The deal is the deal. The slow part is the offering memorandum: turning a rent roll, a few photos, and the asset facts into a clean, persuasive document an investor will actually read. That marketing prose is exactly where AI is fast, and it is good at it. The trap is that the same tool will also invent the numbers. Ask it to draft the OM and it will produce a confident NOI, a cap rate, a rent comp, a market growth statistic, and a five-year projection out of thin air, and you are the one who hands that document to investors with your name on it. The prose AI can draft. The numbers are yours, sourced and verified. This is a plain look at where AI genuinely helps with offering memorandums, where it does not, and a set of prompts you can paste in and test on your own real deals.
The honest picture
AI is genuinely useful for the writing, turning your verified facts into a document:
- What AI does well today: take the asset facts, the verified financials you supply, and the property details and write a clear, consistent OM narrative: the property description and investment highlights, the location and market section from data you paste, and a readable summary of the rent roll. It compresses the drafting hours, not the underwriting.
- What AI does not do: source, verify, or vouch for a single number. It does not know the real NOI, the actual cap rate, the true rent roll, or the comps, and any figure it produces on its own is a guess that ends up in a document investors rely on. It cannot pull market statistics, find comparable sales, or build a projection from anything but the inputs you give it. The financials, the representations, and every number in the OM are yours. AI drafts the prose around figures you have already verified.
The right way to think about it: AI is a fast writer working from your verified numbers, not an analyst and not a data source. The figures and the representations are yours. The drafting is what you hand off.
The line: it will fabricate the numbers you have to stand behind
The useful job and the document-killing mistake sit right next to each other, and the failure shows up in a few specific ways:
- Inventing the financials. Ask it to "write the financial summary" with a thin prompt and it will produce an NOI, a cap rate, an expense ratio, and rent roll figures that look right and came from nowhere. In an OM, a plausible-looking number you did not verify is a representation you cannot back up.
- Generating market stats and comps. It is tempting to ask "what is the rent growth in this submarket" or "give me three comparable sales." A model answering that is inventing data, often stale or fabricated, presented as fact in a document an investor underwrites against. AI cannot source market data. It can only arrange the data you paste.
- Fabricating projections. Ask for a "five-year proforma" and it will generate a tidy growth curve with no basis in your actual assumptions. The projection is your underwriting, not the model's invention.
The fix is the same in every case: AI uses only the figures you provide, invents no statistic, comp, or projection, and leaves any number you did not give as a blank for you to fill rather than guessing it.
The setup that keeps the numbers yours
Two habits make AI safe for an OM, and the prompts below build them in:
- Give it your verified figures, and only those. Hand it the financials you have already confirmed, the rent roll, and the market data you sourced yourself, and tell it to use those verbatim. Never ask it to estimate a cap rate, find a comp, or pull a market statistic. A model handed your verified numbers arranges them; a model asked to source data fabricates it.
- Make it blank, not guess. Tell it that if it needs a figure you did not provide, it leaves a clearly marked blank for you to fill and never invents the number. You want a draft that surfaces every gap, not one that papers over it with confident fiction.
How to test it on your own work
Do not trust a polished demo, including this one. Pull two or three of your own recent deals, the verified financials and the asset facts, a timer, and the prompts below. Rate each output 1 to 5 on usefulness and accuracy, and check every number against your source. Compare the time against how you draft an OM today, and keep what wins. Use your firm's enterprise or no-train settings before you paste anything you consider sensitive, and confirm you are comfortable with that.
Paste-ready prompts
Copy these as written. Bracketed text is what you swap per deal.
Test 1: Draft the property description and highlights (text model)
I am giving you the verified facts for one commercial property. Write a clear,
investor-facing property description and investment highlights section from only
what I provide.
Rules:
- Use only the facts I paste. Do not add a figure, a tenant, an amenity, a
condition, or a selling point I did not give you. If something is unclear,
leave it as "[confirm]" rather than assuming it.
- Invent no statistic, no comp, and no number. Leave any number I did not give
as a blank for me to fill, written as "[figure to fill]".
- Organize it as: property overview, investment highlights, and a short
positioning paragraph, in plain professional prose.
Verified property facts: [paste them]
Watch for: did it stick to your facts and leave every number to you, or did it add a highlight or slip in a figure you never gave it?
Test 2: Draft the location and market section (text model)
Write the location and market section of an offering memorandum using ONLY the
data I paste below. Rules:
- Use only the figures and facts I provide. Invent no market statistic, no
growth rate, no demographic number, and no comparable sale. If a figure would
normally go here and I did not give it to you, leave a blank marked
"[market data to source]" instead of producing a number.
- Do not characterize the market as "strong," "growing," or any judgment my data
does not directly support.
- Write it as clear, readable prose an investor will skim.
Verified market data and location facts: [paste them]
Watch for: did it write the section from only your data, or did it produce a growth statistic, a comp, or a "thriving submarket" claim you never supplied?
Test 3: Turn the rent roll into clean summary prose (text model)
Here is my verified rent roll and operating financials: [paste them]. Write a
clear summary of the income and the tenancy in prose an investor can read.
Rules:
- Copy every figure exactly as I gave it. Do not round, recalculate, sum, or
adjust any number. If a calculated figure is needed (for example a total or a
ratio) and I did not provide it, do not compute it: mark it "[figure to fill]"
for me to add.
- Flag any gap: list every place the rent roll is missing a value, a lease term,
or an expiration so I can fill it before this goes out.
- Add no tenant, no income, and no expense that is not in what I pasted.
Watch for: did it copy your figures verbatim and flag the gaps, or did it quietly compute a total or invent a missing lease term?
Test 4: Fabrication audit (text model)
You are auditing an offering memorandum draft against its source. I will paste
(A) my verified source figures and facts, and (B) the OM draft. Find every
number, statistic, comparable, projection, or factual claim in the draft that is
NOT present in the source.
For each one, quote the exact line from the draft and say what is unsupported.
Specifically catch:
1. Any dollar figure, cap rate, NOI, rent, or ratio not in the source.
2. Any market statistic, growth rate, or demographic number not in the source.
3. Any comparable sale or comp not in the source.
4. Any projection or forward-looking figure not in the source.
Do not rewrite the draft. Only list what cannot be traced to the source.
(A) Source figures and facts: [paste them]
(B) OM draft: [paste it]
Watch for: does it catch the cap rate, the comp, or the projection that crept in with no source behind it? Run this on every draft before it leaves your desk.
What success looks like, and where it could go
If your own testing shows real time savings, the next step is a small pilot: draft your next few OMs with the prompts and measure the hours. If that holds up, the natural next step is a simple agent, running on your brokerage's own cloud, that you use in plain language. The most useful version takes the verified financials and asset facts you supply, drafts the OM narrative around them, and cites the exact source figure behind every number it uses, leaving a clearly marked blank for anything you did not provide rather than inventing it. The deals and the financials stay in software your brokerage owns and runs, not rented per seat. It drafts and arranges; you source, verify, and stand behind every figure.
The principle holds the whole way through: AI gives you a faster, cleaner OM draft and a second set of eyes on what cannot be traced to a source. It does not pull the market data, and it does not certify the numbers. Keep that line clear and the rest is upside.
This is not investment, securities, or compliance advice. You are responsible for the accuracy of every figure and representation in an offering memorandum and for meeting the disclosure rules that apply to your deal.
Want a straight answer for your brokerage?
I build practical AI and custom software for businesses, on Google Cloud. If you want a second set of eyes on how AI could fit your deal-marketing workflow, or on a tool you are considering buying, tell me what you are working with. No pitch, just a straight answer.