AI-Powered Offering Memorandum Generation for CRE Brokers

The offering memorandum is where a broker's time goes to die. The signed listing sits in a folder while days pass gathering due diligence, days disappear into design-and-revision email loops, and more days leak away correcting errors that accumulate across successive versions. The document that should have been marketing the deal two weeks earlier is still in production, and every hour spent reformatting a rent roll is an hour not spent on the phone with a buyer.

This is not a failure of discipline. It is a structural property of how OMs have always been made. The workflow documented by Loren Keim, of CENTURY 21 Keim Realtors and Lehigh University, puts the traditional span at up to several dozen hours and a cost of several thousand dollars per document. The tasks that consume that time are structurally insight-free: manual data entry, chart building, reformatting, rewriting the same property description across templates. None of it is judgment. None of it closes a deal.

That is why the OM became AI's best use case in brokerage. Every strong OM tells the same story in the same sequence, market context, comparables, property narrative, financials, executive summary, cover. The structure is stable and well-defined, which makes it automatable. AI does not change that structure. It builds it faster and with more consistency. And because the traditional process wastes skilled time on work that a machine can do, automation targets exactly the right problem.

What AI actually does inside the OM production workflow, step by step

AI handles the reading, extracting, drafting, and formatting layers of OM production. The broker handles judgment, strategy, and the final layer that differentiates one document from another. The division of labor is clean, and it maps directly onto the steps where traditional production stalls.

The first step is document ingestion. AI reads the raw deal folder, comprising offering memoranda, rent rolls, T12 statements, lease agreements, and debt term sheets, using natural language processing and computer vision to extract and normalize data directly into structured form. The key fields are the inputs that populate every financial summary: NOI figures, occupancy rates, lease expirations, expense categories, tenant information. Well-implemented platforms cite every extracted number back to its source page, creating an auditable lineage that matters when an investor starts scrutinizing the financials.

The second step is narrative drafting. AI turns bullet points, comps, and market data into a competent first draft of market write-ups, property descriptions, and executive summaries. The broker's role shifts from staring at a blank page to editing, which is faster and less cognitively draining. The third step is financial summarization, where AI populates the charts and summary tables from extracted data, so you remove the manual spreadsheet work. The fourth is assembly and formatting, where AI handles the slide and layout logic and produces a document that looks institutional-grade without a designer in the loop.

Keim's documented workflow uses Google NotebookLM to transform a due diligence folder into a structured draft deck, a repeatable process already in use by working brokers. What remains with the broker after all four steps is the pricing judgment, the submarket read, the investor-specific framing, and the verification pass. That is the irreplaceable portion that wins the deal.

The stack that supports this workflow is modular. A harness such as Claude or ChatGPT sits on top of a model, with connectors to deal files and CRE data sources, and knowledge layers that encode CRE methodology. Adventures in CRE's Fall 2026 edition documents how leading data providers now publish MCP servers, so you can plug licensed market data directly into Claude or ChatGPT for OM drafting. Purpose-built tools like the A.CRE Offering Memorandum Creator Assistant go further, encoding a structured OM format so the AI produces consistently organized offering memorandums, not generic output. This is an assembled, working system, not a prescription.

How much time the automation actually recovers, and where it goes instead

The time recovered is real. Keim's documented process compresses a first-draft OM from a multi-day production cycle to a fraction of the time previously required, with the broker's hours concentrated in review and customization rather than data entry and formatting. AI handles the vast majority of the mechanical work. What remains, the polish, the judgment, the strategy, is where the broker's value is concentrated and where it becomes visible to clients.

But the per-deal saving is not the point. The scalability consequence matters more. A broker who produced two or three polished OMs per week can now produce many more without adding staff, or can produce the same number and reclaim hours for prospecting, client meetings, and market analysis. The competitive advantage comes from redirecting that time toward the relationship and judgment work that actually sources and closes deals.

Freed broker time compounds. More OMs to market faster means more listings active simultaneously, more investor conversations in flight, and more opportunities to respond quickly when a buyer emerges. The document that once consumed two weeks of production time becomes a vehicle for being in more conversations, more often, with more decision-makers.

Why the data foundation matters more than the AI tool itself

The quality of an AI-generated OM is bounded by the quality of the underlying deal data. This is the condition that determines whether the workflow delivers a draft that needs light review or one that requires as much work as manual production. And in CRE, the data problem is not hypothetical. Rent rolls, lease amendments, and trailing financials are routinely inconsistent across documents. That is the ordinary state of a deal file, not an edge case.

Agentic OM workflows depend on fresh, accurate, real-time data. An agent working from stale or inconsistent inputs will produce unreliable outputs and compound errors across automated steps. Each downstream step inherits whatever was wrong upstream, and by the time the draft reaches the broker's review, the correction burden has multiplied.

Organizations with successful AI initiatives invest meaningfully more in data and analytics foundations than those that fail. In CRE terms, that means fixing fragmented property data and inconsistent rent rolls before layering on AI tools. The broker who feeds clean, current deal data into the workflow gets a draft that needs light review. The broker who feeds disorganized, mismatched documents gets a draft that requires as much effort as starting from scratch. The practical takeaway is to get the deal files in order before running the workflow.

The real objection to AI-generated OMs, and why it is an implementation problem, not a technology problem

Investor distrust of AI-generated analysis is real and it is legitimate. The trust objection takes two forms: brokers uncertain whether to use AI at all, and investors uncertain whether to trust what they receive. Neither form is unreasonable, given the stakes of a high-dollar acquisition decision.

The evidence on AI project failure points in one direction. Most initiatives that fail do so because firms expect too much too fast, a pattern cited by 57 percent of leaders who experienced at least one failed initiative, and because they deploy tools on top of inadequate data foundations. The gap between adopting an AI platform and producing better OMs is an execution gap, not a capability gap. The technology is not the limiting factor. The preparation and the review discipline are.

The answer to investor distrust is not to abandon AI-assisted production. It is to make the verification layer visible. Every extracted number gets cited back to its source document. Every financial figure gets cross-checked by the broker before the document leaves the firm. Keim's workflow frames the human review step explicitly: trust, but verify. AI can do the bulk of the work, but final review of financials, disclosures, and investor-specific messaging is non-negotiable before the document reaches a client.

The generic-output objection is answered by the knowledge layer. Purpose-built CRE tools that encode deal-specific methodology produce documents that follow the professional structure investors recognize. And the risk of not adopting AI has its own form: competitors who produce polished, market-ready OMs in a fraction of the time are already in an investor's inbox while a firm still cycling through revisions is still in production.

How brokers at smaller firms can build an AI OM workflow without enterprise resources

The mid-market broker does not need an enterprise AI contract to produce institutional-quality OMs faster. A modular stack assembled from available tools delivers the core workflow at a fraction of the cost of hiring additional analyst capacity. Adventures in CRE's Fall 2026 stack framework identifies five layers: a harness such as Claude, ChatGPT, or Microsoft Copilot; a model; connectors to email, files, CRM, and property management systems; knowledge layers that encode CRE methodology; and the human judgment that sits on top. A small firm can assemble the first four layers with modest, predictable costs. General-purpose AI tools at a modest monthly per-person cost handle document reading, narrative drafting, and question-answering, the daily productivity layer that touches every part of OM production.

CRE-specific data is now directly connectable. As of the Fall 2026 edition, providers including Green Street, MSCI Real Assets, Yardi Matrix, ATTOM, Regrid, and Moody's publish MCP servers that plug licensed market data into Claude or ChatGPT. This replaces the manual comp research that previously required separate platform logins. Purpose-built tools like the A.CRE Offering Memorandum Creator Assistant add deal-specific structure so the AI produces a CRE-formatted document rather than a generic draft.

Lev's platform takes the stack logic further for capital markets workflows. Its purpose-built AI agents for CRE, including Lev Memo for deal documentation and Lev Match for capital sourcing, run on real-time CRE data rather than generic AI retrofitted to the industry. For a broker who prefers a single platform to a self-assembled stack, Lev is a credible first-class option. The practical starting point for a small firm is straightforward: begin with a general AI subscription for OM drafting and document reading, add a CRE-specific data connector for market comps, and layer in a purpose-built OM tool once volume justifies the added structure.

What brokers should verify before any AI-generated OM reaches an investor

The broker's review pass is not an optional quality check. It is the step that converts an AI-produced draft into a document a broker can stand behind professionally and an investor can rely on. The financial figures come first. Every extracted number, NOI, occupancy, expense categories, the cap rate implied by asking price, must be traced back to its source document. The AI's citation to a source page is the starting point, not the ending point. The rent schedule and escalations deserve particular care. Rebuild the full schedule by hand and compare it to the AI output. This is where models most frequently misread fixed-dollar bumps as compounding, or miss free-rent periods that shift the effective schedule. McKelvey's 2026 analysis identifies rent escalations as one of five clauses no model signs off on without human verification.

Lease terms require the same discipline. Co-tenancy triggers and remedies, assignment and subletting rights, renewal option notice windows, these are the fields where one section quietly rewrites another and where an AI summary that gets the headline right can still miss the operative detail. The market narrative is the broker's alone: the submarket read, the framing of the opportunity, the investor-specific angle. That is the portion that distinguishes a deal document from a data dump. Disclosures and compliance belong in the document, and they are the broker's professional responsibility, not the AI's.

The division of labor that makes this workflow work is the same one that closes the piece. AI produces the draft that would have taken days. The broker's review pass is what makes it theirs. And that review pass now takes minutes, not a day, because the structure and the data are already in order. The deal gets to investors faster, and the broker's expertise goes with it intact.