excludes a certain building. Next quarter, those corrections should carry forward.
That chain of corrections, assumptions, and approvals is institutional knowledge. Today most of it lives in analysts' heads, email threads, and spreadsheet comments. When an analyst leaves, much of it leaves with them.
A well-built AI workflow can capture that knowledge in a form the organization keeps. Where it gets captured is the strategic question. If the source data, the prompts, the corrections, and the approval history all sit inside a single vendor's platform, the knowledge is now an asset of that platform. Switch vendors or models and you start over.
If you don't own your data & digital infrastructure, your vendors do.
How this reaches NOI and asset value
Asset managers are right to be skeptical of claims that AI improves NOI. So it helps to trace the chain carefully.
The first effect is time and accuracy in recurring work. If a variance explanation or market read takes less analyst time and contains fewer errors, the team can cover more assets or go deeper on the ones that matter. That is an operating benefit, and you should measure it directly rather than assume it.
The second effect is decision quality. A hold-or-sell recommendation built on consistent, traceable research, with the reasoning preserved from prior quarters, is easier to defend to an investment committee and easier to revisit when conditions change. Better decisions about expense control, capital timing, and renewals are where operating improvements actually reach NOI.
The third effect shows up at refinancing and disposition. Lenders and buyers ask for operating history and want to understand it. When your operating data and the reasoning behind your numbers are organized, portable, and in your control, the diligence conversation gets shorter and surprises get rarer. When that history is scattered across vendor systems you can't fully export, gaps surface at the worst possible moment, after the price has already started to move.
Each link in that chain depends on the same condition: the data and history have to belong to the owner and remain usable when tools change.
The problem underneath: fragmented data & digital infrastructure
Most owners who try to automate a recurring workflow hit the same wall quickly. The inputs are scattered. Utility data sits in one portal, access control logs in another, work order history in a third, and the building network is often managed by several parties with no shared standard.
We see what that fragmentation costs. In one PPP Review of a 400,000 SF office property, we found roughly $300K of redundant fiber, parallel backbones nobody could explain, each under a different vendor's control. That is a physical example, but the data version of the same pattern is common: overlapping systems, duplicate feeds, and no single party responsible for making the information trustworthy and portable.
AI does not fix that condition. It amplifies it. An AI workflow running on inconsistent inputs produces confident-looking outputs that are hard to audit. Automation without governance is how firms end up trusting numbers they can't trace back to a source.
What a credible approach requires
If the value sits in the workflow, a credible approach has to do four things. It has to preserve source data in a form the owner can access and export. It has to keep a record of decisions and corrections over time. It has to define where humans approve outputs before they reach an investor, lender, or tenant. And it has to remain usable when the model or vendor changes.
That is the logic behind the OpticWise two-layer model and the Peak Property Performance® PPP 5C™ plan.
- Clarify (PPP Review): pick the recurring workflow, define what success means in accuracy and time, and map where every input lives and who controls it.
- Connect (managed data & digital infrastructure): establish secure, owner-controlled connectivity to those sources, repeatable property to property.
- Collect (managed data & digital infrastructure): capture and normalize the inputs into a consistent model you can reuse each cycle.
- Coordinate (Property Brain™ + Portfolio Brain™): govern identity, access, lineage, retention, and the approval rules that decide when a human signs off.
- Control (Property Brain™ + Portfolio Brain™): let any decision engine, whether a vendor platform, internal analytics, or any LLM, run the workflow under owner permissions.
The first three steps build the foundation your team owns. The last two make that foundation usable by whatever AI you choose, today or three years from now. Property Brain™ is vendor- and LLM-agnostic by design, so the corrections and history your team builds stay with the property. Standardize it once and it becomes Portfolio Brain™, where those workflows run across buildings instead of restarting at every address.
Where to start
Resist the urge to launch a portfolio-wide AI program. Choose one recurring workflow that your team already does every quarter and that matters to a capital decision. The asset report, the refinance data package, or the submarket read are good candidates.
Then do four things before you expand. Preserve the source data and the decision history in a place you control. Write down who approves what before it leaves the building. Measure accuracy and analyst time against the way you did it last quarter. And test portability by confirming you could move that workflow to a different model without losing what it has learned.
If the numbers hold up, standardize it and move to the next property. If they don't, you have learned something cheaply, and you still own every input.
The AI market will keep changing. Your recurring workflows, and the knowledge they accumulate, are the part worth owning.
Own your data & digital infrastructure. Operate with strategic foresight. Build for the long game.
