“Our staff is trained on AI” is one of the least reliable sentences in commercial real estate right now, because the word “trained” hides three very different things and the buyer almost never knows which one they bought. It can mean the team has logins. It can mean they sat through a session and can describe what a large language model does. Or it can mean they reach for the tool on a live lease and get back something they will actually send. Those are three different levels of capability, and only the third one moves a deal faster. This piece decodes what fluency actually is, so you can look at your own team, or a vendor’s promise, and tell which level is really on the table.
One bit of orientation before the definitions. The full arc of getting a small team from logins to genuine capability is laid out in our 90-day training playbook, and the wider case for why lean firms can out-operate larger ones sits in the small CRE firm manifesto. This article is the narrow piece underneath both: a precise definition of the thing training is supposed to produce, and how to recognize it when you see it.
Access, Literacy, Fluency: Three Rungs People Call the Same Thing
The confusion starts because three separate states all get described with the same words. Sorting them is most of the battle.
Access is the first rung: the firm has paid for tools and everyone has a login. This is the easy part, and the industry is mostly here. NAR’s 2025 Technology Survey of Realtors found 68% now use AI tools. Access is real, and it is also the least meaningful of the three, because a login is not a skill.
Literacy is the second rung: the team understands what these tools are, can hold a basic conversation with one, and knows that you type a request and get text back. Most “AI training” delivers exactly this and stops. People leave the room able to explain a large language model and paste a question into a chat box. It feels like a result. It is genuinely better than nothing. But literacy is knowing about the tool, not being able to do your job faster with it.
Fluency is the third rung, and it is a different kind of thing entirely. A fluent employee does not think about the tool; they think about the task, and the tool disappears into the work the way a spreadsheet formula does for someone who lives in Excel. They can take a real lease, a real LOI, a real market question, and get back output they trust enough to act on, then fix it when it is wrong.
The gap between rung one and rung three is exactly where the industry’s disappointment lives. JLL, surveying more than 1,500 CRE decision-makers, found 88% of firms piloting AI but only about 5% achieving their program goals. NAR’s own numbers show the same shape from the other side: 68% using tools, but only 17% reporting a significant positive impact on their business. Those firms did not fail to get access. They got stuck one rung below fluency and assumed the climb was over.
A Working Definition of Fluency
Here is a definition you can actually test against, rather than a mood. A fluent employee can take a real task from their own desk, describe it well enough to an AI assistant to get a usable first result, judge whether that result is good enough to use, and iterate until it is, without supervision and inside the time the task normally takes.
Read that sentence slowly, because every clause is load-bearing.
A real task from their own desk rules out practice on generic examples. Fluency is domain-bound; being fluent at summarizing a made-up document is not the same as being fluent at abstracting your actual leases.
Describe it well enough to get a usable first result is the skill people mistake for the whole thing. It matters, but it is one clause of four.
Judge whether the result is good enough is the clause almost nobody trains and the one that separates fluency from everything below it. A literate user accepts what comes back or rejects the whole tool. A fluent user reads the output the way an experienced broker reads a rent roll: knowing where the errors hide.
Iterate until it is good enough is the recovery skill. Fluency is not getting a perfect answer on the first try; it is getting to a send-ready answer reliably, which usually takes two or three passes.
Without supervision, inside the normal time is the economic test. If a task takes longer with AI than without, or only works when the trainer is in the room, the firm has literacy, not fluency, and the workflow has not actually changed.
What a Fluent Broker Actually Does at the Desk
The definition gets concrete fast when you map it onto the four jobs that fill a small CRE firm’s day.
On a letter of intent, a literate broker asks the tool to “write an LOI” and gets a generic template they then rewrite by hand, saving little. A fluent broker feeds it the deal terms, the property specifics, and a prior LOI in the firm’s own voice, gets a draft that is 80% there, and spends the saved time on the two clauses that actually needed a human. Same tool, different rung.
On a lease, a literate user pastes a page and asks “what does this say.” A fluent user asks a 60-page lease for the specific things that matter, commencement and expiration, renewal options, the operating-expense structure, assignment language, then spot-checks the answers against the source because they know a summary can quietly miss a clause. We go deeper on that discipline in our work on where AI training programs go wrong, because practicing on real documents rather than samples is what builds this exact skill.
On a market write-up, a literate user gets a paragraph of confident-sounding text and ships it. A fluent user gets a draft, notices the tool has no access to current comps, supplies the real numbers, and uses the assistant for structure and language rather than facts, because they know precisely which part of the job the tool is good at.
On a client email, a literate user is happy the draft sounds professional. A fluent user catches that it promised something the deal cannot deliver and fixes it before sending, because judgment, not generation, is the skill they are exercising.
The pattern across all four is the same. Literacy produces output. Fluency produces trustworthy output, because the fluent person is doing the reading and the deciding, and the tool is only doing the typing.
The Four Capabilities Hiding Under One Word
Break fluency into its parts and it stops being mysterious. It is four teachable capabilities, and a training program either builds all four or produces literacy wearing fluency’s name tag.
- Framing. Describing a task in the firm’s own terms, with the context the tool needs, precisely enough to get a usable result. This is the part most courses teach, and the part people wrongly treat as the whole skill.
- Judgment. Reading the output critically and knowing where these tools reliably go wrong, invented specifics, missed clauses, confident-but-stale facts. This is the rarest and most valuable capability, and it comes from practicing on material the person actually understands.
- Iteration. Fixing a mediocre first result in a pass or two rather than discarding the tool. Fluent users expect to iterate; literate users expect magic and quit when they do not get it.
- Restraint. Knowing when not to reach for AI, on a task where it adds no value, or with confidential deal data that should not go into a general tool at all. At a small firm handling tenant financials and unannounced deals, this last one is not optional, and a program that never draws the line has not produced fluency; it has produced a risk.
Notice that only the first of the four is about prompting. The other three are judgment, and judgment only forms when someone practices on their own work, gets feedback, and repeats it enough times that the pattern sticks. That is why fluency takes weeks of real use, not an afternoon.
Why Tool Certificates Measure the Wrong Variable
A certificate of completion tells you someone was present. It does not tell you whether any of the four capabilities transferred, and the two questions have almost nothing to do with each other.
The research on training transfer is blunt about this. Only a small fraction of course content survives the month after a class, roughly a third of trainees are still applying a taught skill a year later, and the single strongest predictor of whether a skill sticks is relevance, whether the person practiced on the material they genuinely handle. A completion certificate measures attendance on day one. Fluency is a question about day thirty, on real work, and the certificate is silent on exactly the days that matter.
There is a second reason certificates mislead, specific to this technology. A course that certifies someone on one product’s current menu ages badly, because the interface changes and the memorized steps break. Durable fluency is tool-agnostic: it transfers cleanly across ChatGPT, Claude, Gemini, and Microsoft Copilot, because all of them reward the same clear framing and the same critical reading of the output. If you are still deciding which of those your firm should standardize on, we compare the two most common choices for a small office in our look at picking a primary AI platform. A team fluent in the underlying skill gets better results as the models improve; a team certified on one product’s buttons is fragile by design.
A Five-Minute Fluency Test You Can Run Yourself
You do not need to be technical to check where your team actually sits. Sit next to someone, hand them a real task from their own desk, a lease to summarize, an LOI to draft, and watch four things.
- Do they know how to start? A fluent person frames the task with context without being coached. A literate person freezes, or types something generic and hopes.
- Do they read the output critically? Watch whether they scan the result for errors or accept it whole. Acceptance is the tell for literacy; scrutiny is the tell for fluency.
- Do they fix it? When the first result is mediocre, do they iterate, or do they give up and do it by hand? Recovery is fluency; surrender is not.
- Do they know the line? Ask what they would not put into the tool. A fluent CRE employee names confidential categories immediately. A blank look is a risk you have not addressed.
Run that on three people and you will know, in fifteen minutes total, whether your firm bought fluency or attendance. It is the same diagnostic a good program is built to pass, and the same one most “trained staff” claims quietly fail.
Measuring Fluency Without an IT Department
The five-minute test tells you about an individual. To know whether the firm as a whole is fluent, you want a couple of numbers you can track without any technical infrastructure, and completion rates are not among them.
Track three behavioral signals instead. First, weekly active use on real work: what share of the team reaches for the tool on an actual deal task in a given week, aiming for near-universal within a couple of months rather than a spike after training that fades. Second, task-time deltas on a fixed set of real jobs: time a lease summary or an LOI draft before and after, on the same kind of document, and see whether the number actually moved. Third, throughput on one revenue-gating workflow, deals screened, LOIs out, listings written, because that is the number the training was ultimately bought to change. None of these needs software you do not already have; a shared note and an honest weekly glance will do. The broader system for turning those signals into an adoption plan is laid out in our adoption framework for small CRE firms.
The reason to measure behavior rather than certificates is that behavior is the only thing that correlates with the result you wanted. Deloitte’s 2026 commercial real estate outlook, surveying more than 850 executives at firms with real budgets, still found 27% blocked by implementation challenges tied to expertise and resistance to change. A 10-person shop faces the same wall with fewer people to lean on, which is exactly why a clear, testable definition of fluency, and a cheap way to measure it, is worth more to a small firm than a stack of certificates ever will be.
Frequently Asked Questions
What does AI fluency actually mean for a real estate team?
AI fluency means a team member can take a real task from their own desk, a lease to abstract, an LOI to draft, a market note to write, describe it well enough to an AI assistant to get a usable first result, judge whether that result is good enough to act on, and iterate until it is, without supervision and inside the time the task normally takes. It is domain-bound and judgment-heavy: being fluent on your firm’s actual leases is not the same as being able to summarize a generic document. Fluency is the point where the tool disappears into the work and the workflow genuinely changes.
What is the difference between AI literacy and AI fluency?
Literacy is knowing about the tool; fluency is being able to do your job faster with it. A literate employee understands what a large language model is and can paste a question into a chat box. A fluent employee reaches for it on a live deal and gets back something they trust enough to send, then fixes it when it is wrong. Most AI training delivers literacy and stops, which is why so many firms have high adoption and low impact: they climbed one rung and assumed the climb was over.
How do I know if my staff is really trained or just attended a session?
Watch them work. Hand someone a real task from their own desk and observe four things: whether they know how to frame the request without coaching, whether they read the output critically instead of accepting it whole, whether they iterate when the first result is mediocre, and whether they can name what they would never put into the tool. Fluency shows up as framing, scrutiny, recovery, and restraint. If someone freezes at the start, accepts whatever comes back, or gives up on a weak first draft, you have attendance, not capability.
Does a completion certificate prove fluency?
No. A certificate tells you someone was present on training day; fluency is a question about day thirty, on real work. Training-transfer research shows only a small fraction of course content survives the following month and that relevance, practicing on the material you genuinely handle, is the strongest predictor of whether a skill sticks. A certificate measures attendance, not the days that matter. Certificates tied to one product’s current interface age especially badly, because the durable skill is tool-agnostic, not menu-specific.
How long does it take a small CRE team to become fluent?
Longer than a single session and shorter than most people fear, typically several weeks of real, repeated use with light reinforcement rather than one event. Fluency is built from four capabilities, framing, judgment, iteration, and restraint, and three of those are judgment that only forms when someone practices on their own work and repeats it enough for the pattern to stick. A half-day workshop can start the process, but the capability consolidates in the weeks after, on live deals, which is why a program with no plan for week two rarely produces fluency.
Can we measure fluency without an IT department?
Yes, and you should, because the useful measures are behavioral, not technical. Track weekly active use of the tool on real deal work and aim for near-universal within a couple of months, time a fixed set of real tasks such as a lease summary before and after training to see whether the number moved, and watch throughput on one revenue-gating workflow like deals screened or LOIs drafted. A shared note and an honest weekly review are enough; you do not need any software you do not already own.
Is fluency tied to one AI tool like ChatGPT?
No. Durable fluency is tool-agnostic. The underlying skill, framing a task clearly, judging the output, and iterating, transfers cleanly across ChatGPT, Claude, Gemini, and Microsoft Copilot, because all of them reward the same clear instruction and the same critical reading. A team fluent in the underlying skill gets better results as the models improve; a team certified on one product’s exact buttons has to relearn every time the interface changes, which makes tool-specific certification a fragile thing to buy.
What should we do before booking AI training?
Decide what fluency should look like at your firm before you buy the program meant to produce it. Name the specific tasks a trained team should do faster, lease summaries, LOI drafts, market write-ups, agree on how you will observe fluency and what you will measure afterward, and make sure the training will practice on your own documents rather than samples. Buying a session before you have defined the target is how firms end up with a certificate and no change in the work.
Where to Start
Every confusion in this piece traces back to one missing step: nobody defined what fluency should look like at the firm before buying the training meant to produce it. That definition is the real starting point, and it is exactly what a free AI-readiness assessment produces, a working session that maps which of your workflows would actually change with a fluent team, sets an observable target for each, and gives you a cheap way to measure whether the training moved it. Book a free AI-readiness assessment and you will leave with that target, plus an honest read on whether your team needs a full workshop, a lighter course, or simply a clear set of rules for the tools it already has.
Arthur Wandzel
