AI for Commercial Real Estate

Commercial real estate runs on documents, relationships, and judgment — three things most generic AI advice gets wrong. You are not flipping condos. You are reading 80-page leases, pulling rent rolls out of CoStar, drafting a Broker Opinion of Value the seller will actually trust, and chasing a tenant rep relationship on LinkedIn for nine months before a deal closes. This guide shows where Claude (Anthropic’s AI assistant, and the model we recommend first for CRE work) saves real hours in your week, plus the tools, prompts, and guardrails that keep you out of trouble.

Where Claude pays for itself in commercial real estate

If you only adopt one AI tool this quarter, make it Claude. CRE work is text-heavy and judgment-heavy — exactly where Claude outperforms. It handles long documents (a full 60-page office lease fits in one prompt), it produces calmer, more professional prose than competing models, and it is willing to say “I’m not sure” instead of inventing a cap rate. That last part matters when your name is on the BOV.

The five tasks where brokers and owner-operators see the fastest payback are: lease abstracting, tenant prospect research, BOV narrative drafting, due diligence document review, and replying to LinkedIn messages and email at 6am before showings start. None of these require you to learn to code. You paste in the document or the data, you ask Claude what you need, you edit the answer. That is the entire workflow.

Here is a prompt to start with this afternoon. Copy it, paste a lease into Claude, and you have your first abstract done before lunch.

You are an experienced commercial real estate paralegal helping me abstract a lease.

I will paste the full lease below. Produce a one-page lease abstract with these sections:
1. Parties (landlord, tenant, guarantor)
2. Premises (address, suite, rentable square feet)
3. Term (commencement, expiration, options to renew)
4. Base rent schedule (with escalations)
5. Operating expenses / CAM / NNN treatment
6. Use clause and exclusivity
7. Assignment and subletting
8. Default and remedies (in plain English)
9. Five things I should flag for my client before they sign

Cite the section number from the lease for every item. If a clause is unusual or unfavorable to my client, say so.

LEASE:
[paste full lease text here]

Run that once and you’ll see what we mean by payback. For more starter prompts, see our best Claude prompts library and the how to use Claude walkthrough.

Lease abstracting and document review at speed

Lease abstracting is the single biggest time sink in commercial brokerage and asset management. A senior associate bills three to four hours per lease. Across a 40-property portfolio that’s a full work week, every renewal cycle. Claude can cut that to twenty minutes per lease — you read the abstract, you spot-check three clauses against the source, you sign off.

The workflow is straightforward. Export the lease as a PDF or Word file. Open Claude. Paste the document into the chat (Claude handles long PDFs natively in the web app and in Claude Pro). Use the prompt from the previous section. Ask follow-up questions like “what is the tenant’s true rent obligation in year five including escalations and CAM” or “summarize every option, contingency, and consent right in this lease” — Claude will pull line-cites for you.

Where it shines is the boring stuff that brokers skip and then regret: estoppel certificates, SNDAs, options to expand or contract, exclusive use carve-outs, percentage rent triggers in retail, holdover rent multipliers, and the difference between modified gross and full service gross. Claude reads every page. You don’t.

For rent roll analysis, paste the rent roll (or a sanitized version with addresses and tenant names removed if you’re worried about confidentiality) and ask Claude to flag tenants approaching renewal in the next 18 months, calculate weighted average lease term, and identify tenant concentration risk. Pair this with property management platforms — Yardi, AppFolio, and Buildium all export rent rolls cleanly to Excel or CSV. Claude reads CSV directly.

One caveat that we’ll cover more below: do not paste live CoStar data into ChatGPT or any consumer chatbot. Your CoStar license has terms about machine extraction. The safer pattern is to copy the fields you need (address, NRA, asking rent) into your own spreadsheet and feed Claude the spreadsheet, not the CoStar export. Same goes for CREXI and Reonomy — read your terms.

The 2026 CRE Broker’s Claude Stack

Commercial real estate is research-and-relationship work at high stakes. The 2026 Claude stack reshapes the research layer dramatically without touching the relationships. See also our AI for Real Estate Investors deep guide, the Claude for Real Estate Leasing, and the full RE cluster.

• Opus 4.7 with 1-million-token context — drop in a full property offering memorandum + comp set + tenant rent roll + financial statements in one conversation. Ask Claude to reconcile what the broker is showing against what the documents actually say.
• Claude Projects per active deal or per market — one Project per active acquisition or per submarket you cover. Every conversation about that deal is grounded in the full document set.
• Claude Skills for BOV + LOI + lease-abstract automation — encode YOUR firm’s BOV narrative voice, your standard LOI structure, your lease-abstract template. Skills mean every junior analyst drafts at the senior level.
• Vision-enabled property analysis — drop interior + exterior + aerial photos. Claude (with vision) flags deferred maintenance, surfaces likely capital-expenditure items, identifies parking-and-access-pattern issues most desktop reviews miss.
• MCP connectors for CoStar, LoopNet, Crexi, RealQuest, ALN — as MCP servers ship for CRE-data platforms, Claude reads live comp data without you bouncing between five dashboards.

Tenant prospecting: from CoStar list to personalized outreach

Tenant prospecting is where most brokers waste the most hours producing the worst output: 200 cold emails that all sound the same, 200 LinkedIn connection requests with no context, 200 voicemails. Claude lets you flip the ratio — fewer touches, far more personalized, much higher reply rate.

Start with your CoStar tenant list, your LoopNet watchlist, or whatever skip-trace tool you use to find decision-makers. Pull a small batch — 25 prospects, not 250 — with name, title, company, current premises, lease expiration if you have it, and a public LinkedIn URL. Drop that into a CSV.

Now ask Claude to do the part you’re bad at. For each prospect, you want a reason-to-call that is real: their company just announced a new product line that needs warehouse space, their HQ lease is expiring in 14 months, they posted on LinkedIn about hiring 30 people in Phoenix, the parent company was acquired and they’re consolidating offices. Claude won’t have all of that — but if you paste the LinkedIn profile text and the last three company press releases, Claude will surface the angle in seconds.

From there, your CRM does the lifting. HubSpot and Salesforce both let you paste in a personalized opener, set a sequence, and track replies. Brokers who only need a lighter setup can run the same play out of a Google Sheet plus Gmail merge. The point is the personalization, not the platform.

Two voice tools earn their keep here. Wispr Flow turns dictation into clean text in any app, so you can talk through 25 prospect notes between showings instead of typing them at 11pm. Otter.ai transcribes your tour calls and listing presentations — feed the transcript to Claude and ask for a follow-up email and a CRM note. The hour you used to spend at the end of the day disappearing.

Broker Opinion of Value: drafting the BOV narrative

The BOV is where new brokers freeze. You have the comps, the cap rate, the NOI, the DCF — the math is in Excel. The hard part is the narrative: explaining to the owner why this strip center should trade at a 6.75 cap and not a 7.25, why the vacant junior anchor box is opportunity not risk, why the local submarket is in a different cycle than the headlines suggest. That is writing. That is what Claude was built for.

The pattern: you do the analysis in Excel. You bring conclusions to Claude. You ask Claude to draft the narrative — market overview, property positioning, valuation methodology, comparable analysis, recommendation. Then you edit. Claude will not invent comps if you don’t give it comps. It will not pull a cap rate out of the air. It will turn the inputs you provide into clean, defensible prose that reads like it took you a day, not three hours.

Specific tasks Claude handles well in a BOV: writing the executive summary in the owner’s tone, translating cap rate compression into language a non-broker family trust can understand, building a “risks and offsets” section that doesn’t sound like boilerplate, and generating a tenant-by-tenant credit narrative for an industrial multi-tenant property where the rent roll has eight names and you need to explain each one.

Claude also handles the visual side, sort of. For graphs and rent comp tables, stay in Excel. For the cover page, the property profile sheet, and the OM-style spreads that go in front of an institutional buyer, Canva with Claude-drafted copy is faster than fighting InDesign. Have Claude write the headline, the deal highlights bullets, and the property description paragraph. Drop it into a Canva CRE template. Done.

If you want a starting structure for your BOV prompt and you’re newer to CRE writing, work through our how to write AI prompts primer first — it’ll save you the four or five iterations most brokers go through before their prompts produce useful output.

10 CRE Plays Most Brokers Don’t Run

1. OM interrogation Skill

Drop the OM. Claude with your underwriting Skill surfaces every assumption that needs verification, every claim that contradicts the T-12, every comp that’s outside the realistic range. The diligence layer that historically required senior-analyst hours.

2. Lease-abstract automation at scale

The worst job in commercial. Drop a 60-page lease. Claude produces a 1-page abstract: term, options, base rent, escalations, expense responsibilities, exclusive-use, assignment, default triggers. Cuts analyst-time per deal by 80%.

3. BOV narrative generator

Broker Opinion of Value drafts in your firm’s voice. Claude with the comp set + the subject property profile + your firm’s standard BOV format produces the defensible narrative in 30 minutes vs. 4 hours.

4. Tenant prospecting from CoStar

Most tenant outreach is generic. Claude with the prospect’s public business profile + recent expansion announcements + their current lease term (if discoverable) drafts a personalized first-touch email demonstrating you understand their specific situation. Response rates 5-10x cold outreach.

5. The Voss Never Split the Difference framework for landlord/tenant negotiations

The TI battle. The free-rent dance. The exclusive-use clause. Chris Voss’s Never Split the Difference framework, encoded as a Skill, gives you the calibrated questions and tactical empathy moves for the high-stakes commercial moments.

6. Capital-stack waterfall modeling

For sponsor + LP deals: drop the proposed senior debt + mezzanine + preferred equity + common equity. Claude builds the waterfall and tells you whether the projected IRRs hold up under realistic assumptions or rely on aggressive exit caps.

7. Tenant-credit analysis Skill

For NNN and credit-tenant deals: Claude with the tenant’s public financials builds a defensible credit-quality narrative + flags any red flags that change the underwriting.

8. Comparable-sale + comparable-lease normalization

Drop your comp set. Claude normalizes each comp for the adjustments (base year, escalations, TI, free rent) and surfaces the three closest comps with the math shown. Defensibility that survives the LP underwriting committee.

9. Entitlement + permitting pre-check

For value-add and ground-up: drop the parcel data + the zoning code + recent permit-approval data. Claude produces a defensible “here’s the realistic permitting path and timeline” briefing before you offer.

10. Year-end broker portfolio + book-of-business review

Drop your year’s deal flow + closed deals + dead deals + time-per-deal. Claude surfaces which submarkets are productive, which deal types are unprofitable for you, which client relationships are worth deepening.

For broader framing on where the regulatory environment is heading for commercial finance, this newsletter recently covered the Bank of England joining regulators concerned about AI in financial services — useful framing for any CRE broker thinking about commercial-property exposure to fintech/AI tenants and the regulatory pressure on their underwriting.

Three Claude prompts every CRE broker should save

These three prompts cover the bulk of what a working broker does in a normal week. Save them in a notes app or in Claude Projects so you’re not rewriting them every Monday.

PROMPT 1 — Personalized tenant prospect outreach from CoStar data

You are my CRE business development assistant. I will paste a CSV of 25 tenant prospects below with these columns: Contact Name, Title, Company, Current Address, Current SF, Lease Expiration, LinkedIn URL, Public News Snippet.

For each prospect, write a 90-word LinkedIn message and a separate 110-word email. Both must:
- Open with a specific, real reason I am reaching out (use the news snippet or the lease expiration — never both in the same message)
- Reference one tangible market fact about their submarket
- End with a soft ask: a 15-minute call, not a tour
- Sound like a human broker who has been in the market 10 years, not like a template

My signature: [your name, brokerage, market, phone].

CSV:
[paste here]

PROMPT 2 — BOV narrative for an industrial property the owner is considering selling

You are drafting the narrative sections of a Broker Opinion of Value for a [SIZE] SF [SHALLOW BAY / BULK / FLEX] industrial property in [SUBMARKET, METRO]. The owner is considering a sale in the next 6-12 months.

Inputs I am providing:
- In-place NOI: [$]
- Stabilized NOI: [$]
- Concluded value range: [$ low - $ high]
- Concluded cap rate range: [low% - high%]
- Three sale comps (address, date, price, cap, PSF, brief story)
- Tenant roster with credit notes
- Two market data points I trust (vacancy, asking rent trend)

Write four sections, ~250 words each: (1) Executive Summary, (2) Property Positioning, (3) Valuation Methodology and Comp Analysis, (4) Recommendation. Tone: confident, plain-English, no jargon the owner won't recognize. No invented numbers.

PROMPT 3 — Respond to a 1-star Google review claiming you ghosted a prospective tenant

You are helping me draft a public reply to a Google Business Profile review. The reviewer is a prospective tenant who claims I never returned their calls about a retail space last fall. I did follow up — I have the email thread — but I am not going to argue facts in public.

Write a 90-word reply that:
- Opens by thanking them for the feedback
- Acknowledges that any prospect who feels unheard is a real problem regardless of what happened
- States, neutrally, that I do have records of follow-up and would welcome a direct conversation to make it right
- Provides my email and direct line
- Does NOT name the reviewer, does NOT call them wrong, does NOT mention the property
- Sounds like a confident broker, not a defensive one

Drop these into the same conversation thread when you’re done with the inputs and Claude will keep your tone consistent across all three. Update your Google Business Profile reply within 48 hours of any review — search rankings reward responsiveness.

What AI shouldn’t do in CRE

Three hard lines, and they are non-negotiable. First, AI should not draft binding LOIs, leases, purchase and sale agreements, or 1031 exchange documentation that goes to counterparty signature. Use Claude to summarize, redline, and explain — never to produce the final document. Your attorney signs off. Always.

Second, do not paste licensed CoStar, CREXI, Reonomy, or MLS data wholesale into ChatGPT, Claude, or any consumer AI. Read your data license. Most CRE platforms restrict machine extraction and bulk reuse. The safe pattern is to extract specific facts you’ve earned the right to use into your own spreadsheet, then work from that.

Third, AI does not replace site visits, phone calls, or relationships. CRE deals close on trust built over coffees and tours, not on email sequences. Claude makes you faster at the desk so you can spend more hours in the market — that’s the win. For more on the broader playbook, see our guides on AI for real estate and AI for small business, browse our full AI tools directory, and join the weekly Beginners in AI newsletter for plays we test before they hit the blog.

You May Also Like

• AI for stock trading: what works and what is hype
• AI for YouTube Creators: Thumbnails, Scripts, and Editing
• AI for Real Estate Investors: Deal Analysis & Portfolio Management
• AI for Property Managers: Tenants, Maintenance & Leasing
• AI for Engineering Students: Problem Sets, Lab Reports & Design
• How to Use AI: A Step-by-Step Guide