For a 4–20 person commercial real estate firm, the first AI dollar should almost always buy fluency, not software. Get the team competent with general AI tools on your own LOIs, leases, and market write-ups first; buy or build software second, once you can tell which vendor features are real and which of your workflows actually deserve a project. Software-first is the right call only in a few honest exceptions — a burning, well-understood pain, or a commodity workflow with an obvious off-the-shelf answer. Order it wrong and you pay twice: once for licenses nobody adopts, and again for a build scoped against hope instead of evidence.
Most AI-spend advice for small firms is written by someone selling one side of this decision: software vendors tell you to buy the platform, training companies tell you to book the workshop. This piece takes a position instead, built for the 4–20 person firm with no IT department that has to make the call once and defend it in a partner meeting. It sits alongside our broader buy-vs-build playbook for small CRE firms and the wider small-firm AI manifesto.
It Is a Sequencing Question, Not an Either/Or
The framing “workshop or software” is wrong on its face. A serious firm does both within the first year. The real question is order, because order changes what each dollar buys.
Spend on fluency first and the software you buy afterward is chosen by people who can now judge it. Spend on software first and you are asking a team that has never worked with AI to evaluate, adopt, and get value from a tool they cannot yet assess. One sequence compounds; the other gambles.
This matters more at small scale, not less. A 200-person firm can absorb a failed software rollout as a line item. A 10-person firm that sinks its annual AI budget into the wrong platform has spent the whole budget, and the appetite to try again is gone for a year. The order is the decision.
The Default: Fluency First
For nearly every firm this size, fluency is the correct first spend. Three reasons carry the argument.
Fluency Makes You a Better Software Buyer
Proptech AI features change every quarter, and the marketing runs ahead of the reality. A team that has spent a few weeks using ChatGPT, Claude, Microsoft Copilot, or Gemini on real deal documents develops a working sense of what these models do well (summarizing a lease, drafting a first-pass LOI, reformatting a rent roll) and where they still need a human. That sense is the single most useful thing you can bring to a vendor demo.
Fluent buyers ask the questions that separate real capability from a slide. They notice when a “proprietary AI” feature is a thin wrapper over the same general model they already use for free. They can tell whether a $500-per-seat platform earns its price over a $20 subscription plus a good prompt library. Non-fluent buyers cannot, and vendors price accordingly.
Fluency Is Cheap, Fast, and Reversible
A focused fluency effort — a workshop plus a few weeks of applied practice — runs roughly $2,000 to $15,000 at market rates, lands in weeks, and leaves you with a capability the firm keeps regardless of which software you later choose. If it underdelivers, you are out a modest sum and a bit of time.
A custom automation build, by contrast, runs roughly $25,000 to $150,000, takes months, and is far harder to unwind. The asymmetry is the point. When one bet is small, fast, and reversible and the other is large, slow, and sticky, you run the small one first and let what you learn size the large one.
Software Bought Before Fluency Becomes Shelfware
Unused software licenses are one of the most reliably documented waste lines in business technology; a large share of purchased SaaS seats go unused across industries, and AI tools bought on hype are especially prone to it. The pattern in CRE is familiar: a firm buys an AI-enhanced platform, two people try it, nobody builds it into daily work, and the subscription renews on autopilot.
The cause is almost never the software. It is that the team had no working model of what the tool was for, because they had never used AI on their own work. Fluency is what turns a purchased license into an adopted one. Buy the capability to adopt before you buy the thing to be adopted.
When Software Should Come First
The default is not a law. There are honest cases where spending on software first is the right sequence, and pretending otherwise would be dishonest. Three qualify.
A Burning, Well-Understood Pain
When a specific, expensive problem is on fire right now — a compliance deadline, a portfolio you cannot report on, an AP backlog bleeding late fees — and an off-the-shelf tool solves exactly that problem, buy the tool now. Fluency training will not stop the bleeding this month. Solve the emergency, then circle back to fluency for the rest of the firm.
The test is honesty about the word burning. A workflow that is merely annoying is not burning. If the pain has been tolerated for two years, it can wait the few weeks fluency takes.
A Commodity Workflow With an Obvious Answer
Some workflows are so standard that the buy decision is trivial and needs no fluency to get right. Listing marketing, accounting, e-signature — the market has mature, cheap, well-understood tools, and no amount of AI training will change which one you pick. Buy the obvious tool and move on, and save the fluency-first discipline for the decisions where judgment actually matters.
A Workflow You Already Know Cold
If the firm already understands a workflow in complete detail — every rule, every exception, every edge case is documented — then you have, in effect, already done the hard part that fluency and discovery exist to produce. A build scoped against that clarity is far less risky. This is rare at small firms, because most workflows live in one person’s head rather than on paper, but where it is genuinely true, software-first can be defensible.
The Cost of Getting the Order Wrong
Both wrong orders have a price, and naming them makes the sequencing rule concrete.
| Wrong order | The tax you pay |
|---|---|
| Software first, no fluency | Licenses nobody adopts (shelfware), a platform chosen by buyers who could not evaluate it, and per-seat fees renewing on autopilot for value you never captured. |
| Build first, no proven usage | A custom automation scoped against a guess. The team could not tell the builder what the workflow really is, so the build solves the imagined problem, not the real one, and the rework starts on day one. |
| Fluency first, then nothing | The rare failure mode of over-caution: the team gets competent, then never commits to the tooling that would scale the gains. Fluency is the start of the sequence, not the whole of it. |
The first two taxes are the expensive ones, and both come from spending on tools before the firm could judge them. The third is a reminder that fluency-first does not mean fluency-only.
For the full three-year cost math behind the second row — what a build actually costs versus a subscription — the buy-vs-build playbook works the numbers at 5, 10, and 20 seats. And for how the fluency spend fits a whole-year budget, see how much a 10-person CRE firm should budget for AI.
A Sequencing Rule for One Partner Meeting
You can settle the order in a single meeting by walking these questions in order and stopping at the first “no.”
- Is there a burning, well-understood pain with an obvious off-the-shelf fix? If yes, buy that tool now, then resume at question 2 for everything else. If no, continue.
- Is the team fluent with general AI tools on our own documents? If no, that is the first spend — a workshop on prompting for LOIs, lease summaries, market write-ups, and email. Nothing else gets bought until this is true.
- Now that we are fluent, which two or three workflows actually cost us the most hours? Inventory them. This list did not exist before fluency, because you could not see which problems AI could touch.
- For each, is the honest answer buy, build, or wait? Standard workflows: buy. Firm-specific workflows that are how you make money and cross systems no vendor connects: consider a build. Everything else: wait, and keep using the general tools.
- Have we proven the volume before committing build money? No usage data means you are guessing. Buy cheap or wait; do not build on a guess.
If the room cannot agree on question 2, the firm is not ready for questions 3 through 5. That disagreement is itself the signal that fluency comes first. The rules for buying software without an IT department to lean on are worth reading before you sign anything — our ten rules for buying AI without an IT department cover the traps.
What “Workshop First” Actually Means
Fluency-first is a real, bounded thing, and it is easy to oversell it into something it is not. What we mean by a fluency workshop is narrow on purpose: getting your team competent using general AI tools — ChatGPT, Claude, Microsoft Copilot, Gemini — on the actual documents your firm produces. Prompting for a first-draft LOI. Summarizing a 90-page lease into the ten terms that matter. Turning a stack of comps into a market write-up. Cleaning a broker email before it goes out.
That is the whole of it. A fluency workshop is not a tool-stack audit, not a custom assistant build, and not an AI-policy exercise. Those are separate pieces of work with their own value; conflating them inflates the “training” line and muddies the sequence. The point of fluency-first is small, fast, and specific: a team that can now see where AI helps and judge every tool and build decision that follows. For a sense of where the market sits and which off-the-shelf capabilities are already mature, our read on the state of AI proptech for small CRE firms is a useful companion.
Frequently Asked Questions
Should we spend on AI training or AI software first?
Training first, in almost every case, for a 4–20 person firm. Fluency with general AI tools on your own documents is cheap, fast, and reversible, and it makes every later software decision better because the team can finally judge what a tool is worth. Software bought before the team is fluent tends to go unused. The main exception is a burning, well-understood problem with an obvious off-the-shelf fix — buy that now, and get fluent on everything else.
How much does AI fluency training cost for a small CRE firm?
Market rates for a focused workshop plus applied practice run roughly $2,000 to $15,000, depending on team size and depth. It lands in weeks, not months, and leaves you with a capability the firm keeps no matter which software you buy later. Compared with a custom automation build at $25,000 to $150,000, the training spend is small, fast, and low-risk, which is exactly why it belongs first in the sequence.
Isn’t buying software faster than training the team?
Buying is faster to sign, not faster to value. A license the team cannot use produces nothing until adoption happens, and adoption is what fluency creates. A firm that trains first often reaches real productivity sooner than one that buys first and then spends months trying to get anyone to use the platform. Speed to a purchase order is not speed to results.
What counts as a fluency workshop, exactly?
Getting your team competent with general AI tools — ChatGPT, Claude, Microsoft Copilot, Gemini — on your firm’s real work: LOIs, lease summaries, market write-ups, and email. That is the scope. It deliberately excludes tool-stack audits, custom assistant builds, and policy work, which are separate engagements. The goal is a team that can see where AI helps and judge the tool and build decisions that follow.
When does it make sense to buy software before training?
Three cases. A burning, well-understood pain with an obvious off-the-shelf fix — buy it now, the emergency will not wait for training. A commodity workflow like accounting or listing marketing where the tool choice is trivial and training would not change it. Or a workflow the firm already understands in complete, documented detail, which is rare at small firms but does happen. Outside these, fluency first.
What is shelfware and why does it matter here?
Shelfware is software you pay for but do not use. A large share of purchased SaaS seats go unused across industries, and AI tools bought on hype are especially exposed. For a small firm, shelfware burns both the annual AI budget and the appetite to try again. Fluency-first is the most reliable defense, because it turns the team into people who can actually adopt what they buy.
Can we do training and software at the same time?
You can, and for a burning pain you should buy that one tool immediately. For everything else, a short head start on fluency pays off, because even a few weeks of the team using AI changes which software you choose. Think of it as a sequence measured in weeks, not a year-long gate.
Does fluency training reduce what we spend on software later?
Often, yes. It stops you buying platforms whose “AI” is a thin wrapper on the same general models your team can already use cheaply, and it keeps you from scoping a custom build against a guess. Fluent teams buy less software and scope better builds, so the training spend frequently pays for itself in avoided purchases.
How do we know when we are fluent enough to start buying?
When the team can name, without prompting, the two or three workflows where AI would save the most hours, and can explain what a tool would have to do to help. That clarity is the output of fluency and the input to a good buy-vs-build decision. If the room cannot produce that list, the firm is not fluent yet, and buying software would be premature.
Where to Start
The sequence — fluency, then buy or build against proven usage — is easy to state and easy to get wrong under vendor pressure. The honest starting point is a clear-eyed look at your own workflows: which ones cost the most hours, which the team could handle today with general tools, and which genuinely warrant a purchase or a build.
That inventory is what we produce in a free AI-readiness assessment: a working session that maps your firm’s workflows, sets the order of spend, and returns a ranked plan with honest fluency, buy, build, or wait calls — including the ones where the answer is a $20 subscription and a good prompt, not a project. Book a free AI-readiness assessment if you want the sequence set for your firm before the next vendor demo, not after.
Arthur Wandzel
