The skeptic in your office is usually right, and that is the part most rollout advice gets backwards. When a top producer says she is not pasting a client’s deal terms into a chat window, or when a partner points out that a confident-sounding AI summary got a lease clause wrong, they are not being difficult. They are pressure-testing a tool that handles confidential information and carries real liability, which is exactly what you want people doing before they trust it. A 2026 industry pulse check of 255 commercial real estate professionals by First American Data & Analytics and DealGround found that despite near-universal usage, only 5% trusted AI enough to inform an actual deal decision. That gap between using and trusting is the whole problem, and you do not close it by winning an argument. You close it by running the rollout so the work itself answers the objection. These are the ten rules for doing that at a firm where you are the entire change-management department.
A note on where this fits before the rules. The full 90-day arc of getting a lean team fluent lives in our CRE AI training playbook, and the broader case for why small shops can out-operate larger competitors sits in the small CRE firm manifesto. This piece is about the human friction those plans run into: what to do when half your office is in and half is folded arms.
Rule 1: Assume the Skeptic Is Right Until Proven Otherwise
The fastest way to lose a rollout is to treat doubt as a character flaw. In commercial real estate the loudest objections are usually the smartest ones: a hallucinated cap rate, a misread lease clause, a client name sitting in a consumer tool’s training data. Those are not excuses to avoid change; they are the exact risks a careful operator should flag.
So start by conceding the point. When someone says the output can be wrong, agree, because it can. When someone says the confidential data question is not settled, agree, because until you write the rule it is not. Granting the objection does two things. It signals that this is not a loyalty test, and it reframes the skeptic from obstacle to quality control. The person poking holes in the tool is the person you most want checking its work later.
Rule 2: Name the Fear Before You Name the Tool
Underneath the technical objections are three quieter fears, and every rollout that leads with a product demo steps right over them. People worry that the tool will replace them, that it will make their job harder before it makes it easier, and that they will look foolish for not already knowing how to use it.
Say those out loud before you open a single app. Be specific about the first one, because in a brokerage or a lean investment shop nobody is getting replaced by a chat assistant; the work is judgment, relationships, and accountability that no model carries. What changes is the hour of document grunt work that surrounds the judgment. Naming the fear does not eliminate it, but an unspoken fear runs the meeting from the back row. A named one can be answered.
Rule 3: Never Mandate; Let the Work Convince
At a 12-person firm you cannot fire your way to adoption, and you certainly cannot order a co-principal or a top producer to change how they work. Authority is the weakest tool you have here, and reaching for it converts a fixable culture problem into a standoff.
Voluntary adoption is not softness; it is the only mechanism that actually holds. McKinsey’s research on organizational change has long put the failure rate of large change programs near 70%, with employee resistance and management behavior the barrier most often blamed. A mandate produces compliance theater, people who open the tool when you are watching and abandon it when you are not. Demonstrated value produces pull, where a broker asks the colleague next to her how she turned that market write-up around so fast. You are aiming for the second one, and it only arrives if using the tool is a choice people make because it plainly helps.
Rule 4: Write the Data Rule Before Anyone Touches a Real Deal
The single most common skeptic objection in this industry is some version of “I am not putting our clients’ information into ChatGPT.” That objection is correct, and it stays correct until the firm writes down what is safe to share and where. Skip this step and you validate the skeptic permanently.
It does not take a governance committee. It takes one page: which tools are approved, what data is fine to paste (public listings, generic prompts, hypotheticals), and what never leaves the firm (client names tied to financials, unexecuted terms, anything under NDA). Default the firm to business-tier accounts, whose settings keep your inputs out of model training, rather than personal free logins; confirm the current terms for whichever assistant you approve, since they change. Getting this right early is enough of a discipline on its own that it is worth treating as its own project, which is the argument in our piece on why banning ChatGPT backfires and what to write instead.
Rule 5: Rig the First Task So It Cannot Fail
The first thing a skeptic tries with AI decides everything, so do not leave it to chance. The instinct is to showcase the tool on something impressive, a full underwriting memo or a complex lease abstraction. That is precisely the task most likely to produce a confident, subtly wrong answer that confirms every doubt in the room.
Pick the opposite. The right first task is high-annoyance, low-stakes, and low-confidentiality: drafting a first-pass reply to a routine listing inquiry, cleaning up notes into a readable email, turning a few bullet points into a serviceable market blurb. The work should be something the skeptic finds tedious and the tool handles in seconds, so the first experience is relief rather than risk. Save the high-stakes documents for after trust is earned, not before. For a menu of these safe entry points mapped to actual brokerage tasks, our field guide to AI use cases for a small brokerage lays them out.
Rule 6: Start With Your Loudest Doubter, Not Your Biggest Fan
Every office has an early adopter who will use whatever you put in front of them. Converting that person proves nothing, because the room already discounts their enthusiasm. The endorsement that moves a firm comes from the person who was against it.
So invest your best attention in the skeptic, not the enthusiast. Sit with them, pick a task they personally hate, and let them drive while you help them get one clean, useful result on their own work. A doubter who gets a genuine win becomes the most credible advocate you have, precisely because everyone knows they were not inclined to like it. The enthusiast’s praise sounds like personality; the former skeptic’s praise sounds like evidence.
Rule 7: The Principal Cannot Be the Visible Skeptic
If you are leading this rollout while privately rolling your eyes at the tool, the firm will read it instantly and follow your real position, not your stated one. Senior-level doubt is the most contagious kind, because it gives everyone permission to opt out without consequence.
This does not mean performing fake enthusiasm, which reads as hollow and erodes the trust you need. It means doing the work yourself, visibly, on your own tasks, and being honest about both the wins and the limits. A principal who says “this saved me twenty minutes on the LOI draft, and I still checked every number” models the exact behavior the rollout is trying to install: real use, paired with real judgment. Your team calibrates to what you do with the tool, not to what you say about it in a meeting.
Rule 8: Practice on Your Own Leases, Not Vendor Samples
Skepticism hardens fastest when training happens on a vendor’s tidy sample documents that look nothing like the messy PDFs your team actually works with. The skeptic’s whole point is “that is not how our deals look,” and a demo on clean fixtures proves them right.
Train on the firm’s real material instead: an actual lease from your files, a real LOI, a market write-up in the format your clients expect. This does two things at once. It shows the tool coping with your genuine complexity, scanned pages, odd clauses, local conventions, and it builds a skill that transfers, because the team is practicing on the exact work they do all day. The ability to describe a task clearly and judge the result carries across ChatGPT, Claude, Gemini, and Microsoft Copilot, so you are building durable fluency rather than muscle memory for one menu.
Rule 9: Keep a Human Check, and Say So Out Loud
The reliability objection deserves a permanent answer, not a one-time reassurance. Build a visible verification step into every workflow and name it as policy: a human reads and owns every output before it goes to a client or into a deal file. Nothing leaves the firm on the model’s authority alone.
Stating this openly does more than manage risk. It tells the careful people on your team that their caution is now part of the process rather than an argument against it, which is the fastest way to bring a skeptic inside. It also matches how the industry actually behaves: that same First American and DealGround pulse found only 5% of professionals ready to let AI inform a live deal decision, which is a rational stance, not a laggard one. The workflow that assumes a human check is the one that both earns trust and deserves it.
Rule 10: Measure Behavior, Then Standardize the Wins
Rollouts drift because firms track the wrong thing. Attendance at a training session, logins provisioned, a satisfied exit survey, none of these tell you whether the work changed. Watch behavior instead: weekly active use of the firm’s primary tool, aimed at near-universal within a couple of months, and the time it takes to produce a fixed set of real jobs before and after.
Then capture what works so it survives the next busy season. When a broker lands on a prompt that reliably produces a solid first-draft LOI, write it down where everyone can grab it and name someone to own the shared library. This is what keeps the effort from evaporating when the person who figured it out gets busy or leaves. The NAR 2025 Technology Survey, drawing on a large national sample of Realtors, found 68% now using AI but only 17% reporting a significant positive business impact. The firms in that 17% are not the ones with the most enthusiasm; they are the ones that measured real work and standardized what moved it.
The Pattern Behind the Rules
Read the ten back to back and one idea runs through all of them: you do not beat the skeptic, you let the work beat the skeptic. Every rule removes a reason to doubt rather than an argument against doubting. You concede the risks, write the data rule, rig the first task, keep the human check, and measure real output, and the objections dissolve because they have been answered in practice rather than in a meeting.
This is also why order matters as much as the rules themselves. Conceding the objection buys you the room to write the data rule; the data rule makes the first safe task possible; the safe task converts the doubter; the doubter’s win pulls the rest of the firm. The full ordered version of that sequence, from mapping your workflows to compounding the wins into automation, is laid out in our AI adoption framework for small CRE firms. Skepticism, handled this way, stops being the thing blocking your rollout and becomes the thing that makes it trustworthy.
Frequently Asked Questions
How do I get a resistant team to actually use AI?
Stop trying to persuade and start removing reasons to doubt. Concede that the risks are real, write a one-page rule for confidential data, and pick a first task that is tedious and low-stakes so the initial experience is a clear win rather than a subtle failure. Resistance at a small firm rarely yields to a better pitch; it yields to demonstrated value on the person’s own work. A single genuine result on a task they personally hate does more than any deck, because it answers the objection in practice instead of in argument.
Should AI use be mandatory or optional?
Optional, especially at a small firm. You cannot meaningfully mandate a co-owner or a top producer, and a mandate tends to produce compliance theater, people who use the tool when watched and abandon it otherwise. Voluntary adoption driven by visible value is the only version that holds, which is consistent with change research showing roughly 70% of forced change programs miss their goals largely because of resistance. Make using the tool an obviously easier way to do a hated task, and pull will do what a rule cannot.
What if my most skeptical person is a top producer or co-owner?
Treat them as your most valuable convert, not an obstacle to route around. At a lean firm the loudest doubter often has the most sway over the firm’s revenue, so their eventual endorsement carries more weight than anyone’s. Invest direct time helping them get one clean result on a task they find tedious, on their own real work. A top producer who was against AI and then found it genuinely useful becomes your single most credible advocate, precisely because the room knows they were not inclined to like it.
How do I handle the fear that AI will replace jobs?
Name it directly and be honest, because in a brokerage or lean investment shop it is not true and people can tell when you are dodging. The work that defines these jobs, judgment, relationships, accountability for a deal, is not something a chat assistant carries. What changes is the hour of document drafting and email triage that surrounds the judgment. Framing AI as removing the grunt work rather than the person is both accurate and the only message that survives contact with a skeptical, experienced team.
What is the safest first task to roll out?
A task that is high-annoyance, low-stakes, and low-confidentiality: drafting a first-pass reply to a routine inquiry, tidying rough notes into a clean email, or turning bullet points into a serviceable market blurb. These give a fast, obvious win without risking a client relationship or exposing sensitive data. Avoid leading with complex, high-stakes work like full underwriting or detailed lease abstraction, where a confident but subtly wrong answer confirms every doubt in the room. Earn trust on the easy work first, then graduate to the documents that matter.
How long should a rollout to a small firm take?
Weeks to a few months for working fluency, not a single event. Writing the data rule and picking first tasks takes days; the first wins can come the same week. What takes time is the reinforcement that turns a first success into a habit, which is why the standardizing stage runs on the order of a couple of months, aiming for near-universal weekly use. Anyone promising firm-wide transformation from one afternoon workshop is selling attendance, not capability.
How do I know the rollout is actually working?
Measure behavior on real work, not attendance or logins. Track weekly active use of the firm’s main tool and aim for near-universal within a couple of months, and time a fixed set of real jobs, such as an LOI draft or a market write-up, before and after. If those numbers move, adoption is real. The national picture shows the gap plainly: most Realtors now use AI, but only about one in six report a significant business impact, and the difference is measurement and standardization, not enthusiasm.
Do we need to hire someone or buy software first?
No, and doing either first is a common mistake. The rollout is a set of owner decisions, not a purchase: concede the risks, write the data rule, choose safe first tasks, convert the skeptic, keep a human check, and measure. The tools most firms need to start are business-tier accounts of assistants your team can already use, such as ChatGPT, Claude, Gemini, or Microsoft Copilot. Custom software comes much later, only once a workflow is standardized enough that hand-work is the bottleneck, and market rates for that kind of build run well into the low six figures.
Where to Start
The first move is not a purchase or a mandate; it is figuring out which objections at your firm are the real ones and which safe first task will earn a win. That is exactly what a free AI-readiness assessment produces: a working session that identifies your highest-friction workflows, flags the confidential-data rules you need before anyone starts, and gives you an honest read on whether you need a full LLM-fluency workshop, a lighter course, or just clearer rules for the tools your team already has. Market rates for structured fluency training run in the low thousands to low tens of thousands, but the assessment costs nothing and often shows a divided office is closer to aligned than it looks. Book a free AI-readiness assessment and you will leave knowing which rule to run first, and which skeptic to start with.
Arthur Wandzel
