Disclosure
Pacific Software Ventures (PSV) builds AI systems that read and draft offering memoranda for brokers and buyers, so yes, we have a stake in what you pick. The steps, checks and safety rules below work with any tool, ours included, and each result links to its case study.
What does AI do with an offering memorandum?
AI offering memorandum analysis means an AI reads the OM the way a careful analyst would. On the buy side, it pulls the deal into your screening sheet, checks every claim against the rent roll and T-12, splits in-place income from the broker's pro forma, flags the assumptions doing the heavy lifting, and cites a page for every number. On the sell side, it drafts the property, tenant and financial sections from the seller's data room, in the broker's template. The broker still writes the thesis and the pricing. The analyst still decides every adjustment. And your own model does the math.
How does AI review an offering memorandum on the buy side?
Every OM reads like the best deal of the year. That's its job. It's accurate about some things and generous about others. So figure out which before anyone burns a week underwriting it. Here's the order we'd run it in.
- Extract the deal into your screening sheet. The AI reads the whole OM, scanned pages too, and fills in your screening sheet: price or guidance, property type, location, size, vintage, occupancy, in-place and pro forma NOI, the stated cap rate, tenancy, assumable debt and the call for offers date. Every field cites its OM page. If the OM doesn't say it, the field stays blank and flagged. No guessing.
- Check the OM against the rent roll. Occupied units or suites, in-place rent, expirations and concessions should match the rent roll line by line. When they don't, look for these gaps: a signed lease that hasn't started counted as occupied, asking rents sitting where in-place rents belong, or a big tenant's termination right missing from the tenant summary. Each gap comes back with both page references. AI rent roll analysis
- Check the OM against the T-12. It maps the seller's trailing twelve into your categories and lists every place the OM's in-place NOI drifts from it: a missing expense line, a management fee below market, taxes or insurance held at an old bill, a one-time item left in, or other income annualized off one good month. Your analyst decides every adjustment. AI T-12 analysis
- Separate in-place income from the broker's pro forma. OMs mix trailing, in-place, first-year and stabilized NOI. The AI labels each one, notes which one the price is quoted on and lists the steps from in-place to pro forma. The bridge itself is computed in your model or a calculator, not typed by the AI. Cap rate calculator
- Flag the broker's assumptions. A pro forma rests on a few assumptions that carry the value: rent growth, mark-to-market on rollover, lease-up pace, renovation premiums, expenses growing slower than income, taxes held flat after sale, and the exit cap. The AI lists each with the OM's wording and page, and marks the ones with nothing behind them.
- Screen against your box. It runs the deal against your written criteria (markets, asset types, size, vintage, business plan, return bar) and returns pursue, pass or needs information, with the rule behind the call. AI for CRE acquisitions
- Write the one-page screen and the broker questions. One page. The deal in a paragraph, the screening sheet, the gaps between the OM and its backup, the assumptions to test, and broker questions (the missing rent roll month, the termination option, where the renovation premium came from). Then an analyst decides whether it goes to full underwriting.
| Field | Where the OM states it | What the AI checks it against |
|---|---|---|
| Price or guidance | Cover, executive summary, call for offers | Stated cap rate recomputed in your model from the in-place NOI |
| Occupancy | Executive summary, property overview | Occupied units or suites on the rent roll, excluding leases not yet started, and economic occupancy from T-12 collections |
| In-place rent | Rent roll summary, unit mix table | Rent roll line items and the T-12 rental income |
| In-place NOI | Financial summary | The T-12, normalized to your categories |
| Expense recoveries | Financial summary, tenant summary | Each lease's recovery clause and the latest CAM reconciliations |
| Pro forma NOI | Financial summary, pro forma page | Each assumption in the bridge, listed with its page |
| Tenancy and rollover | Tenant summary, lease abstracts | The leases themselves: options, termination rights, co-tenancy, and the weighted lease term by year and tenant |
| Capital work | Property overview, value-add section | Seller's capex history and the property condition report, when provided |
| Assumable debt | Financing section | The loan documents: rate, maturity, prepayment and assumption terms |
Which OM claims should you verify before trusting them?
These are the lines that move value, and it's the broker's job to show them in the best light. A good AI review puts each one next to the document that settles it.
- Why the table matters. The OM claims that cost buyers money are seldom false. They're true statements about a different number than the one you'll underwrite: stabilized occupancy quoted as current, an asking rent quoted as achieved, a cap rate on next year's income. What is an offering memorandum
| The OM claims | Verify against | Flag when |
|---|---|---|
| The cap rate | In-place NOI from the normalized T-12 divided by the price, in your model | A cap rate quoted on first-year or stabilized NOI without saying so |
| Occupancy | The rent roll as of its date | Signed but not commenced leases, model units or employee units counted as occupied |
| Economic occupancy | T-12 collections against gross potential rent | Concessions, bad debt or non-paying tenants counted as occupied |
| Rents below market | The OM's own rent comps and your recent leasing data | Comps from newer buildings, better locations or old dates |
| Loss-to-lease upside | Rents achieved on recent new leases and renewals | A mark-to-market that assumes every renewal reaches asking rent |
| Expense recoveries | Lease recovery clauses and the CAM reconciliations | A gross-up, recovery cap or exclusion the OM leaves out of its recoveries |
| Weighted lease term | Expirations on the rent roll, by year and by tenant | Rollover concentrated in one year or one tenant behind a healthy average |
| Expense ratio | T-12 expense lines by category | Missing reserves, a management fee below your floor, taxes not reassessed on sale |
| Value-add upside | Renovation scope, cost and downtime | A rent premium with no budget or no vacancy during the work |
| Lease-up timing | Absorption history in the submarket and at the property | A stabilization date the property's own history does not support |
| Credit tenants | The lease: guarantor, term remaining, options, termination rights | A parent company named in the OM that is not on the lease |
| Recent capital work | Seller's capex ledger and invoices, when the data room has them | A new roof or system named with no date or cost behind it |
How do brokers use AI to draft an offering memorandum?
On the sell side, the work runs backwards. The broker has a data room, a deadline, and an OM where every page has to agree with the rent roll, the T-12 and the leases. AI drafts the sections that come straight from documents, then checks them against each other.
- Read the data room. The AI reads what the seller sends (rent roll, operating history, leases, condition report, capex history) and lists what's missing, so the broker asks the seller once. How PSV works with brokers
- Draft the property and location sections. Building facts, site, systems, parking and recent capital work, in the firm's template and voice, every fact tied to its document.
- Draft the tenant and rent roll summary. Tenant profiles, the expiration schedule and the unit mix or stacking plan, from the rent roll and lease abstracts. Every lease term that lands in the OM, from renewal options to termination rights, gets checked against the lease before release. AI lease abstraction
- Lay out the financial summary. In-place figures come from the seller's T-12. The broker's assumptions go in as labeled inputs to the firm's model, which computes the pro forma. The AI just places the outputs and labels every NOI as trailing, in-place or pro forma.
- Run a consistency check across the whole OM. Before release, the AI reads the draft against the data room: cover occupancy against the rent roll, square footage wherever it appears, the tenant summary against the leases. Buyers will run this exact check. Catch it first.
- Answer buyer questions from the same record. The AI drafts each answer from the data room with the source attached, and the broker approves every one.
What does a broker still write in an AI-drafted OM?
The sections that sell the deal are judgment and positioning. Those stay with the broker.
- The investment thesis. Why this asset, why now, and for which buyer. The AI can point to facts that back it. The broker decides the story.
- Pricing guidance. Where to set guidance, whether to price it at all, and how to frame the cap rate. That's the broker's market read and the seller's goals.
- Pro forma assumptions. Rent growth, lease-up and exit assumptions are the broker's to defend. The AI doesn't pick them.
- Disclaimers and confidentiality language. The firm's standard language, reviewed by its own counsel. The AI drops in the approved text. It doesn't write legal language.
How does an OM become an IC memo?
But the OM is only where an investment committee memo starts. The memo keeps what the OM proved, swaps the broker's numbers for yours and adds what the OM leaves out.
- Carry over the verified facts. Property, location and tenancy facts move over with their sources. You checked them during the screen.
- Replace the broker's numbers with your underwriting. Every income, expense and return figure comes from your model, never the OM. Where you depart from the broker's pro forma, the memo says why.
- Add what the OM leaves out. Your market view, risks and mitigants, business plan, capital structure and sensitivities. Broker Q&A goes in as an appendix.
- Draft in your own template. The AI drafts the memo in your committee's format, cites every figure to the model or a source page and flags any field the file doesn't support. Your deal lead writes the recommendation. IC decides. AI investment committee memos
Examples: the same work at real firms
Broker clients draft in their own templates. Investor clients screen deals and fill their own models. Each line below is the published result.
- Saunders Real Estate, broker opinions of value in its own template. Saunders Real Estate cut senior time per broker opinion of value from 4-8 hours to about 30 minutes, drafted in its own template off its own comparables. Read the case study
- CapRock Real Estate, leasing outreach and a branded leasing brief. CapRock Real Estate's AI-assisted leasing outreach drew 20 prospect replies and 17 follow-up openings, plus a branded leasing brief in about four minutes. Read the case study
- Westlake Realty Group, underwriting in its own model. Westlake Realty Group cut per-deal underwriting from 6-8 hours of senior time to a 20-minute review, in its own model with every number traced to source. Read the case study
- Silverado Interests, deal screening from one inbox. Silverado Interests' AI analyst works deal screening, underwriting prep and investor reporting from one email address the Dallas team forwards work to. Read the case study
Is it safe to put a seller's OM into AI, and what is prompt injection?
Two risks. Confidentiality: you signed an agreement to see this OM. And injection: someone else wrote it, and text inside it can try to steer the AI reading it.
- Read the confidentiality agreement first. Check what the confidentiality agreement allows. Then use an AI inside your firm's own Microsoft or Google tenant, or a business Claude, ChatGPT or Copilot workspace whose terms say your files don't train a shared model. Never a personal chat account. Is AI safe for deal documents
- Prompt injection in seller documents. Prompt injection is text written for an AI, not a person. Think white text on a white page, a note in the file's metadata or a footnote in tiny type, telling any AI reviewer to describe the deal favorably. You probably won't spot it in the PDF. And a model that obeys it can soften its review without telling you.
- How a careful system defends against it. It treats everything in a document as data, never as instructions, and reports any text aimed at an AI as a finding, with the page. The model reading untrusted documents can't send email, move files or touch your model. Want to test a tool? Plant an instruction in a sample PDF and see if it reports it or follows it.
Have it built, do it yourself, or buy OM software?
- Have it built around your sheet, model and templates (ours). For buyers, we build the OM screen around your screening sheet, criteria, model and memo template. For brokers, we build OM drafting around your template, plus the data room checks. Either way it runs inside your tenant, and we stay on as your AI team as your box, templates and markets change. How PSV works with CRE firms
- Do it yourself in Claude, ChatGPT or Copilot. Cheap to try, and fine for a first read if you demand a cited page for every number. Your team owns the prompts and checks. ChatGPT prompts for CRE
- Buy point OM software. Each product covers one step, like summarizing an OM, in the vendor's screens and format. Wiring it into your rent roll, T-12, model and memo is on your team.
Common questions
- What is AI offering memorandum analysis?
- An AI pulls the OM into your screening sheet with a cited page for every number, checks it against the rent roll and T-12, splits in-place income from the pro forma and lists the assumptions you need to test.
- Can AI review an offering memorandum before I underwrite it?
- Yes. A good review hands back a one-page screen: the gaps between the OM and its backup, plus a list of broker questions. Then your team picks what gets underwritten.
- Which OM sections can AI draft from a data room?
- Property facts, tenant and rent roll summaries, and the financial summary placed from the firm's model. The broker writes the thesis, the pricing guidance and the pro forma assumptions, and signs off on the whole OM.
- How do you turn an OM into an IC memo with AI?
- Carry over the verified facts, swap every broker figure for your model's output, add your view, risks and plan, and draft it in your committee template. Your deal lead still writes the recommendation.
- Can a seller hide instructions to AI inside an OM?
- Yes. White type, metadata, footnotes. A careful system treats document text as data, reports anything addressed to an AI and checks the OM against its backup.
- Should a confidential OM go into ChatGPT, Claude or Copilot?
- Check the confidentiality agreement first, then use your firm's business workspace or tenant. Never a personal account. Keep data vendors' exports out unless you have their written permission. Work from the OM, the seller's data room and your own files.
- Does AI replace reading the OM yourself?
- No. It tells you where to look. You still read what it flags, and you decide what the deal is worth.
- Is there a measured time saving for AI on OM work?
- We haven't published an OM-specific time figure. The closest measured results come from the work around it. For broker opinions of value: Saunders Real Estate cut senior time per broker opinion of value from 4-8 hours to about 30 minutes, drafted in its own template off its own comparables. For underwriting: Westlake Realty Group cut per-deal underwriting from 6-8 hours of senior time to a 20-minute review, in its own model with every number traced to source.