AI workflows for a commercial real estate brokerage

The pitch decks all say the same thing. “AI will transform commercial real estate.” Then the page turns and there is a screenshot of someone asking ChatGPT to write a property description for a strip mall. I want to talk about the parts that actually move a brokerage forward, written from inside one. None of this is hypothetical. These are the workflows I am wiring up for principals this quarter.

The tour memo, three minutes after you pull out of the lot

This is the easiest dollar in the business and almost nobody captures it. The acquisitions broker spends two hours touring an asset with a seller broker. They get back in the car. They have eleven minutes before the next call. The tour notes get scribbled into Notes, sometimes Voice Memo, sometimes nothing.

Claude turns those eleven minutes into a real artifact. A Slack channel or an iMessage thread with Claude on the other end. Broker dictates for two minutes. “Class B office, 1986 vintage, 102K square feet, two stories, full asphalt repave needed, HVAC is mostly Carrier 2008 rooftop units, the lobby renovation looks like a 2019 job and they still have not leased the second floor, tenant on the first is paying 19 a foot triple net but the broker says they have a renewal option at 17, asset manager confirmed seller wants to close before year end, the seller broker mentioned the listing is also being shopped to one other group quietly.”

Claude returns a structured tour memo. Sections for physical condition, lease economics, deal dynamics, action items. It pulls from the firm’s prior memos so it matches the house style. It flags the things that need a callback (“you mentioned the seller is talking to one other group, who, and is this exclusive or open”). It generates the followup email back to the seller broker requesting the rent roll and operating statements before the broker is out of the parking lot.

The broker does not have to remember to write it up that night. It is already done.

Comps and OBV in fifteen minutes instead of three hours

Broker opinions of value are the single most leveraged tool a brokerage produces. They are also where 70% of associate hours go to waste. The pattern: pull comps from CoStar, scrub them, build the model, write the cover memo, redo the cover memo because the principal wants a different cap rate range, redo the model because three of the comps are not actually comparable.

A Claude workflow does not replace the model. It eats the parts around it. Pull comps from CoStar via API. Filter for true comparability (vintage range, size range, asset class, distance, sale within 24 months). Build the comp set table. Write the cover narrative with the three points the principal always makes. Generate the page that explains why two excluded comps were excluded.

What used to be three associate hours becomes fifteen minutes of associate review. The principal still owns the cap rate range and the closing argument. The grunt work is gone.

The LOI back and forth, finally tracked

This is the workflow that surprised me the most when I started building it. Every brokerage thinks of the LOI as a single document. It is not. It is a string of nine emails over eleven days where each side moves on rent, term, escalations, options, free rent, TI, and broker commission, usually two terms at a time. By the end nobody is sure what the current standing terms are without scrolling.

Claude reads the entire thread and produces a running terms summary after every email. Each side’s current position, every term, what changed in the latest round, and what the spread is. The broker sees it in their inbox the moment the latest email lands. They do not need to scroll. They do not need to open the prior draft.

The second-order effect is that the broker brings a much better recommendation to the principal. “Tenant is at 32, we are at 36, they have moved twice in the last three rounds, the gap is 12% but the gap two rounds ago was 18%. Recommend countering at 34 with two free months added back.” That used to require half an hour of email rereading. Now it is a chat message.

Investor and landlord updates that the principal does not have to write

The recurring monthly update is the workflow nobody wants to admit eats their Sunday evening. Big landlords want a status memo on every active listing. Investor clients want pipeline updates. The principal has the context but not the time. The associate has the time but not the context.

Claude reads from the firm CRM, the pipeline tracker, and the active email threads, and produces a draft update in the principal’s voice for every account. The principal edits for fifteen minutes per update, signs, sends. The structural work is automated. The judgment work stays with the principal.

The downstream effect is that the updates actually go out monthly. Today most shops produce them quarterly because the time cost is too high, and quality suffers. Pulling that cadence to weekly or monthly without burning a Sunday is a real revenue lever.

Where Claude does not belong yet at a brokerage

I am skeptical of two use cases that get pitched a lot:

Cold outbound at scale. Every brokerage wants to multiply their canvassing. The reality is that the personalization signal that makes outbound work is exactly what scales worst. Sending 200 mediocre AI-personalized notes a week to landlords does not get you on a listing. It gets you ignored. The right answer is to use Claude to write the three best personalized notes a day, not 50 worse ones.

Buyer matching. “Claude will match buyers to listings.” This is a CRM problem dressed up as an AI problem. You do not need a language model to know that the family office with a $40M industrial mandate wants to see the $38M industrial deal. You need a CRM that does not lose the mandate. Fix that first.

The order of operations

If you run a brokerage with under 30 brokers, the order I would tackle this is:

First, the tour memo workflow. It is the highest ratio of value-to-effort and you get instant adoption because the broker is already in the habit of voice notes.

Second, the LOI tracker. This one creates a moment in every active deal where the broker brings a measurably sharper recommendation to the principal. That changes how the principal sees the broker.

Third, the OBV grunt work. Pull this out of associate hours and you reclaim 5 to 10 hours per associate per week. That is more billable time and more deals worked.

Fourth, the recurring updates. This one is the easiest sell to the principal but the hardest to get right, because the principal’s voice is doing 40% of the work. Save it for last and do it carefully.

The whole thing kicks off in about 30 days and gets sharper every quarter on a written-SOP cycle. None of it requires a brokerage to swap its tech stack. It plugs into what is already on the desk.

If you want to walk through how this would look at your shop, I do a 60-minute working session that pulls a real deal of yours and shows the workflow end to end. Calendar link is at the bottom of the page.