The demo always lands. A senior person pastes a messy lease into an assistant, asks for a one-page summary, and the room goes quiet in the good way. Someone drafts a letter of intent in ninety seconds. Someone else turns a folder of comps into a clean market paragraph. Everyone leaves the session convinced this changes how the firm works — and for about two weeks, it does. Then a deal heats up, an inspection deadline lands, three tenants call the same afternoon, and the whole thing quietly evaporates. A month later the assistant sits unused and the team is back on the old workflow, a little embarrassed to bring it up. If that is the arc you recognize, the workshop did not fail. You hit the plateau, and the plateau is predictable. This piece explains why enthusiasm stops working, what the decline actually looks like week by week, and the small-firm system that restarts the climb without an IT department or a training team.
The demo high is real, and it is not adoption
The excitement after a good session is genuine, and it is worth having — it buys you permission to change how people work. The mistake is reading it as the finish line. A demo proves the tool can do the work. It says nothing about whether your team will reach for it at 4:45pm with an LOI due at five. Capability and habit are different things, and the gap between them is where most firms stall.
The numbers around this gap are stark. JLL’s 2025 Global Real Estate Technology Survey found that roughly nine in ten CRE firms are piloting or using AI, yet only about 5% report achieving all of their program goals. Deloitte’s 2026 Commercial Real Estate Outlook gives the feeling a name — “AI pilot fatigue” — the exhaustion that sets in when a firm launches with energy, sees no durable change, and quietly moves on. Almost everyone starts. Almost no one sustains. The plateau is not a rare failure mode; it is the default outcome, and knowing that changes how you plan the weeks after the demo.
Why enthusiasm stops working
Enthusiasm is a spike. It is high right after the session and it decays on its own, the way any motivation does, whether or not the tool is useful. Building your rollout on it is like building a budget on a single good month.
Two well-documented curves explain the fade. The first is memory. In the 1880s the psychologist Hermann Ebbinghaus measured how fast people forget newly learned material, and the “forgetting curve” he described has held up for more than a century: without deliberate reuse, most of what you learn in a session is gone within days. A broker who watched someone summarize a lease on Tuesday cannot reconstruct the prompt on the following Monday, tries once, gets a mediocre result, and concludes the tool “doesn’t really work for my deals.” The skill did not fail. The reps never happened.
The second is the arc every new technology travels. Gartner’s hype cycle calls the drop after the initial excitement the “trough of disillusionment” — the predictable slump when a tool that looked magical in a demo meets the friction of real work. Individual teams live a compressed version of that curve in the weeks after a workshop. The peak is the demo; the trough is the third week, when the novelty is gone and the habit has not formed. Firms that understand this plan for the trough. Firms that don’t read the trough as proof the whole thing was overhyped and stop.
There is a wider skills story underneath, too. The World Economic Forum’s Future of Jobs Report 2025 estimates that 39% of workers’ core skills will change by 2030, with AI and data skills the fastest-growing category. Fluency here is not a fact you memorize once; it is a practiced skill in a field that keeps moving. Skills like that live or die on repetition, and enthusiasm does not supply repetition. Only a system does.
The anatomy of the plateau, week by week
The decline has a recognizable shape. Naming it helps you spot it before it hardens.
- Week one — the high. Usage spikes. People try the assistant on real tasks, share screenshots, and talk about it in the hallway. Output quality is uneven but the energy is high. This looks like success, and it is the most misleading week.
- Week two — the narrowing. The whole team stops experimenting; two or three genuinely curious people keep going. Everyone else drifts back to their inbox and their spreadsheet, meaning to return to it when things calm down. Things do not calm down.
- Week three — the trough. A deadline hits. Under pressure, people reach for the workflow they trust, not the one they half-learned. The assistant goes untouched for days. The curious few start to feel like they are using something the firm has silently abandoned.
- Week four and beyond — the flatline. Use settles at a low baseline concentrated in one or two enthusiasts. The tool is not gone, but it is no longer a firm capability. When the enthusiast gets busy, takes vacation, or leaves, even that baseline disappears.
The tell is concentration. Healthy adoption spreads across roles — a broker, a property manager, an analyst all using it for their own work. A plateau concentrates: the same one or two names, doing the same one or two things, while the firm-wide capability the demo promised never materializes. We traced the deeper reasons this pattern repeats in why most AI training programs stall at real estate firms; the point here is that the flatline is a systems outcome, not a motivation problem you can pep-talk your way out of.
The three forces that pull a CRE team back
Three specific forces cause the reversion, and all three are stronger at a small commercial real estate firm than the generic change-management advice assumes.
Deadline reversion. Deal work runs on hard, external deadlines — an LOI due before a competing offer, a due-diligence clock, an estoppel that has to go out today. Under that kind of pressure, people default to the fastest known path, and a skill that is two weeks old and half-practiced is not the fastest known path. It is a risk. So the assistant gets skipped exactly when it would help most, and every skip makes the next one more likely. This is the force generic advice misses entirely, because most workplaces don’t run on the same relentless external clock a brokerage does.
No default. After the session, nobody decided which tool the firm uses for which task. One person opens ChatGPT, another has Microsoft Copilot in their email, a third remembers something about Claude from the demo. With no default, every use starts with a small decision, and small decisions under deadline pressure resolve toward the old way. A habit needs a default; a firm that leaves the tool choice to each person on each task has not set one.
No reps in the real work. The workshop created the skill in a sandbox — practice leases, sample LOIs. But the firm never changed how an actual deliverable gets produced. The market report still gets written the old way; the lease abstract still gets keyed by hand. If the new method is not the way the real work is now done, the reps never accumulate, and without reps the forgetting curve wins. The teams that retain the most, as we found in the lessons from a dozen AI workshops, are the ones who left the room having already rewired one real recurring task, not the ones who saw the most impressive demo.
What actually restarts the climb
The fix is not more enthusiasm and it is not more tools. It is a small operating system that replaces motivation with habit. None of it requires an IT department or a learning team — which is precisely why a lean firm can run it and a slow institution often cannot.
Set one default assistant and stop there
Pick one business-tier assistant for the whole firm — ChatGPT Business, Claude Team, or Microsoft Copilot — and make it the default for language and lookup work. Business-tier access, at roughly $20 to $30 per user per month, matters for a second reason beyond consistency: under those terms your confidential deal data stays out of model training, which is the security floor for a firm handling non-public financials. One assistant, firm-wide, is the entire day-one stack. Resist the urge to sign up for five specialized apps before the team is fluent on one; tool sprawl deepens the plateau, it does not cure it.
Build the reps into work that already exists
Take one recurring deliverable — the weekly market update, the lease abstract, the LOI first draft — and make the assistant the standard way it gets produced, starting now. Not a suggested option; the default method. When the reps live inside work that has to happen anyway, repetition takes care of itself and the forgetting curve never gets its footing. Start with one task, prove it holds for a month, then add the second. A shared, maintained prompt library turns one person’s good result into the whole firm’s starting point, so nobody faces a blank box under deadline.
Name a champion, not a committee
Adoption at a small firm needs one accountable person, not a working group. The champion is not the most technical person; it is the respected operator who keeps the prompt library current, answers “how do I get it to do X,” and gently notices when someone has quietly stopped using it. This role is the difference between a habit that spreads and one that concentrates in whoever happened to be most excited. We lay out the champion’s remit and the full ninety-day sequence in the CRE AI training playbook.
Run a thirty-minute demo every month
Put a standing thirty-minute session on the calendar where two or three people show one real thing the assistant did that month — a tricky lease clause it caught, a market paragraph it drafted, a tenant email it softened. This does three jobs at once: it refreshes the skill against the forgetting curve, it surfaces what changed in the tools that quarter, and it keeps the practice social so the trough never sets in. It is the cheapest reinforcement you can buy, because the cost is protected time, not a fee.
Measure recovered hours, not logins
Track the outcome that matters to a partner, not vanity usage. Ask producers how many hours a week they are getting back on drafting and lookup, and watch that number rather than login counts. Recovered hours tie the habit to money, which keeps the effort funded when the initial excitement is long gone. Funding that reinforcement deliberately, rather than paying for a session and hoping, is the reframe we made in rethinking the training budget for the AI era — the money that matters is the small, continuous amount that keeps the habit alive, not the one-time cost of the demo everyone remembers.
Enthusiasm is a starting condition, not a strategy
The healthiest way to think about the post-demo high is as fuel for the launch, not the engine of the flight. It gets you off the ground. What keeps you climbing is a default tool, reps embedded in real deliverables, a named champion, a monthly rhythm, and a metric a partner can see. Put those five in place and the plateau flattens into a slope, because the team is no longer relying on how they feel about the tool — they are just doing their work the new way.
This is where a small firm has a real edge over a large one. The broader case for that edge runs through the small-firm AI manifesto: a 200-person company answers the adoption problem with a change program and a committee; a 10-person shop answers it with one default, one champion, and one recurring task rewired this week. You can move before the excitement fades. The firms that do are the ones still using AI a year after the demo, quietly faster than the competition that ran the same workshop and let the enthusiasm run out.
FAQ
Why does AI enthusiasm stop working after a workshop?
Because enthusiasm is a temporary spike, not a durable system. Motivation decays on its own within days, the forgetting curve erases an unpracticed skill, and deal deadlines pull people back to familiar workflows. Unless the new method becomes the default way a real recurring task gets done, usage concentrates in one or two enthusiasts and then flatlines. The fix is a habit-based system — a default tool, embedded reps, a champion, and a monthly rhythm — not another burst of excitement.
Is a plateau after the demo a sign the workshop failed?
No. The plateau is the default outcome, not a rare failure. JLL found roughly nine in ten CRE firms piloting AI but only about 5% hitting all their goals, and Deloitte named the broader pattern “AI pilot fatigue.” A good demo proves capability; sustained use requires a separate system built in the weeks afterward. Reading the trough as proof the workshop was overhyped is the mistake that turns a temporary dip into a permanent stop.
How long after training does AI usage typically drop off?
The decline usually shows within three to four weeks. Week one is a usage high, week two narrows to a few curious people, week three hits a trough when a deadline forces reversion to old habits, and by week four use settles at a low baseline concentrated in one or two enthusiasts. Planning reinforcement before that third week is what keeps the curve from flattening.
What is the single most important thing to fix the plateau?
Embed the reps into work that already exists. Take one recurring deliverable — a market update, a lease abstract, an LOI draft — and make the assistant the standard way it is produced starting now. When repetition lives inside work that has to happen anyway, the forgetting curve never gets traction, which matters more than any one prompt technique or tool choice.
Why do deadlines make AI adoption harder at a CRE firm?
Because deal work runs on hard external clocks, and under pressure people default to the fastest known path. A two-week-old, half-practiced skill is not the fastest known path — it feels like a risk — so the assistant gets skipped exactly when it would help most, and every skip makes the next one more likely. This deadline-reversion force is stronger in commercial real estate than in workplaces without the same relentless external timelines.
Do we need more AI tools to get past the plateau?
No — more tools usually deepen the plateau. With no default, every use starts with a small decision about which tool to open, and under deadline pressure those decisions resolve toward the old way. Pick one business-tier assistant for the whole firm and make it the default. Add a purpose-built tool only when a specific recurring task clearly justifies it, after the base habit holds.
Who should own AI adoption at a small firm?
One named champion, not a committee. The champion should be a respected operator rather than the most technical person: someone who keeps the shared prompt library current, answers quick “how do I get it to do X” questions, and notices when usage is quietly fading. A single accountable person is what turns a habit that would otherwise concentrate in one enthusiast into one that spreads across roles.
How do we know if adoption is actually sticking?
Watch for spread, not just volume. Healthy adoption shows up across roles — a broker, a property manager, and an analyst each using the assistant for their own work — while a plateau concentrates in the same one or two names doing the same one or two tasks. Track recovered hours per producer rather than login counts, because that ties the habit to money and keeps the effort funded after the initial excitement fades.
How much should reinforcement cost after the workshop?
Reinforcement is mostly protected time, not a large fee. Business-tier tool access runs about $20 to $30 per user per month, and the monthly thirty-minute demo, the maintained prompt library, and the champion’s attention cost staff hours rather than invoices. A facilitated fluency session itself typically falls in the roughly $2,000 to $15,000 range depending on format, but the money that determines whether it sticks is the small, continuous amount spent keeping the habit alive.
Key takeaways
- The post-demo high is real but it is not adoption; a workshop proves the tool can do the work, while sustained use is a separate habit built in the weeks afterward.
- The plateau is predictable and has a shape — a week-one high, a week-two narrowing, a week-three trough at the first deadline, and a week-four flatline concentrated in one or two enthusiasts.
- Three forces pull a CRE team back: deadline reversion to known workflows, no default tool, and no reps embedded in real deliverables — the forgetting curve does the rest.
- The restart is a small operating system, not more enthusiasm: one default assistant, reps built into existing work, a named champion, a monthly thirty-minute demo, and recovered hours as the metric.
- Enthusiasm is a starting condition, not a strategy — and running that system without an IT department or a learning team is exactly where a lean firm out-executes a slow institution.
Want to know where your team actually stands after the demo — and what would make it stick? A short conversation about your workflows, your deadlines, and where usage is fading will tell you more than any survey average. Book your free AI-readiness assessment → and we will map what durable fluency would take for your firm.
Arthur Wandzel
