An AI note-taker records a meeting, turns the audio into a transcript, labels who said what, and writes a short summary with action items. For a commercial real estate firm the real question is not how it works but which meetings it belongs in. On internal calls like pipeline reviews, team check-ins, and vendor conversations, it is close to free value: a written recap and a set of follow-ups nobody had to type. On client meetings, where the conversation covers NDA-bound deal terms, seller motivations, and pricing strategy, the same tool raises two questions the vendors rarely surface. Is it legal to record this call, and should a third-party vendor hold the recording? Both are answerable, and neither is a reason to avoid the technology. They are reasons to decide, once, where it runs and on what account, which is exactly the kind of decision a small firm can make in an afternoon.
How AI Note-Takers Actually Work
Every AI note-taker, whatever the brand on it, runs the same four steps. Understanding them is enough to judge the tool, because the failure points sit inside the steps.
First, it captures the audio — either by joining the call as a participant or by listening through the meeting platform you already use. Second, it runs speech-to-text, converting the spoken words into a written transcript. Third, it does speaker labeling, the part that guesses which voice belongs to which person so the transcript reads “Sarah:” and “the buyer:” instead of one undifferentiated block. Fourth, it feeds that transcript to a language model that writes a summary and pulls out action items — the two-paragraph recap and the “follow up on the estoppel by Friday” list that most people actually read instead of the full transcript.
The quality drops off in predictable places. Speech-to-text mishears proper nouns, so tenant names, submarket names, and dollar figures come through wrong more often than common words do. Speaker labeling struggles when people talk over each other or dial in from one conference-room phone. And the summary step, because it is a language model, can quietly compress away a caveat someone said once, or state something more firmly than the speaker did. None of this makes the output useless. It makes it a first draft of the record, the same way these tools produce a first draft of a lease summary or a market write-up. If some of those terms are unfamiliar, the plain-English versions are in our glossary of AI terms every CRE principal should know.
Two Shapes: A Bot That Joins vs. a Built-In Assistant
There are two ways a note-taker gets into your meeting, and the difference matters more than any feature list.
The first shape is a standalone bot, such as Otter, Fireflies, or Fathom, that connects to your calendar and dials into the call as its own named participant. You see it in the attendee list. It works across whatever platform the meeting is on, which is its main advantage: one tool covers your Zoom calls, your Google Meet calls, and your Teams calls. Its main drawback is that it is a separate vendor holding your recordings, and that named participant is visible to everyone on the call, including a client.
The second shape is a built-in assistant already inside the platform you pay for. Microsoft Copilot can take notes inside Teams, Google’s Gemini offers a “take notes for me” function in Meet, and Zoom has its AI Companion. Nothing new joins the call; the summary comes from the same tool that hosts the meeting. For a small firm this is usually the better starting point, because it adds no new vendor, no new subscription, and no unfamiliar name in the attendee list. If your office runs on Microsoft 365 or Google Workspace, you may already own a capable note-taker and not know it.
Because these features change every few months, confirm the current capability and the data terms on the vendor’s own page before you standardize on one. The shape you pick — a broad standalone bot or the assistant already bundled with your platform — sets up every question that follows about who can see it and who holds what it records.
What They’re Genuinely Good At in a CRE Firm
Stripped of the hype, a note-taker does one thing well: it removes the tax of writing things down after a conversation. In a lean firm where the same people sell, manage, and close, that tax is real, and the tool pays for itself in a handful of specific places.
| Meeting | What the note-taker produces | Why it lands |
|---|---|---|
| Internal pipeline review | A recap of which deals moved and the next step on each | Nobody has to reconstruct decisions from memory later |
| Property tour or site visit debrief | A written summary of what the team observed, ready to drop into the deal file | The observations survive past the drive back to the office |
| Vendor or contractor call | An action-item list of who owes what by when | Follow-ups stop falling through the cracks |
| Listing-pitch or team strategy call | Notes a broker can turn into CRM entries and a follow-up email | The CRM gets updated without a broker typing at 9 p.m. |
The through-line is that these are mostly internal conversations, where the content is your own and the only people present work for you. That is the zone where a note-taker is close to pure upside. The value is the same lever that runs through all early AI adoption at a small firm: taking the language-heavy, repetitive parts of the week off a person’s plate so the person spends the hour on judgment instead of transcription. That shift, from a private habit a few people have to a shared firm capability, is the subject of the small-firm AI manifesto.
The Line That Matters: Internal vs. Client Meetings
The one rule worth internalizing is a line, not a ban. Note-takers are low-risk on internal meetings and require judgment on client meetings. Almost every hard question about these tools dissolves once you sort the meeting into one of those two buckets before the call.
| Internal meeting | Client / external meeting | |
|---|---|---|
| Who’s present | Only your team | A client, counterparty, or their broker |
| Content | Your own pipeline and notes | Often NDA-bound deal terms, seller motives, pricing |
| Consent question | You control both sides | You must consider the other party’s consent |
| Default | Record freely | Disclose first, or take manual notes |
For internal calls, the note-taker can run by default. For client calls, such as an acquisition discussion, a lease negotiation, or a listing pitch where the seller shares confidential motivation, the default flips to caution for two separate reasons that people tend to blur together. One is legal: recording a conversation is a regulated act in many states. The other is confidentiality: the content is not yours to hand to a third-party vendor without thought. The next two sections take them one at a time, because the answer to each is different.
Consent Is a Legal Act, Not a Setting
Turning on a note-taker in a client meeting is not a software preference. It is a decision to record a conversation, and US law treats that seriously and inconsistently across state lines. This is the single thing the tool vendors say least about and the thing a principal most needs to hold.
US states divide into two camps. In one-party consent states, one participant in the conversation — you — can consent to recording it, and the recording is lawful. In all-party consent states (sometimes called two-party consent), every participant has to consent. California, Florida, Illinois, and Pennsylvania are among the all-party states, and a client call can easily cross state lines when a broker in one state talks to a seller in another, which pulls the stricter rule into play. A bot that silently joins a client meeting in an all-party context, with no disclosure, is exactly the situation the law is written about.
None of this is legal advice, and the exact rules turn on your state and the specific circumstances. The practical habit that keeps you clear of the whole question is simple: disclose by default. A one-line “I’ve got a note-taker on this call to capture the follow-ups, any objection?” at the top of the meeting does two things at once. It satisfies the consent requirement in every state, because everyone present has now been told and can object. And it removes the second problem, the one no statute covers: a client discovering after the fact that an AI silently transcribed a confidential negotiation. In a relationship business, that discovery costs more than any efficiency the tool bought. When the honest answer is that disclosure would change the conversation, as in a delicate acquisition where the seller would go quiet, the right call is to close the transcript and take notes by hand.
Who Holds the Recording
Consent handles whether you may record. Confidentiality handles what happens to the recording after you do, and it follows the same principle that governs every other AI tool in the firm.
When a standalone bot records a client meeting, a transcript of a confidential conversation now sits on a third-party vendor’s servers. Whether that is acceptable depends on the same two questions you would ask of any AI tool: is the firm on a business-tier account where the vendor does not train its models on your data, and does the content clear your own sensitivity bar. A note-taker recording an internal standup is trivial. A note-taker recording an NDA-bound deal call on a free personal account that trains on the audio is the meeting-room version of pasting a confidential offering memorandum into a consumer chatbot. The exposure is the same, and so is the fix: a firm-controlled business account and a rule about what content is allowed on it. The full version of that setup, the account tier that matters and the sort-every-document rule, is laid out in our guide to using AI without leaking client data.
Built-in assistants change this calculus in the firm’s favor. If your Teams or Google Workspace meetings are already governed by a business agreement you trust with your email and files, the note-taker inside that platform is not introducing a new party to your confidential data — it is the party you already vetted. That is the main reason a small firm should look at the assistant it already owns before adding a standalone recorder: it collapses the “who holds the recording” question into one you have already answered.
Setting It Up Without Overthinking It
This is not a technology project, and it does not need a policy binder. A firm of ten can settle it in one short conversation and one afternoon of setup.
Start with the tool you already own. If you run Microsoft 365 or Google Workspace, test the built-in note-taker on a few internal meetings before you evaluate anything you would pay extra for. Put whatever you choose on a business-tier account, not a personal login, so the no-training commitment covers every recording by default. Write down the one line that does the real work: internal meetings, record freely; client meetings, disclose first or take manual notes. Then say the disclosure sentence out loud once so it feels natural when a client is on the line. That is the whole system. It is worth deciding deliberately rather than letting each person improvise, because the gap between firms that get real value from these tools and firms that get burned by them is almost never the tool, and almost always whether anyone decided how to use it. That gap is the subject of why the AI literacy gap is widening in commercial real estate, and closing it across a small team in ninety days is the whole project of our CRE AI training playbook.
Where to Start
You do not need a meeting-technology strategy. You need to pick the note-taker you likely already own, put it on a business account, and hold one line between internal and client calls. A free AI-readiness assessment is a short working session that looks at your actual mix of internal and client meetings, finds where a note-taker would save your team real time and where it would create exposure, and points you at the right setup for your existing Microsoft or Google environment. Book a free AI-readiness assessment and you will leave knowing exactly which of your meetings an AI should be in the room for — and which ones it should stay out of.
Frequently Asked Questions
How do AI note-takers work?
An AI note-taker runs four steps. It captures the meeting audio by joining the call or listening through your meeting platform. It runs speech-to-text to produce a written transcript. It labels which participant said each line. Then a language model reads the transcript and writes a short summary with action items, which is what most people use instead of the full transcript. Accuracy is strongest on ordinary speech and weakest on names, jargon, dollar figures, and calls where people talk over each other.
Is it legal to record a client meeting with an AI note-taker?
It depends on your state, and this is not legal advice. US states split into one-party consent states, where you alone can consent to recording, and all-party consent states, where everyone on the call must consent. California, Florida, Illinois, and Pennsylvania are among the all-party states, and a call that crosses state lines can pull the stricter rule in. The safe habit that works everywhere is to disclose at the top of the call — a one-line “I have a note-taker on to capture follow-ups, any objection?” — so every participant has been told and can opt out.
Should I use an AI note-taker in a confidential deal negotiation?
Usually only after disclosing it, and sometimes not at all. If a plain “I’ve got a note-taker on this call” would change how the other party speaks — a delicate acquisition where the seller would go quiet — take notes by hand instead. The efficiency a transcript buys is never worth a client discovering afterward that an AI silently recorded a confidential negotiation. For internal deal reviews, where only your team is present, a note-taker is low-risk and worth running by default.
What’s the difference between a bot that joins the call and a built-in note-taker?
A standalone bot, like Otter, Fireflies, or Fathom, connects to your calendar and joins the meeting as a visible named participant, and works across Zoom, Meet, and Teams. A built-in assistant, like Microsoft Copilot in Teams, Gemini in Google Meet, or Zoom AI Companion, lives inside the platform you already pay for and adds no new vendor or attendee. For a small firm, the built-in option is usually the better starting point because it introduces no new party to your confidential data.
Do AI note-takers train on my meeting recordings?
It depends entirely on the account tier. On free or personal plans, some tools may use your data to improve their models. On business-tier accounts, the reputable vendors commit contractually not to train on your inputs, the same way business-tier chatbots do. Because these terms change often, confirm the current language on the vendor’s own business-data page before you standardize. For any firm handling confidential deal data, the business tier is the version to use.
Which AI note-taker should a small CRE firm use?
Start with the one you may already own. If your office runs on Microsoft 365, test Copilot’s note-taking in Teams; if you run Google Workspace, test Gemini’s note-taking in Meet; if you live in Zoom, test its AI Companion. These add no new vendor and are already covered by an agreement you trust with your email and files. Only evaluate a standalone tool like Otter, Fireflies, or Fathom if you need one recorder that spans several meeting platforms, and put it on a business account if you do.
How accurate are AI meeting summaries?
Good enough to be a useful first draft, not good enough to trust unread. Speech-to-text handles ordinary conversation well and stumbles on proper nouns, industry jargon, and numbers, so tenant names, submarket names, and dollar figures are the errors to watch for. The summary step can also compress away a caveat or state something more firmly than the speaker did. Treat the output as a draft you skim and correct, especially before anything from it lands in a client-facing document.
Do I need to tell clients an AI is taking notes?
Yes, and it protects you twice. Legally, disclosure satisfies the consent requirement in every state, including the all-party states where silent recording is unlawful. Relationally, it prevents the worse problem no statute covers: a client learning after the fact that a confidential conversation was transcribed by a tool they never saw. A single sentence at the start of the call handles both. If disclosing would visibly change the conversation, that is your signal to keep the note-taker off and write by hand.
Where do AI note-takers add the most value for a brokerage?
On internal and operational meetings, where the content is your own. Pipeline reviews, property-tour debriefs, vendor calls, and team strategy sessions all produce a clean recap and an action-item list without anyone typing after the call, and several tools can push those notes straight into your CRM. That removes a real after-hours tax in a lean firm. The value is narrower on external client meetings, where consent and confidentiality turn it from an automatic yes into a judgment call.
Dirk Jan van Veen, PhD
