Justin McKelvey
Fractional CTO · 15 years, 50+ products shipped
AI for Commercial Real Estate: Lease Abstraction, OMs and Comps Memos, With a Worked Example (2026)
The short answer
How is AI used in commercial real estate? As of 2026, mostly for the reading and the first draft: abstracting leases into a one-page summary, screening offering memorandums against your buy box, drafting comps and broker opinion of value memos from numbers you supply, and drafting tenant-rep follow-ups. A general assistant like ChatGPT Business or Claude Team ($20 a seat billed annually) handles all four. What AI doesn't do is price the building or sign off on a lease. A person still owns both.
This page is for the 5 to 50 person shop: a brokerage principal, a small owner-operator, an asset or property manager. If you're on the residential side, the AI for real estate hub is the better starting point. Here it's leases, OMs and memos, plus a worked example on a 40-page office lease, because "AI will transform CRE" isn't something you can do on a Tuesday.
What jobs does AI actually do well in commercial real estate?
Four, and they share a pattern. Each one is a long document going in and a short structured thing coming out, with a person checking the short thing.
- Lease abstraction. A 40-page lease becomes a one-page abstract: parties, premises, term, rent schedule, escalations, CAM, options, assignment, default. This is the biggest time sink and the biggest win.
- Deal screening. Paste in an offering memorandum and your buy box (asset type, size, market, minimum cap rate, tenancy rules) and get back a pass, a maybe, or a no with the page numbers behind it. It saves the first read, not the underwriting.
- Comps and BOV memo drafting. You supply the comps and the rent roll; the assistant drafts the memo in your format. It should never supply a comp, a cap rate or a value it found on its own. Why that matters is covered in can AI value a property.
- Tenant-rep follow-up. Post-tour recaps, lease expiration nudges 18 months out, and the "here are three spaces that fit" email. Short, repetitive, and easy to get wrong when you send 40 of them by hand.
Building management is its own world (work orders, owner reports, tenant messages), and AI for property management covers it.
How does AI lease abstraction work on a real lease?
Here's the worked example. Take a 40-page office lease: 5,000 rentable square feet, five-year term, a base rent schedule, a 3% annual bump, a base-year operating expense structure, one five-year renewal option, an anchor-tenant co-tenancy clause, and an assignment section that runs a full page. Two ways to abstract it.
Option A: a general assistant with a template
ChatGPT Business or Claude Team, a fixed abstract template saved in a shared project, and three rules in the instructions:
- Fill every field from the lease only. If it's not in the lease, write "not found".
- Quote the section number and the exact sentence for every field.
- List every place where one section modifies another (an amendment, an exhibit, a "notwithstanding").
Rule 2 is the one that matters. An abstract with citations takes minutes to check. An abstract without them takes as long as reading the lease, which defeats the point.
Option B: a dedicated lease abstraction tool
Purpose-built CRE tools extract a standard field set and push it into a lease administration system or a portfolio database. The win is consistency across hundreds of leases and data you can query ("every lease with a co-tenancy clause expiring in 2027"). The cost is a vendor contract, and the price is usually behind a quote form. Occupier, for example, lists AI lease abstraction among its products, and its pricing page (read October 7, 2026) describes "Simple per-lease, per-module pricing with no hidden fees, no per-seat charges" but prints no number, just a "get pricing quote" button. Ask for the per-lease quote in writing and ask what year two costs.
The five clauses no model signs off on
Whichever option you pick, a person reads these five against the lease, every time. They're where the money is, and they're where one section quietly rewrites another.
| Clause | What the abstract can get wrong | What the human checks |
|---|---|---|
| Rent escalations | Treats a 3% bump as compounding when it's fixed-dollar, misses a CPI floor or cap, or misses free rent that shifts the schedule | Rebuild the full rent schedule by hand and compare it month by month |
| CAM reconciliation | Misses the base year, the expense exclusions, the cap on controllable expenses, or the gross-up | The base year, the exclusion list, any cap, and the audit right with its deadline |
| Co-tenancy | Summarizes the trigger but drops the remedy (rent reduction, then a termination right after a cure period) | The named tenants, the trigger, the remedy, and the clock |
| Assignment and subletting | Writes "landlord consent required" and drops the permitted transfers, the recapture right, or the profit split | Consent standard, permitted transfers, recapture, and who keeps the sublease profit |
| Renewal options | Gets the option right and the notice window wrong, or misses that the option is personal to the original tenant | The notice window (put it in the calendar), how the renewal rent is set, and any conditions |
One legal note, said plainly: an AI lease abstract isn't legal review. It's a summary to speed up your reading. It isn't legal advice, it can miss a clause, and anything that drives a decision goes back to the lease and to your attorney.
What does AI lease abstraction save a small CRE firm?
Here's the math with numbers you should replace with your own. The hours and the hourly rate below are assumptions, not measurements. Only the seat prices are published: Claude Team read on Anthropic's pricing page October 7, 2026 (Standard seat, teams of 2 to 150), ChatGPT Business on OpenAI's October 6, 2026.
| Line | Assumption or price | Per lease or per month |
|---|---|---|
| Manual abstract | Assume 4 hours at a $60 loaded hourly cost | $240 per lease |
| AI draft + human check of the five clauses | Assume 1.5 hours at $60 | $90 per lease |
| Saved | The difference | $150 per lease |
| 5 seats, ChatGPT Business or Claude Team | $20 per seat billed annually ($25 monthly) | $100 a month ($125 monthly), $1,200 a year |
On those assumptions, one lease a month covers five seats, and three leases a month saves about $450 a month against $100 of seats. If your manual abstract takes 2 hours, not 4, the saving per lease falls to $30 and it takes four leases a month to cover the seats. Run it with your own times before you believe anyone's ROI slide, including mine.
And the seats do the other three jobs too, so the lease math is the floor, not the ceiling.
Which AI tools should a commercial real estate firm start with?
One general assistant on a business plan, for everyone, before anything else. Pick ChatGPT Business or Claude Team; both list $20 a seat billed annually, and the business plans keep your documents out of personal accounts, which matters when an OM arrives under an NDA. The residential AI tools for real estate agents table has both vendors' plans side by side if you want the detail.
Add a dedicated lease tool only when one of these is true: you abstract leases every week, you need the data in a lease administration system, or you're standardizing fields across a portfolio. Before that, a template and a checker beat a contract.
Where should a CRE team be careful with AI?
- Confidential documents. OMs under NDA, rent rolls and tenant financials go into a business workspace with admin controls, never a personal account. Check your NDA's terms before any of it goes into any AI tool.
- Numbers it invents. An assistant asked for "market rent" will give you one. Every number in a memo comes from your comps or your data room, with the source next to it.
- Values. A drafted BOV is a draft of your opinion. The value is yours, and so is the liability.
Who can help a commercial real estate firm roll out AI?
I can. I'm Justin McKelvey, an AI consultant in Austin, Texas, and I help teams pick one assistant, set the rules on what goes into it, and build the first workflow, like the lease abstract template above. Start with the free guide, 6 AI tactics for real estate, which covers residential and commercial operators and lands in your inbox. If you want a written plan for your firm after that, the $2,500 AI Readiness Assessment is the paid path: two weeks and a roadmap you can act on.
What to do this week
Take one lease you already abstracted by hand. Run it through a business-plan assistant with the three rules above, then compare the five clauses line by line. Where it matches, you've found time. Where it doesn't, you've found your checklist.
Sources
- OpenAI, ChatGPT Business pricing, openai.com/business/pricing (read October 6, 2026)
- Anthropic, Claude pricing, claude.com/pricing (read October 7, 2026)
- Occupier, pricing plans, occupier.com/pricing (read October 7, 2026)
- DataForSEO, US search volume and keyword difficulty for "ai for commercial real estate", "ai in commercial real estate" and "ai lease abstraction" (read October 7, 2026)
Next step Get the 6 tactics →
6 AI tactics for real estate
Listing copy, lead follow-up, document parsing and three more plays for agents, teams and brokerages. Free.
Frequently Asked Questions
- How is AI used in commercial real estate?
- As of 2026, mostly for document work: abstracting leases into a one-page summary, screening offering memorandums against a buy box, drafting comps and broker opinion of value memos from data you supply, and drafting tenant-rep follow-ups. It reads and drafts fast. It doesn't price buildings, sign off on lease terms or replace a lawyer.
- Can ChatGPT or Claude abstract a commercial lease?
- They can produce a solid first draft of a lease abstract if you give them a fixed template and make them quote the section number for every field. Where they slip is in the clauses that depend on other parts of the lease: rent escalations, CAM reconciliation, co-tenancy, assignment and renewal options. A person checks those five against the lease every time.
- Should a small CRE firm use a dedicated lease abstraction tool or a general AI assistant?
- If you abstract a handful of leases a month, a business plan for ChatGPT or Claude with a good template is usually enough. A dedicated lease abstraction tool earns its price when volume is steady, when you need the data in a lease administration system, or when a portfolio needs the same fields across hundreds of leases. Get any dedicated vendor's per-lease or per-portfolio quote in writing.
- What does AI cost for a 5-person commercial real estate team?
- At published prices (Claude Team read October 7, 2026, ChatGPT Business read October 6, 2026), both list $20 per seat per month billed annually, or $25 monthly. Five seats is $100 a month ($1,200 a year) on annual billing or $125 a month billed monthly.
- Is an AI lease abstract a substitute for legal review?
- No. An AI abstract is a summary to speed up your own reading. It isn't legal advice, it can miss or misread a clause, and anything that affects a decision (a renewal, an assignment, a dispute over CAM) goes to the lease itself and to your attorney.
More on AI for Business
AI for Universities (2026): How to Roll It Out to Staff, With ChatGPT Edu, Claude, Gemini and Copilot Side by Side
How a university or college should roll AI out to its administrative staff: one approved tool, a written data line, and three offices first. Plus the four vendor education programs compared from their own pages, and what a 150-person staff rollout costs on each.
Claude for Google Workspace (2026): Docs, Sheets and Slides Add-On, Pricing, and Whether You Still Need It With Gemini
Claude now runs in a sidebar inside Google Docs, Sheets and Slides, in public beta on every paid Claude plan as of October 6, 2026. What it does in each app, what it costs a 10-person team on top of Workspace, when Gemini is enough, and how to roll it out without regretting it.
ChatGPT Business vs Plus (2026): Which One a Small Business Should Actually Pay For
ChatGPT Plus vs Business, as of September 2026: Plus is $20 a month for one person and OpenAI may train on your chats unless you opt out; Business is $20 a seat billed annually or $25 monthly, two seats minimum, with no training by default, admin controls, shared projects for groups, and the Company Knowledge plugin. The side-by-side from OpenAI's pages, and three owner scenarios with a verdict each.
ChatGPT for Customer Service (2026): What Works, What Breaks, and What It Costs a Small Business
ChatGPT for customer service, as of September 2026: it drafts replies, answers policy questions from your own documents, and triages an inbox for $20 to $25 a seat a month on ChatGPT Business. It can't look up an order, issue a refund, or remember a customer unless you connect the system that holds them. The honest capability line, the plan you need, a setup in six steps, the failure modes, and the point where a real agent takes over.
Written by
Justin McKelvey
Fractional CTO & AI consultant in Austin, TX. 15 years building software, 50+ products shipped, $53M+ in client revenue generated. I help $1M–$50M founders ship production software and automate operations with AI, without hiring a full-time executive team.
Work with me