The real cost of a small commercial real estate firm waiting a year on AI is not the software it did not buy — because buying software, by itself, is exactly what failed for most firms that tried. JLL’s 2025 technology survey found that roughly 88% of real estate owners and investors had started AI pilots, yet only about 5% reported hitting most of their goals; Deloitte’s 2026 outlook found the share of CRE executives claiming a transformative impact from AI had collapsed to around 1%, from roughly 12% a year earlier. So the honest cost of waiting is not “your competitors bought a tool and you didn’t.” It is subtler and, for a 4–20 person shop, more expensive: a year of un-built team fluency and un-captured proprietary advantage, both of which compound, set against the fact that the correct first step is cheap and reversible. This piece prices that gap and shows you the arithmetic to price your own.
The Urgency Story, and Why a Skeptic Should Distrust It
Search “cost of delaying AI” and you will find a wall of urgency: a “delay tax” of eight to fifteen thousand dollars a month, competitors pulling away, a compounding gap that never closes. Most of it is written for enterprises by firms selling the cure, and most of it quotes a headline dollar figure that turns out to be a vendor’s illustration rather than a measured result.
A commercial real estate owner is right to distrust it, because the same body of research contains a fact the urgency pieces leave out. Adoption was nearly universal and results were not. When roughly nine in ten firms run pilots and one in twenty gets what it wanted, “everyone else is doing it” is not a reason to move — it is a warning about how the move usually goes wrong. If the argument for acting now is only fear of missing out, the honest counter is that the firms who acted out of fear are the ones staring at a stalled pilot.
That does not mean waiting is free. It means the cost of waiting has to be measured against the right thing. The firms that got value did not simply buy more software than everyone else. They built a capability — people who know how to use the tools, and workflows and data shaped around that skill. That capability is what a year of waiting actually costs you, and it is worth being precise about why.
What Actually Compounds Over a Year
Three things compound for a small CRE firm that starts, and stay flat for one that waits. None of them is the software.
Team fluency. The single largest return in the first year comes from the general tools your team can already open — ChatGPT, Claude, Microsoft Copilot — applied to daily work: drafting letters of intent, summarizing a lease, turning comps into a market write-up, clearing an inbox. Industry adoption research puts the inflection around the 90-day mark, when a team that has practiced deliberately starts producing noticeably better output, and shows heavy users saving on the order of nine hours a week against roughly two for occasional dabblers. That gap is not a tool feature. It is a skill curve, and a skill curve only advances while someone is climbing it. A firm that waits a year does not start the year behind on software; it starts a year behind on the curve, and the curve is the expensive part to catch up on.
Proprietary data and workflow understanding. Your firm’s edge is not the model everyone can rent. It is your deal history, your document stack, and the specific way you underwrite, market, and close. The moment your team works with AI on real files, you begin learning which of your workflows are standard and cleanly automatable and which are the judgment-heavy, document-heavy processes where a build would pay off. That map is proprietary and it takes real reps to draw. A competitor who spent the year drawing it can scope a custom automation with precision; a firm that waited is still guessing. The groundwork for that discipline is the whole point of our guide to mapping the workflow before buying any tool.
Judgment about vendors. A fluent team reads a proptech demo differently. It knows what a general model plus a good prompt can already do, so it can tell which vendor features are substance and which are a thin wrapper priced like a platform. A firm that waited a year has not built that discrimination, which is precisely why its eventual buying spree tends to land in the failed-pilot statistics. The buying discipline that fluency unlocks is the subject of our field guide to evaluating AI vendors fairly.
The through-line: waiting a year does not pause the cost at zero and let you resume where you left off. It lets a gap open on skill, on proprietary understanding, and on judgment — three assets that grow with practice and cannot be bought back in a weekend.
The Cost-of-Delay Model You Can Run on Your Own Numbers
Ignore the headline dollar figures. The only credible number is the one built from your own inputs, and it takes four of them.
- Hours per week your team spends on tasks a fluent operator would compress with general AI tools: LOI and email drafting, lease and document review, comp research, first-pass market write-ups, formatting and cleanup.
- Loaded hourly rate for the people doing that work — salary plus overhead, not the billing rate.
- Realistic compression you would reach after a real ramp. Do not model the vendor’s fantasy; model something defensible, such as reclaiming a quarter to a third of those hours once the team is genuinely fluent.
- Ramp lag — the weeks before the compression is real. Treat the first quarter as climbing, not saving.
Worked example, illustrative only. Say a six-person firm spends 40 hours a week across the team on those tasks, at a loaded rate of 60 dollars an hour, and reaches a 30% compression after a one-quarter ramp. That is 12 hours a week reclaimed, worth about 720 dollars a week, or roughly 34,000 dollars a year once ramped — with the first quarter mostly ramp rather than return. Waiting a year does not defer that number; it forfeits it, and it also pushes your ramp a full year later, so the quarter you would have already climbed is a quarter you still have to climb whenever you finally start.
Two honesty checks keep this from becoming its own piece of urgency theater. First, the compression only materializes if the fluency is real, which is why the number is an argument for training, not for software. Second, this model deliberately excludes speculative revenue from “more deals screened” — that upside is real but unmeasurable in advance, and a number you cannot defend is worse than no number. Run the four inputs, keep it conservative, and you will have a delay figure you can actually stand behind in front of your partners.
Waiting Is Not the Same as Sequencing
The urgency content collapses a real distinction. “Waiting” and “moving deliberately” look similar from the outside and are opposites underneath.
Waiting means the clock on fluency, proprietary understanding, and vendor judgment has not started. Every week is a week not climbed. Sequencing means the clock has started on the cheap, compounding pieces while the expensive, irreversible ones are deliberately deferred until the evidence justifies them. A firm that runs a fluency workshop this quarter, spends two quarters learning which of its workflows actually bleed hours, and only then scopes a custom build is not waiting on AI — it is refusing to buy the wrong thing first. Its cost of delay on the compounding assets is near zero, because those assets are already growing.
This is the same discipline behind treating your firm’s technology functions as jobs to be covered rather than a department to be hired, which we lay out in rethinking IT for a firm with no IT department. Covering the enablement function — making the team fluent — is the piece that starts the compounding clock, and it costs a workshop, not a project. The full framework for which decisions to make now and which to defer runs through our buy-versus-build playbook and the small-firm operating manifesto that frames the whole approach.
When Waiting Is Actually the Right Call
An honest piece on the cost of waiting has to name where waiting is correct, because in one specific area it usually is.
Waiting on a custom build is frequently the right call. A bespoke automation for a proprietary workflow is expensive — market rates run roughly 25,000 to 150,000 dollars depending on scope — and largely irreversible once commissioned. If your team is not yet fluent enough to specify what the automation should do, or you have not confirmed which workflow bleeds the most hours, commissioning the build now is how firms end up in the failed-pilot column. Deferring that spend until the evidence is in is not delay; it is diligence.
What is almost never the right call is waiting on fluency. The cost of entry has fallen far enough that price is no longer the excuse — capable tools that cost around fifty dollars a month in 2019 now sit near twenty to thirty, and the business tiers your team would use publish enterprise data terms that, by default, do not train on your inputs. The barrier is not money and it is not tooling. It is the skill curve, and the skill curve is the thing that punishes waiting.
So the useful rule is a split decision. Defer the irreversible, evidence-hungry spend as long as you honestly lack the evidence. Do not defer the reversible, cheap, compounding capability, because that is the one whose cost of delay is real and whose downside if you start too early is trivial — a few thousand dollars and a team that got better at its job.
The Cheapest Move That Starts the Clock
The asymmetry is the whole argument. Starting small is cheap and reversible: a hands-on fluency workshop on your firm’s real work product runs a workshop budget, roughly 2,000 to 15,000 dollars at market rates, and its worst case is a modestly better team. Waiting is not reversible in the same way, because the skill curve, the proprietary map, and the vendor judgment you would have built do not wait for you to catch up — they simply do not exist yet, and a competitor’s do.
That is why the correct response to “should we wait a year” is almost never yes on the whole question and almost always yes on one part of it. Start the compounding clock now with the cheap move. Defer the expensive build until your own newly fluent team can tell you exactly what to build. You lose nothing you cannot afford to lose, and you stop forfeiting the one number in this whole debate you can actually defend.
Frequently Asked Questions
What does it actually cost a small CRE firm to wait a year on AI? Not the software — buying software, on its own, is what failed for most firms that tried. The measurable cost is a year of forfeited efficiency from an un-fluent team, plus a year of un-built proprietary understanding and vendor judgment that compound with practice. Price the efficiency piece with four inputs: weekly hours on AI-compressible tasks, loaded hourly rate, a conservative compression estimate, and a one-quarter ramp lag. That yields a figure you can defend; the headline “delay tax” numbers online cannot.
Isn’t it smarter to wait until the tools mature and prices drop? Prices already dropped — capable tools fell from around fifty dollars a month in 2019 to roughly twenty to thirty by 2025 — so price is no longer the barrier. Tool maturity does not help the part that matters, because what compounds is your team’s skill and your proprietary workflow understanding, neither of which a more mature tool builds for you. Waiting for maturity postpones the skill curve you will have to climb regardless.
If most AI pilots fail, doesn’t that argue for waiting? It argues for starting differently, not for waiting. JLL found roughly 88% of real estate firms had started pilots and only about 5% hit most of their goals, and Deloitte found reports of transformative impact collapsing to around 1%. Those failures cluster where firms bought software before building fluency. Starting with team fluency on real work, and deferring custom builds until the evidence is in, is the pattern that avoids the failed-pilot column.
What is the cost of delay and how do I calculate mine? Cost of delay is the value you forfeit for each period you postpone. Calculate it from weekly hours your team spends on tasks a fluent operator compresses, times a loaded hourly rate, times a conservative compression percentage, adjusted for a ramp period before the savings are real. A six-person firm spending 40 hours a week at 60 dollars loaded, reaching 30% compression, forfeits on the order of 34,000 dollars a year once ramped. Keep the estimate conservative and exclude speculative new revenue so the number survives scrutiny.
What compounds over a year — the tool, or something else? Not the tool. Three things compound: team fluency (heavy users save on the order of nine hours a week versus about two for dabblers, with the inflection near 90 days), proprietary understanding of which of your workflows are automatable, and judgment about which vendor features are real. All three grow only while someone is practicing, and none can be repurchased quickly after a year off.
Does waiting a year really put us behind bigger competitors? On raw tooling, no — you rent the same models they do. The gap that opens is on the compounding assets: a competitor who spent the year building fluency and mapping its workflows can scope automation with precision and read vendors accurately, while a firm that waited is still guessing. For a small firm, that judgment gap matters more than any tool a larger rival can afford.
What is the cheapest first step that starts the clock in the right direction? A hands-on fluency workshop on your firm’s real work — LOIs, lease summaries, market write-ups, email — using the general tools your team already has. At market rates that runs roughly 2,000 to 15,000 dollars, and its downside is trivial: a modestly more capable team. It starts the skill curve and gives you the fluency to make every later buying and building decision better.
When is waiting actually the right call? Waiting on a custom build is often correct. A bespoke automation runs roughly 25,000 to 150,000 dollars and is largely irreversible, so if your team is not yet fluent enough to specify it or you have not confirmed which workflow bleeds the most hours, deferring is diligence, not delay. Waiting on fluency is almost never right, because that is the cheap, reversible, compounding capability.
How long until a small firm sees real value from starting? Fluency gains tend to show around the 90-day mark for teams that practice deliberately on real work, not just attend a session. Build the ramp into any expectation: treat the first quarter as climbing the curve rather than banking savings, and model the return only after the team is genuinely fluent.
Does waiting a year change the buy-versus-build decision? It changes your ability to make it well. The buy-versus-build call depends on knowing which workflows are standard, which are proprietary, and which bleed enough hours to justify a build — knowledge you only accumulate by working with AI on real files. A year of waiting leaves that knowledge un-built, so you enter the decision guessing. A year of deliberate fluency lets you make the call on evidence.
Where to Start
Do not start by buying a platform, and do not start by commissioning a build. Start by pricing your own delay honestly: run the four inputs on your firm’s real hours, and the number will tell you whether the compounding assets are worth starting on this quarter. For almost every 4–20 person firm, they are, because the move that starts the clock costs a workshop and the thing it protects grows every week you own it.
The next step is to find the specific hours that number is made of. That is exactly what a free AI-readiness assessment does: a working session that inventories the workflows where your team’s hours disappear, estimates your real cost of delay, and returns a plain-language plan — including the honest calls about where the answer is a subscription, where it is a workshop, and where a build should wait until you have the evidence to scope it. Book a free AI-readiness assessment if you want that map for your firm. You will leave with a defensible number and a ranked plan, whether or not any of it turns into a project.
Arthur Wandzel
