AI is not the problem, and it is not the whole answer either. The answer is knowing which lane each technology belongs in.
The question every CRE leader is asking
Scroll through LinkedIn this month and you’ll find the same question again and again: what happens when the next generation of CRE professionals uses AI for everything, but never actually learns CRE?
Most of us learned the hard way. We reviewed leases, built models that broke, fixed them, and had a senior appraiser or analyst explain exactly why we were wrong. That apprenticeship produced judgment. The worry is that if AI does the work, the apprenticeship disappears, and so does the judgment.
It’s a fair worry. But I think the industry is framing it wrong. The real question isn’t AI or no AI. It’s which work belongs to AI, and which work never should.
Two different kinds of technology
AI and deterministic software are fundamentally different tools, and most of the trouble comes from treating them as interchangeable.
AI is probabilistic. It is extraordinary at reading, summarizing, extracting, drafting, and finding patterns in messy information. But ask it the same question twice and you may get two different answers. It cannot show you a governed, auditable chain of reasoning, and it does not have decades of valuation experience. Ask it, and it will tell you so.
Deterministic software is rule-based. Same inputs, same output, every time. Every calculation can be traced, printed, reviewed, and defended. It does not improvise, which is exactly what you want from a valuation model.
A value conclusion has to be defensible, reproducible, and supportable. Only one of these technologies can guarantee all three. When AI is allowed to produce the numbers, you get the black box everyone is worried about: a model nobody can QA, assumptions nobody chose, and an answer that might change tomorrow.
Lane 1: The deterministic core
The deterministic lane holds everything that becomes part of the value conclusion. That means the methodology itself, encoded by credentialed experts with decades of real assignments behind them:
-
The valuation approaches. Cost, sales comparison, direct capitalization, and yield capitalization/DCF.
-
Models for every property type and stage of development. Stabilized, non-stabilized, partially completed, ground-up, entitled and unentitled land, subdivisions, and mixed use. Each model segment has its own logic, and none of it gets stuffed into one oversized spreadsheet.
-
Interconnection. Change an input and every model and every waypoint value along the development timeline and holding period updates consistently.
-
Bidirectional checks. A reconciled value is reverse-engineered so the present value of future benefits equals it, revealing the implied going-in cap rate, going-out cap rate, and yield. If the rate hierarchy breaks, something upstream is wrong.
-
Lease math. Reimbursements, expense stops and caps, gross-ups, capital expenditures, renewal probabilities, and fee simple vs. leased fee cash flows to quantify rent loss or intangible value.
-
Feasibility and holding periods. If a development isn’t feasible today, the system determines the as-is value and re-values every future waypoint when it becomes feasible.
-
Value sustainability. Testing whether a well-supported value, or an investor’s expected IRR, holds up against fundamental value over a typical holding period.
This is where experience lives. Once it is encoded, it’s available on every assignment, to every user, every time.
Lane 2: AI at the edges
The AI lane holds the work around the valuation: the time-consuming, language-heavy tasks that slow analysts down but don’t determine the answer.
-
Due diligence. Reading leases, rent rolls, operating statements, and title documents, and organizing what the core needs.
-
Research and retrieval. Searching a firm’s own database and market data in plain language.
-
Comparable data sheets. Formatting verified comps into the firm’s house style.
-
Report writing. Turning the engine’s results into a narrative report in minutes rather than days.
-
Review. Flagging inconsistencies, missing support, and compliance gaps before a report goes out.
-
Scenario requests. Translating a plain-English question like “what if the cap rate moves 50 basis points?” into inputs. The core still does the math.
AI makes the professional faster. It never decides what the property is worth.
The rules of the road
Keeping each technology in its lane comes down to a few simple rules:
-
AI proposes. The professional verifies. The engine calculates. No number generated by AI goes into a value conclusion without passing through the deterministic core.
-
Every value has a model behind it. If you can’t print it and walk through it, it isn’t defensible.
-
Same inputs, same answer. A saved scenario reopened next month must produce identical results. That’s the test of a governed system.
-
Don’t spend tokens on arithmetic. Running the valuation through AI every time is slow, expensive, and inconsistent. Deterministic math costs pennies and runs in minutes.
-
The signature stays human. Technology supports the conclusion. The credentialed professional owns it.
What this means for the next generation
Here’s where the two-lane model answers the apprenticeship question.
When experience is encoded in the deterministic core, a trainee doesn’t start with a blank spreadsheet and a prayer. They start inside a glass box built by people who have done this for decades. Every waypoint value has a model behind it they can open. Every check is visible. Every scenario can be changed and compared, so they can see what moves value and why.
Their job shifts from building formulas to the work that actually requires a professional: gathering good due diligence, questioning the inputs, interpreting the results, and defending the conclusion. AI takes the drudgery off their desk, so they have more time to learn the parts that matter.
The mentor doesn’t disappear. The mentor is now available on every assignment, at any hour, in a consistent way. Senior professionals spend their time on judgment calls instead of checking VLOOKUPs.
That’s not a generation that never learns CRE. That’s a generation that learns it faster, with better support than most of us ever had.
What firms can do now
-
Draw the line on paper. List every task in your workflow and decide which lane it belongs in.
-
Keep AI out of the math. Use it for documents, research, and writing, not for producing value conclusions.
-
Demand reproducibility. If a tool can’t give you the same answer twice, it can’t be your valuation engine.
-
Make the model visible to juniors. Tools that hide the logic create black boxes. Tools that show it create professionals.
-
Measure it. Put your least experienced analyst on a governed system and your most experienced one on the old workflow, then compare speed and defensibility.
The bottom line
AI is going to change CRE valuation, and it should. But the profession’s credibility depends on keeping each technology where it performs best. Deterministic at the core. AI at the edges. That’s how we bring AI into the valuation process without giving up defensibility, and how we pass decades of experience to the next generation instead of losing it.
At ValueXpose, that principle is the foundation of everything we’ve built. If you’d like to see the two lanes working together on a live assignment, send me a message.
The challenge: your newest appraiser vs. your most experienced
If you doubt that a practical blend of AI and deterministic technology can deliver on all this, good. Don’t take my word for it. Pilot it on your own work.
Schedule a time to preview ValueXpose, and then accept this challenge:
-
Give us your least experienced appraiser. They’ll pilot ValueXpose on a live appraisal assignment, with our team supporting them.
-
At the same time, have your most experienced appraiser complete the same assignment using your company’s typical workflow. That’s the report that goes to your client, so your client relationship is never at risk.
-
Both appraisers start with the same raw due diligence collected on the property. Same property, same information, same purpose of the appraisal.
When both are finished, we put the two reports side by side and compare:
-
Speed: how long each appraiser took to go from raw due diligence to a finished report.
-
Presentation: how clearly and professionally each report communicates its conclusions to the client.
-
Depth and accuracy of the analysis: how thoroughly and accurately each report addresses the scope of work.
You’ll see on a real property, with your own people, what your newest appraiser can do when experience is built into the system and AI handles the busywork.
Are you willing to put your workflow to the test? Here is a link to schedule a time:
https://calendly.com/ray-nicolas/valuexpose-demo-team
Deterministic at the core. AI at the edges. Judgment endures