The question of disclosure in AI-assisted deal work has moved from theoretical to urgent. When an adviser deploys AI tools to extract clauses, compare terms, or flag risks across lease abstracts and offering memoranda, the client on the other side of that transaction deserves a clear account of what the technology did, where it looked, and who verified the output before it shaped a recommendation.
Why Disclosure Is a Professional Obligation, Not a Courtesy
Clients hire advisers for judgment, and judgment depends on accurate information. When AI is introduced into the document review process, the source of that information changes in ways the client cannot see without being told. A model may extract a rent escalation clause correctly nine times and misread the tenth because of inconsistent formatting in the lease draft — and the client has no way to know which version of that process produced the analysis they are relying on.
The professional relationship in commercial real estate is built on fiduciary principles, even where formal fiduciary duties vary by jurisdiction and engagement letter. Advisers routinely make recommendations that move significant capital. Any tool that materially shapes those recommendations must be disclosed, because the client's ability to give informed consent to the advice depends on knowing how it was generated.
Disclosure is also a risk management act for the adviser. If an AI extraction error later surfaces — a missed tenant termination right, an overstated remaining term, an incorrect base rent figure — the absence of any disclosure creates an asymmetric liability position. The client can credibly claim they had no knowledge the analysis was machine-assisted and no opportunity to request independent verification.
Disclosure frameworks are still being developed across commercial real estate associations and legal bar bodies internationally. What firms can do now is establish their own clear written standard, communicate it to clients at engagement, and revisit it whenever AI tools are added to the workflow. Waiting for regulatory guidance to arrive before building that standard is an avoidable error.
What the Phrase Means in Practice
"What Clients Should Be Told When AI Is Used to Review Their Deal Documents" is not a single disclosure line buried in a terms-of-service update. The phrase describes an ongoing communication practice that begins at the engagement stage, continues through the transaction, and is confirmed in the closing package. Each stage of disclosure carries different content and a different purpose.
At engagement, the client needs to understand the general framework: that AI-assisted review may be used, what categories of documents it typically touches, and what human review steps exist before any output reaches the client. This level of disclosure sets the expectation without requiring the adviser to predict every tool decision the team will make over the following months.
During active deal execution, the disclosure becomes transactional. The notation does not need to be lengthy; it needs to be present, honest, and consistent.
At closing, disclosure becomes archival. The final deal package should include a record of which documents were reviewed with AI assistance, what tools or processes were used in general terms, and who on the advisory team verified the conclusions. That record protects the adviser, informs the client's future asset management team, and creates the kind of audit trail that sophisticated occupiers and investors increasingly expect.
The Specific Facts That Must Be Communicated
Clients do not need a technical brief on large language models. They need operational facts. The adviser should be prepared to communicate, in plain language, which documents were processed through AI review, including lease agreements, letters of intent, amendment schedules, and diligence questionnaires.
The scope of the AI's task matters too. There is a material difference between using AI to extract defined fields from a standardized lease abstract template and using it to analyze free-text legal language in a bespoke ground lease with unusual carve-outs. Clients should understand which of those scenarios applied, because the second carries more inherent uncertainty.
Advisers should also communicate what the AI was not asked to do. If the tool reviewed economic terms but the team did not use it to analyze landlord liability provisions or environmental representations, that boundary should be explicit. Clients who assume that AI review is comprehensive when it was actually scoped narrowly may draw false comfort from a qualified document that was only partially machine-reviewed.
Human verification steps must be described specifically. Specificity builds confidence and creates accountability.
Finally, clients should know what to do if they want independent verification. The adviser should affirmatively state that the client has the right to have any AI-extracted analysis reviewed by their own counsel or technical experts, and the adviser should not treat that request as a challenge to their credibility. It is a reasonable exercise of due diligence.
Building a Deal-Level Disclosure Framework
A disclosure framework is not a legal disclaimer appended to every email. It is a structured set of communications matched to the stages of a transaction. The framework begins with the engagement letter, which should include a paragraph describing the firm's general use of AI-assisted tools, the categories of work where those tools may be applied, and the firm's human-review standard.
The second layer is the document delivery notation. Every AI-assisted extraction, comparison, or risk summary delivered to a client should carry a brief disclosure note in the document header or footer. The note should state that AI-assisted processing was used, name the reviewing professional, and note the date of human verification. This takes seconds to include and answers the core disclosure question before the client needs to ask it.
The third layer is the exception log. When AI output is overridden during human review — when an extracted clause is corrected, supplemented, or flagged as uncertain — that correction should be recorded internally and, where material, communicated to the client. Exception logs serve two purposes: they demonstrate that the human review step is substantive rather than perfunctory, and they create a record of the advisory team's judgment operating alongside the tool.
The fourth layer is the closing disclosure summary. A one-page attachment to the closing package, organized by document category, confirms which documents were AI-assisted, who reviewed them, and what limitations or caveats were noted during review. Many institutional clients, particularly large occupiers managing portfolio-level lease analysis software decisions and capital-markets investors with lender reporting requirements, will eventually require this documentation as a standard deliverable.
What "Human Review Before Delivery" Actually Requires
The statement that a human reviewed AI output before it reached a client is only as credible as the review process behind it. Human review of AI-extracted documents requires a defined scope, documented steps, and a standard for what qualifies as a completed check.
For more complex documents — joint venture operating agreements, development management contracts, ground lease amendments — a minimum review standard is insufficient. The reviewer needs subject-matter competence in the document type, not just familiarity with the AI tool. The firm's review assignment decisions should reflect that distinction.
Review time should be allocated proportionally to document complexity. A three-page letter of intent with four economic variables requires less review time than a forty-page amended and restated lease with a dozen defined rate references, three options, and a co-tenancy clause. Firms that apply a flat review time to all AI-assisted extractions regardless of document length are under-reviewing some documents and over-reviewing others.
Clients may ask how they can trust that the review happened. The honest answer is that the adviser's professional reputation and engagement terms are the primary assurance, supplemented by the exception log and the named-reviewer notation on each document. If a client asks to see the reviewer's checklist for a specific document, the adviser should be prepared to share a sanitized version of that process without treating it as a burdensome request.
Managing Uncertainty and Flagged Confidence Levels
AI document review tools handle uncertainty differently. Some output a confidence score alongside each extracted field. Others flag uncertain extractions with a visual indicator. Others produce a uniform output regardless of internal confidence, leaving the reviewer to identify uncertain fields through contextual judgment. Clients benefit from understanding which approach their adviser's process uses, because it affects how they should interpret the output.
Where confidence scores or uncertainty flags are available, the advisory team should establish a threshold below which the field is considered unverified until confirmed by a full manual review of the source document. That threshold should be fixed, not revisited case by case. A threshold that shifts with deal timeline pressure is not a threshold — it is an aspiration.
When an extracted field cannot be verified to the team's standard — because the source document is illegible, structurally ambiguous, or requires a legal interpretation — the client should be told plainly that the field is unverified and given a recommendation for resolution. That recommendation might be to obtain a clean copy of the document, to request a landlord or seller representation as to the correct term, or to have counsel render a legal opinion. Silence on unverified fields transfers the risk to the client invisibly.
Hypothetically, consider an AI extraction of a retail lease that flags the co-tenancy clause as uncertain because the clause references a "Major Tenant" definition that appears differently in two separate lease amendments. A complete disclosure to the client would note that the co-tenancy rights could not be confirmed from the AI extraction alone, that the team is working to reconcile the amendment language, and that no financial analysis relying on occupancy assumptions should be treated as final until that clause is resolved. That is what real disclosure looks like — not a footnote, but a live transaction flag.
How AI Review Interacts With Legal Advice
One of the most important boundaries in AI-assisted document review is the boundary between factual extraction and legal interpretation. AI tools can extract what a lease says. They cannot reliably interpret how courts in a particular jurisdiction have read similar language, whether a clause is enforceable given the governing law, or what a party's practical remedies are if a representation proves inaccurate.
Clients must be told explicitly that AI-assisted document review is not a substitute for legal review by qualified counsel. This is not a hedge or a limitation — it is an accurate description of scope. The determination of legal rights and remedies is a separate function that requires a licensed attorney operating within the applicable rules of professional conduct.
In practice, some clients, particularly smaller occupiers or private investors moving quickly, treat a thorough AI-assisted lease abstract as equivalent to a legal review. The adviser's disclosure obligation includes correcting that misunderstanding. If the engagement letter and delivery notations are clear about the distinction, that correction is part of the written record rather than an after-the-fact conversation.
Where legal counsel is already engaged on a transaction, the adviser's AI-extracted analysis and the counsel's legal review should be coordinated rather than treated as parallel independent processes. Sharing the economic and structural findings from AI-assisted document review with the client's attorney — and confirming that the attorney's legal conclusions are reflected in the adviser's financial analysis — reduces the risk of the two workstreams operating on different factual records.
Structuring Client Conversations Around AI Disclosure
The initial conversation about AI-assisted review should happen at the adviser-client relationship level, not at the document level. Waiting until a specific lease abstract is delivered to explain that AI was involved puts the client in a reactive position. Raising the topic at engagement — as part of describing the advisory process — makes it a normal feature of the working relationship rather than a disclosure event.
Advisers who are uncomfortable initiating this conversation often underestimate how well-prepared their institutional clients already are. Many large corporate occupiers, real estate investment trusts, and private equity platforms have already developed internal policies on AI use in vendor services. Raising the topic first demonstrates awareness and positions the adviser as a thoughtful operator rather than a firm trying to minimize scrutiny.
For clients who are less familiar with AI tools, the conversation should focus on what the client cares about: accuracy, verification, and accountability. Describing the human-review step in those terms — "every extracted term is cross-checked by a team member before we build the financial model" — is more useful than explaining how the extraction model works technically.
The tone of the disclosure conversation matters. Presenting AI-assisted review as a capability that expands what the advisory team can examine in a given timeframe, while being clear about the verification steps and limitations, is a more accurate and more effective framing than either overselling the technology or apologizing for using it.
Connecting Document Review to Broader Deal Execution
Document review does not exist in isolation. The terms extracted from a lease, purchase agreement, or diligence questionnaire feed directly into financial analysis, risk assessment, negotiation strategy, and client recommendation. That connected chain is exactly why disclosure at the document level has consequences at the decision level.
The client who receives a lease comparison, a financial model, and a negotiation recommendation should understand which parts of that package were informed by AI-extracted data and which were produced by the adviser's independent analysis. Communicating clearly about where in the workflow each verification step occurs is part of what clients should be told about how their deal analysis is produced.
This integration also means that document-level disclosure decisions have downstream effects. The client seeing that financial model has no way to trace the input back to its source unless the adviser's disclosure framework makes that chain visible at each stage. Building that visibility into standard workflow documentation is not a burdensome add-on; it is a professional practice standard consistent with how careful advisers have always documented the sources behind their analysis.
Portfolio-Level Considerations and Repeat AI Use
Disclosure obligations do not reset to zero on each new transaction. The disclosure framework must scale accordingly, addressing how the process handles a portfolio review where individual documents vary significantly in format, age, and complexity.
The disclosure framework must account for how that consolidated output is verified and how the client is informed about confidence levels across the full document set.
Portfolio-level AI review introduces a specific aggregation risk. An extraction error rate that is tolerable on a single-asset basis becomes a material portfolio risk when applied across thirty leases in a single report. The adviser's human-review standard for portfolio work should explicitly address how sampling decisions are made — which leases receive full review and which receive exception-based review — and that standard should be communicated to the client.
Ongoing use does not create ongoing implicit consent — it requires ongoing explicit communication.
The cost of maintaining a portfolio-level disclosure framework is primarily operational rather than financial. The investment is in process design, documentation templates, and consistent team training — not in expensive per-transaction add-ons. Advisers who present this as a standard part of their service structure, rather than as an optional upgrade, are more likely to retain institutional clients who are themselves under governance pressure to document AI use in vendor relationships. The practical implication is that building the framework once, and absorbing its overhead into standard engagement terms, is far more economical than retrofitting disclosure documentation after a client raises the issue mid-transaction.
Maintaining Trust When Errors Occur
Even with rigorous human review and clear disclosure frameworks, errors will occur. AI tools are not infallible, and human reviewers operating under time pressure make mistakes too. The adviser's handling of an error when it surfaces — whether discovered internally before delivery or by the client after delivery — has a greater impact on the professional relationship than the error itself in most cases.
If the timeline allows, inform the client that the internal review process caught an extraction issue and that the corrected analysis will be delivered at a specified time.
When an error is discovered after delivery, the adviser should inform the client promptly, explain what was incorrect, provide the corrected analysis, and describe what in the review process failed to catch the error and how that process is being adjusted. That sequence — acknowledgment, correction, explanation, process improvement — demonstrates that the firm's disclosure framework is genuine rather than performative.
The connection between disclosure and trust is direct and practical. Clients who were fully informed about the AI process and the verification steps are better positioned to treat a disclosed error as a process gap to be corrected, because the existence of a defined process gives them a framework for understanding what went wrong. When no disclosure was made, the same error presents as a more fundamental reliability question, because the client has no prior context for how the analysis was produced or what oversight existed. That asymmetry makes early, clear disclosure one of the most consequential professional decisions an adviser can make, independent of whether any error ever occurs.
Enabling Informed Client Decisions Through Transparency
Ultimately, the standard for what clients should be told when AI is used to review their deal documents is the same standard that governs all material advisory communications: the client should have the information they need to make an informed decision. If a client would make a different decision — about accepting an adviser's recommendation, about requesting independent verification, about adjusting a negotiation position — based on knowing how the analysis was produced, that information is material and must be disclosed.
Advisers who incorporate clear AI disclosure into their standard workflow — rather than treating it as an event triggered only by unusual circumstances — build a more consistent and defensible professional practice over time. The habit of communicating clearly about how analysis was produced extends naturally into other parts of the advisory relationship, including market research, comparable selection, and financial modeling, because the underlying discipline is the same: being explicit about sources, methods, and verification.
Regulatory guidance will continue to evolve, AI tool capabilities will change, and specific disclosure requirements will be refined over time. But the underlying principle — that the client's ability to exercise judgment depends on knowing how the adviser's analysis was produced — will not change regardless of which tools are involved.
For advisers building their disclosure frameworks now, the most useful step is to treat the standard as a floor rather than a ceiling. Document more than you think is required, communicate earlier than the moment of delivery, and design the exception log and closing summary as standard deliverables rather than optional add-ons.
About Advantai
Advantai is a commercial real estate intelligence and operations platform operated by ADVANTAGE AI LLC, a Delaware limited liability company. It connects client relationships, property research, documents and financial decisions in one workspace for commercial real estate teams — advisers and brokerage teams, occupier and facility teams, and portfolio teams. The platform covers CRM and origination, requirements and site selection, Property X-Ray (an interactive 3D building workspace), financial modeling and comparison, document intelligence, transactions and diligence, client collaboration, and portfolio strategy with critical dates. The optional Super Agent upgrade adds specialist, source-backed research and automated scenario analysis.
Get Started with Advantai
Ready to see your next move clearly? Go to advantaico.com, click Request a demo and tell us about your next project. Prefer to start with a single project? Visit advantaico.com/getting-started to plan your first one.