Ask a broker how long an offering memorandum takes and you will hear a number between three days and three weeks. Ask what actually consumes that time and the answer is almost never the writing. It is the assembly: chasing the current rent roll, reconciling it against the leases, pulling the trailing-twelve financials into a clean table, hunting down the square footage that three documents state three different ways, and stitching all of it into a package a buyer will trust. The prose — the investment highlights, the property overview, the market paragraph — is the last and fastest part. An OM is a data-assembly problem wearing the costume of a writing problem, and once you see it that way, the path to a same-day draft becomes obvious: fix the assembly, and the draft follows. The framework below does exactly that, in four passes.
The OM is a data-assembly problem, not a writing problem
Break an offering memorandum into its parts and most of them are numbers, not sentences. The rent roll. The trailing-twelve operating statement. The pro forma. The tenant summary with lease expirations and options. The unit mix or building specs. The cap-rate math that everything else supports. The narrative sections — the executive summary, the property story, the location and market paragraphs — sit on top of those numbers and repeat them in prose. When an OM is wrong or slow, the fault is almost always in the layer underneath the writing: a rent roll that disagrees with the leases, a T-12 line that was transcribed by hand, a square footage that nobody reconciled.
That reframing matters because it tells you where to spend automation. Pointing an assistant at a blank page and asking it to “write an OM” produces confident prose wrapped around numbers it cannot verify — the fastest way to send a buyer a document with a hallucinated figure in it. The work that pays off is upstream: turn the deal’s source documents into clean, structured, verified data first, and the drafting step becomes short and safe because the assistant is transcribing facts you have already checked rather than inventing them. This is the same discipline a lean firm already applies to its lease data — capture the terms once into a structured form and reuse them — described in our guide to turning lease stacks into structured data. An OM is that structured data pointed at a marketing deliverable instead of a rent roll.
The framework in four passes
The OM Production Framework is four passes, run in order. Each pass has one job, and the discipline is not to skip ahead — most blown OMs come from drafting before the data underneath is clean.
- Assemble — gather every source document for the deal into one place, so nothing is retrieved twice.
- Structure — extract the leases, rent roll, and financials into verified fields, the numbers that every section will draw from.
- Draft — generate the narrative sections from that structured data with an assistant, so prose repeats checked facts rather than inventing them.
- Verify — trace every stated number back to the document it came from before the OM leaves the building.
Passes 1 and 2 are the document-intelligence work and consume most of the effort. Passes 3 and 4 are fast once the data is clean. The whole point is that a same-day draft is only possible because the slow part — assembly and structuring — has been made repeatable.
Pass 1: Assemble the deal data in one place
Before any extraction, collect the raw inputs for the deal into a single folder or workspace: the leases and any amendments, the current rent roll, the trailing-twelve operating statement, prior-year financials, the property tax and insurance bills, any survey or site plan, title, environmental reports, and the owner’s own notes. Every OM draws from this set. The failure mode this pass prevents is retrieval churn — an analyst three sections deep in the draft who stops to email the owner for the CAM reconciliation, waits a day, and loses the thread.
Two rules make this pass pay off. First, one canonical version of each document — the executed lease, not the draft; the signed amendment, not the LOI; the final T-12, not the one from two months ago. Second, name the gaps out loud. If the rent roll predates the last two renewals, note it now, because it will change what the leases tell you in the next pass. A small firm without a document-management platform can run this in a disciplined folder structure; the tooling is secondary to the rule that everything the OM needs lives in one place before drafting begins.
Pass 2: Structure the documents into fields
This is the pass that makes or breaks the OM, and it is pure document intelligence. The job is to pull the terms that drive the financials — parties, premises and square footage, base rent and escalations, term and renewal options, CAM and expense treatment, key dates — out of each lease and into structured fields, then reconcile them against the rent roll and the T-12. When those three sources agree, the numeric spine of the OM is trustworthy. When they disagree, you have found the error before the buyer does.
A general assistant (ChatGPT, Claude, or Microsoft Copilot on a business-tier account) with a reusable, well-built extraction prompt turns each lease into a consistent set of fields quickly. What fields to capture, and why each one matters to a downstream reader, is the subject of our lease abstraction framework; the shape of an abstract that actually answers the questions a buyer will ask is covered in our look at the anatomy of a lease abstract. The critical move is to extract once into one structured store and let every OM section read from it, rather than re-pulling the same numbers for the rent roll table, the tenant summary, and the pro forma separately — a habit we make the case against in abstract once, query forever. Re-pulling is how the rent-roll table and the tenant summary in the same OM end up quoting different expiration dates.
The output of this pass is not prose. It is a clean, reconciled table: one row per lease, the rent roll validated against those rows, the T-12 tied out. That table is the OM’s foundation, and every number in the finished document should be traceable to a cell in it.
Pass 3: Draft the narrative from the structured data
Only now does writing enter. With the structured table in hand, an assistant drafts the narrative sections by working from data you have already verified rather than from a document it is reading cold. The executive summary restates the investment thesis over confirmed numbers. The property overview describes the asset from the structured specs. The tenant summary is generated directly from the abstracted lease fields. The market and location paragraphs draw on submarket context — comps, demographics, and trends the broker supplies or the assistant is given.
The prompt discipline here is to feed the assistant the verified data and ask it to write from that data, explicitly instructed not to introduce figures that are not in the inputs. A well-scoped prompt produces a credible first draft of every prose section in minutes. That draft is a starting point, not a finished OM — the broker still owns the pricing narrative, the positioning, the comps selection, and the judgment about what to emphasize for this buyer pool. Making a team fluent enough to write these prompts well — for OM sections, lease summaries, market write-ups, and buyer emails — is the entire content of a focused LLM-fluency workshop, the kind of session that turns a slow, ad-hoc process into a repeatable one. Such workshops sit in the low-single-digit-thousands range in the market; the point is fluency, not a new tool to administer.
Pass 4: Verify every number back to its source
The verification pass is the non-negotiable center of the framework, and the reason a small firm can send an AI-assisted OM to buyers without fear. The rule is simple: every number in the OM must trace to a source you can point to — a lease clause, a rent-roll cell, a T-12 line. If a figure in the draft cannot be traced, it does not go in.
Because Pass 2 built a reconciled table with each value tied to its source document, this pass is a check, not a re-derivation. Walk the OM’s key numbers — in-place rent, escalations, expirations, NOI, the cap-rate math — back to the structured table, and the table back to the documents. This is where AI-drafted material earns trust or loses it. General assistants are capable of stating a plausible wrong number in confident prose; the discipline of tracing every figure is what catches it. The reason this pass is fast rather than exhausting is that the verification cost was paid once, at capture, in Pass 2 — not re-litigated on every section. The same trap and the same fix show up across document AI generally; we treat the pattern at length in the document intelligence playbook.
Where the platforms fit
Several proptech platforms automate parts of OM production, and naming them helps calibrate what “buy” looks like. Because these feature sets change every quarter, treat each specific below as a starting point and confirm it against the vendor’s current documentation.
- Buildout is the best-known marketing-package tool for brokers, built to generate branded offering memoranda and flyers from listing data held in its system, with distribution and reporting on top. It automates the assembly and design end of the OM once the underlying data is in it — which is precisely why the upstream structuring in Pass 2 still matters, because a template only produces a clean OM if clean numbers feed it.
- CoStar and Crexi provide the market and comp data, and listing exposure, that the narrative and pricing sections draw on; Crexi and LoopNet are also where the finished OM gets distributed to buyers.
- Argus and dedicated underwriting tools own the heavy financial modeling that sits behind an institutional OM; most small-firm deals do not need that depth, but it is where the pro forma lives when they do.
The honest read for a 4–20 person firm: a marketing platform is worth it when your listing volume is high enough that the recurring subscription costs less than the hours it saves, or when consistent branding across a team matters. Below that, a business-tier general assistant plus a governed data store runs the same four passes for a fraction of the recurring cost — with the trade that you assemble and design the package yourself. A custom automation that wires extraction, drafting, and your OM template together is a larger project, commonly in the mid five figures to low six figures in the market. Those are ranges; get quotes against your own deal flow.
What “in a day” really means
“From deal data to draft in a day” is a real claim, but a precise one, and it is worth being clear about what it does and does not include. What fits in a day, once the framework is a habit: assembling the source documents, structuring and reconciling the leases and financials, drafting every narrative section, and verifying the numbers. That is a complete, trustworthy draft — the thing that used to take a week of stop-start assembly.
What a day does not include: the broker’s pricing judgment and positioning, the selection and analysis of true comparables, professional photography and drone work, the design polish that makes an OM look like your firm, and the review cycle with the owner. Those remain human, and they should. The framework does not replace the broker; it removes the mechanical assembly that was standing between the broker and the parts of the OM that actually require judgment. The compression is real, and it comes from moving the slow, error-prone data work out of the critical path — not from pretending the whole document writes itself.
Guardrails for confidential deal data
Deal data is exactly the confidential material your firm is bound to protect — leases under NDA, an owner’s financials, sale terms that are not yet public. Three guardrails keep the framework safe, and none require an IT department.
Use business-tier accounts, never consumer ones. Major providers state that inputs on their business and enterprise tiers are not used to train their models by default — verify your plan’s terms, since they change — while free consumer accounts offer no such assurance. Confidential deal documents belong only in the accounts with the right terms.
Set a one-line classification rule. Decide which document types are cleared for which tool before a deal is live, so an analyst under deadline is not making the call in the moment. The most common mistake at a small firm is pasting a strict-NDA document into the wrong account.
Keep the source-link. Store, alongside each extracted value, a pointer to the document and clause it came from. This is what makes Pass 4 fast and what lets you defend any number in the OM to a buyer’s diligence team. The broader posture a lean firm uses to run this kind of disciplined workflow against bigger competitors is the subject of our manifesto on how small CRE shops out-operate institutional giants.
FAQ
What is the OM Production Framework?
It is a four-pass process for producing an offering memorandum: assemble the deal’s source documents in one place, structure the leases and financials into verified fields, draft the narrative sections from that structured data, and verify every number back to its source. The framework treats the OM as a data-assembly problem first and a writing problem last, which is what makes a same-day draft possible.
Can AI really produce an offering memorandum in a day?
It can produce a trustworthy draft in a day once the framework is a habit — assembled documents, structured and reconciled financials, drafted prose, and verified numbers. It does not produce the finished OM: the broker’s pricing judgment, comp analysis, photography, design polish, and owner review remain human. The day-long compression comes from automating the mechanical data assembly, not the judgment.
Where does AI actually help in OM production?
In two places. First, structuring — a business-tier assistant extracts lease terms and financial figures into consistent fields far faster than manual transcription. Second, drafting — with verified data in hand, an assistant writes credible first drafts of the narrative sections in minutes. The assembly discipline and the number-by-number verification stay human-supervised.
How do I stop an AI from putting a wrong number in the OM?
Draft from structured data you have already verified, not from raw documents, and instruct the assistant explicitly not to introduce figures that are not in the inputs. Then run the verification pass: trace every stated number back to its source — a lease clause, a rent-roll cell, a T-12 line — before the OM leaves your firm. Any figure that cannot be traced does not go in.
Do I need Buildout or a platform to use this framework?
No. Buildout and similar platforms automate the assembly-and-design end of the OM and are worth buying at higher listing volume or when consistent team branding matters. The four-pass framework itself runs on a business-tier general assistant and a governed data store for a fraction of the recurring cost, with the trade that you assemble and design the package yourself.
What deal documents do I need before starting?
The executed leases and amendments, the current rent roll, the trailing-twelve operating statement, prior-year financials, tax and insurance bills, any survey or site plan, title, and relevant reports. Gather one canonical version of each into a single workspace before drafting, and name any gaps — such as a rent roll that predates recent renewals — because they change what the leases tell you.
How does lease abstraction connect to producing an OM?
The tenant summary, rent-roll table, and pro forma all draw on the same lease terms. Abstracting each lease once into structured fields — and reading every OM section from that one store — keeps those sections internally consistent. Re-pulling the same numbers separately for each section is how one OM ends up quoting different expiration dates in two places.
Is it safe to put confidential deal data into an AI tool?
With safeguards. Use a business-tier account from a major provider, which states inputs are not used to train the model by default — confirm your plan’s terms — and never use a free consumer account for confidential leases or financials. Set a rule for which document types are cleared for which tool, and keep sensitive material out of any account without the right terms.
How much does it cost to set this up?
The lightest version costs a business-tier assistant subscription and the time to build reusable extraction and drafting prompts. A workshop to make a team fluent in those prompts sits in the low-single-digit-thousands range in the market. A full custom automation wiring extraction, drafting, and your template together is a larger project, commonly mid five figures to low six figures. Get quotes against your own deal flow.
Key takeaways
- An offering memorandum is a data-assembly problem before it is a writing problem; most of its content is numbers, and most of the time it takes is spent assembling and reconciling them.
- The OM Production Framework runs in four passes — assemble, structure, draft, verify — and the discipline is not to draft before the data underneath is clean.
- The document-intelligence work (Passes 1 and 2) is where the time goes and where accuracy is won; drafting and verifying are fast once the structured data is trustworthy.
- Verification is non-negotiable: every number in the OM must trace to a source document, which is what lets a small firm send AI-assisted material to buyers with confidence.
- Platforms like Buildout, CoStar, and Crexi automate assembly, data, and distribution; a business-tier assistant plus a governed store runs the same framework for far less recurring cost. Verify vendor features against current docs.
- “In a day” means the draft — not the pricing judgment, comps, photography, design, or owner review, which stay human.
Not sure whether your firm should buy an OM platform or run this framework on a business-tier assistant and a governed store? The answer depends on your listing volume, how your deal documents arrive, and where your team loses the most time — which is exactly what a short working session sorts out. Book your free AI-readiness assessment →
Dirk Jan van Veen, PhD
