Commercial real estate is late to AI because it is a relationship business built on documents, run by small firms with no IT department and burned enough times by proptech to distrust the next “transformative” tool — and for a 4-to-20-person firm, arriving late is not the handicap it looks like. When a technology is scarce, being early wins: you lock in an advantage before anyone else can. AI in 2026 is the opposite case. The capability is abundant, cheap, and improving by the month, so a firm that waited skipped the expensive, immature phase entirely — the failed pilots, the overpriced point tools, the custom builds on models that are now obsolete. What converts AI into an edge today is not capital or a data-science team. It is fluency and the speed to change how you work, and on both a small firm is structurally advantaged over the institution. This piece explains why the industry lags, and how a lean firm turns lateness into a lead.
Why Commercial Real Estate Is Actually Late
The lag is real, and it is worth naming honestly before reframing it. McKinsey’s Global Institute has for years ranked real estate near the bottom of its industry digitization index, below retail, manufacturing, and every kind of finance. Deloitte’s 2026 Commercial Real Estate Outlook found that roughly 88 percent of surveyed owners and investors are now piloting or using AI, yet the share reporting a genuinely transformative impact sits near 1 percent. The industry is not ignoring AI. It is adopting it slowly and getting little back, which is a different problem — and a more revealing one.
Five structural reasons explain the pace, and none of them are about intelligence or ambition.
It is a relationship-and-analog business. Deals close on trust, phone calls, referrals, and site visits. The core product — matching space to capital — resisted digitization far longer than software-native industries, because the valuable part happened in a conversation, not a database. An industry whose best people succeed by being in the room has little native pull toward automating the desk work around it.
The data lives in documents. Leases, rent rolls, offering memorandums, estoppels, and CAM statements arrive as PDFs and scans, often decades old, formatted a hundred different ways. Older software could not read them, so the payoff from automating anything downstream stayed locked behind a document-parsing problem nobody had solved cheaply. The raw material of the business was, until recently, unreadable by machines.
The firms are small and have no IT function. Most commercial real estate runs through firms of a handful of people. There is no CIO, no analyst bench, no one whose actual job is to evaluate a new tool, run a pilot, and roll it out. New technology in a small firm competes for the owner’s evenings, and the owner is closing deals.
Proptech fatigue is earned. A decade of subscriptions that promised to change everything and mostly added another login left principals rationally skeptical. When the next tool arrives calling itself transformative, the reflex — built from experience — is to wait and see. That reflex slowed AI adoption specifically, because early AI features looked exactly like the last five things that underdelivered.
Confidentiality anxiety is legitimate. Deal terms, tenant financials, and investor information are sensitive, and “don’t put it in a computer you don’t control” was a defensible instinct for years. That caution, sound as it was, kept a lot of firms off the cloud and off AI longer than firms in less confidential lines of work.
Put together, these are not failures of nerve; they are the accurate reasons a document-heavy, relationship-driven, small-firm industry moved last. The interesting question is what moving last actually costs — and this time, the answer is not what it usually is.
When Being Late Costs You — and When It Doesn’t
The fear of being late comes from a real pattern, so it is worth stating the case fairly. In some technology shifts, the early mover wins permanently. When a capability is scarce and compounding — a proprietary dataset, a network of users, a distribution channel nobody else can access — the firm that gets there first builds a lead that later entrants cannot close. Miss that window and you are structurally behind, forever paying to rent an advantage a competitor owns.
That is the model most principals have in their head when they worry about being late to AI. It is the right model for the wrong technology.
Being late is only a penalty when the thing you were late to is scarce. When the capability is abundant, cheap, and still improving, late arrival works the other way: you inherit a mature version of the technology at a fraction of the cost the pioneers paid, and you skip the phase where everyone was paying to learn what does not work. The classic example is emerging markets that never built copper telephone lines and jumped straight to mobile — the lateness was not a lag, it was a leapfrog onto the good version of the technology directly.
AI for commercial real estate is a leapfrog case, not a scarce-capability case. Which means the firms that waited did not miss a window. They avoided a tax.
Why Late Is an Advantage This Time
Consider what the early movers in CRE AI actually bought over the past few years, and what a firm starting now gets to skip.
They bought immature models. Early document-AI tools were trained on models that hallucinated more, cost more per page, and handled messy lease PDFs worse than what is available today. A firm that built its lease-abstraction workflow on a 2023-era model is now maintaining something the current generation does better out of the box.
They bought overpriced point tools. When a general-purpose model could not yet draft a decent market write-up or summarize a lease, single-purpose tools charged hundreds of dollars a month for that one function. Today a principal who can write a clear prompt gets a first-draft lease summary, a letter-of-intent draft, or a market note from ChatGPT, Claude, or Gemini for a few dollars a seat. The firm that waited never paid the premium for the scarce version.
They funded the failed pilots. That 88-percent-piloting, 1-percent-transformative gap in the Deloitte data is not evidence that AI does not work. It is a record of firms that bought tools before they understood their own workflows — a model dropped onto an unexamined process produces a pilot, not a result. A firm arriving now can read that gap as a map of the mistakes already made and paid for by someone else.
The late firm inherits cheaper, better, more reliable tools and a public catalog of what not to do. Its only real disadvantage — that its people are not yet fluent — is the one that is fast and inexpensive to fix, and it was never something you could buy early anyway. The starting line moved toward the late arrivals, not away from them. For the fuller version of this argument — that a disciplined small firm can out-operate a larger, earlier competitor — see the small-firm CRE manifesto.
The Advantage Only a Small Firm Has
The leapfrog helps everyone who waited. The second advantage belongs specifically to the small firm, and it is the one institutions cannot copy.
What converts AI into results in 2026 is not a bigger budget. It is the speed at which an organization can change how it works — adopt a new practice, retrain a habit, redraw a workflow. On that dimension, a 10-person firm is not the underdog. It is the favorite.
At an institutional owner or a national brokerage, adopting AI means a committee, a procurement process, a security review, a legal sign-off, a pilot in one region, a change-management program, and a training rollout across hundreds of people who each have their own way of working. The larger the firm, the more the very things that make it powerful — process, scale, governance — slow the adoption of a fast-moving tool. That is exactly why the biggest players are stuck in the pilot phase; scale is working against them here.
At a small firm, the owner decides on Monday and the firm works differently by Friday. There is no committee to convince, no legacy system that a new practice has to route around, no hundred-person retraining. The whole team fits in one room. A single well-run session can shift how five or eight people handle leases, market notes, and email, and the change sticks because everyone who needs to adopt it was there. How that plays out day to day is the subject of our look at how AI is changing commercial real estate brokerage.
There is also no legacy AI debt. The firm that waited has no half-built integration to maintain, no abandoned tool subscription to unwind, no team trained on a workflow that a newer model made obsolete. It gets to start on the current version, once, with a clean process — and starting clean, late, on mature tooling beats starting early on tooling that aged out from under you.
How to Turn Late Into Ahead
The advantage is real but it is not automatic. Lateness only converts into a lead if the firm moves deliberately now instead of repeating the mistake that produced the 1-percent gap — buying a tool before understanding the work. Three moves, in order.
Get fluent before you buy anything. The highest-return move for a small firm is not a subscription. It is teaching the team to use the general-purpose models it can already access to do real CRE work: prompting ChatGPT, Claude, or Gemini to draft a first-pass lease summary, a market write-up, a letter of intent, or a cleaned-up client email. This is a low-thousands investment, not a capital project, and it does two things at once — it captures value immediately, and it tells you which workflows are genuinely worth automating later, because you can only write a good spec for a task you have already done by hand. What “fluent” actually means for a CRE team, and how to build it in about 90 days, is laid out in the CRE AI training playbook. And if you want the single skill that matters most, it is prompting itself — why every CRE professional needs it.
Scope from the work, not the tool. Once the team is fluent, the firm can see clearly which parts of its week are still slow — usually a document-heavy or cross-system task no prompt fully solves, like reconciling CAM charges or consolidating rent rolls across properties. That is a candidate for automation. The order matters: fluency first exposes the real bottleneck, so you are scoping a problem you understand rather than buying a solution to a problem you assumed you had.
Build only the slice that is yours. Rent the standard, industry-wide layer — market data, e-signature, listing syndication, CRM — from vendors who spread the cost across thousands of firms. Reserve a custom build for the one workflow that is specific to how your firm makes money and that no vendor bridges. A single well-scoped automation for a small firm generally runs in the market range of $25,000 to $150,000 depending on complexity, with the mature tooling now making those builds cheaper and more reliable than they would have been for an early mover.
Done in that order, being late stops being a source of anxiety and becomes what it actually is: a firm that skipped the expensive lessons, inherited the good tools, and can change faster than anyone bigger. The only way to waste the advantage is to keep waiting — or to break the sequence and buy a tool before the team knows what to ask of it.
Frequently Asked Questions
Why is commercial real estate late to adopting AI?
Commercial real estate is late because it is a relationship-driven, document-heavy industry made up mostly of small firms with no IT department. Deals close on trust and site visits, the core data lives in PDFs and scans that older software could not read, principals are rationally skeptical after a decade of proptech that underdelivered, and deal confidentiality made firms cautious about new tools. None of these are failures of ambition; they are the accurate reasons a document-heavy, small-firm industry moved last.
Is being late to AI actually a disadvantage for a CRE firm?
Not this time. Being late is a penalty when a technology is scarce and early movers lock in an advantage. AI in 2026 is abundant, cheap, and still improving, so a firm that waited skipped the expensive, immature phase — the failed pilots, the overpriced point tools, and the custom builds on models that are now obsolete. It inherits mature, cheaper tools and a public record of what does not work. Its only real gap, team fluency, is the fastest and least expensive one to close.
What does the leapfrog effect mean for small real estate firms?
The leapfrog effect is when a late adopter skips an entire generation of expensive, obsolete technology and lands directly on the mature version — the way many markets skipped copper phone lines and went straight to mobile. For a small CRE firm it means you do not have to buy the early, costly versions of AI tools that pioneers paid to learn from. You start on current-generation models that are cheaper and more reliable, having avoided the tax the early movers paid.
If 88 percent of firms are already piloting AI, hasn’t my firm missed the window?
No. The Deloitte figure showing about 88 percent piloting AI but only around 1 percent reporting transformative impact is not evidence of a closed window. It is a record of firms that bought tools before understanding their own workflows, which produces pilots rather than results. A firm starting now can read that gap as a map of mistakes already paid for by others and avoid them by building fluency first.
Why do small firms have an adoption advantage over large institutions?
Because what converts AI into results is the speed of changing how you work, and small firms change fastest. A large institution needs committees, procurement, security review, legal sign-off, and a retraining program across hundreds of people. A 10-person firm’s owner decides and the firm works differently within the week, with the whole team in one room and no legacy systems to route around. Scale, which usually helps, works against the giants on a fast-moving tool like AI.
What is the first step for a CRE firm that feels behind on AI?
Get the team fluent before buying anything. Teach everyone to use general-purpose models — ChatGPT, Claude, Gemini — for real work like drafting lease summaries, market write-ups, letters of intent, and client email. This is a low-thousands investment that captures value immediately and reveals which workflows are actually worth automating later. Buying a tool first is what produced the industry’s poor pilot results; fluency first is the correction.
Do we need to hire a data scientist or build custom AI to catch up?
No. The capability that matters now is fluency with tools you can already access for a few dollars a seat, not a data-science team or a custom build. Rent the standard, industry-wide layer — market data, e-signature, CRM — and reserve a custom automation, typically in the $25,000 to $150,000 market range, only for the one workflow specific to how your firm makes money that no vendor covers. Fluency comes first and costs the least.
Isn’t it safer to keep waiting until AI matures further?
The tools are already mature enough to pay off, and the one thing that still needs building — your team’s fluency — only takes longer the longer you wait. Continuing to wait does not lower risk; it delays the low-cost step that turns your late position into a lead while competitors who move now start compounding the advantage. The disciplined move is to start with fluency and scope from there, not to wait longer.
Can general AI tools really replace the proptech we were told to buy?
For many tasks, yes. A general-purpose model now drafts first-pass lease summaries, market notes, and letters of intent that used to require a dedicated tool, which is why the case for expensive single-function subscriptions has weakened. It cannot replace a market dataset, a CRM’s system of record, or e-signature infrastructure, which depend on data and integrations a model does not have. Use general models for language work, rent the platforms that own the data and workflow, and build only what is specific to your firm.
Where to Start
The mistake is to read “we’re behind” as a reason to rush out and buy something — that is exactly the move that produced a decade of shelfware and the industry’s 1-percent results. The better read is that your firm skipped the expensive phase and can now start clean, on mature tools, faster than anyone bigger. The first step costs almost nothing: get your team fluent with the models you can already access, then let that fluency show you what, if anything, is worth automating. A free AI-readiness assessment gives you that starting map — a short working session that looks at your workflows, where your hours actually go, and where a lean firm’s speed advantage is largest, then returns an honest recommendation on where to begin. Book a free AI-readiness assessment and turn a late start into a lead.
Arthur Wandzel
