This post uses hypothetical scenarios for illustrative purposes only. It does not describe any actual client, transaction, or representation, and is not legal advice.

A predictable failure mode has emerged in AI-assisted legal drafting, and it shows up within the first five pages of virtually every agreement generated by a generic prompt. The instruction is something like “draft a stock purchase agreement for a Series A financing,” and the output is a thirty-page document with a four-page definitions section (defining terms used once), mutual indemnities (in a transaction where only the seller has indemnification exposure), a notice provision requiring delivery by “nationally recognized overnight courier” (for a company where the CEO checks Slack, not the mailroom), and scattered throughout, hedging phrases like “it is important to note that” and “for the avoidance of doubt” deployed as structural filler. The document is complete in the sense that it covers every topic a stock purchase agreement should cover. It’s also wrong in the sense that it covers those topics generically rather than for this specific transaction, these specific parties, and this specific set of risks.

The problem isn’t the AI model. The problem is that a generic instruction to “draft this agreement” provides no constraints on what the output should contain, no guidance on what it should omit, and no framework for the model to evaluate whether its output is correct. The fix is to embed AI completion prompts, role-specific directives, directly in the document template itself.

What a Generic Prompt Produces and Why

Understanding the failure mode is the first step to designing the fix. When an AI model receives an instruction to draft a legal agreement, it draws on patterns in its training data to produce a document that looks like the agreements it has seen. Those agreements include large-company M&A purchase agreements, public-company proxy statements, venture-capital financing documents, commercial real estate leases, and everything in between. The model’s job, absent further instruction, is to produce something that resembles the statistical average of those documents.

That statistical average includes provisions that are common across agreement types even when they’re inappropriate for a specific transaction. Defined-terms sections appear because most training-data agreements have them, regardless of whether this agreement’s complexity warrants one. Mutual indemnities appear because they’re common in commercial contracts, even though most investment transactions have asymmetric risk profiles. Boilerplate notice provisions appear because they’re present in virtually every training-data agreement, unchanged since the era when physical mail was the primary delivery mechanism.

The result is a document that passes a superficial completeness check but fails a fitness-for-purpose review. Every provision is technically valid. Many provisions don’t belong in this specific agreement. And the guardrails needed to prevent these errors can’t be applied after the fact nearly as well as they can be built into the drafting process itself.

The Document-Embedded Prompt Approach

The core insight is that the AI completion prompt should not be a separate instruction that precedes the document. It should be part of the document’s template, embedded at the points where the AI will make drafting decisions. This approach treats the prompt as quality infrastructure: a set of constraints and directives that travel with the document and govern how any AI tool interacts with it.

A well-designed embedded prompt operates at several levels. At the document level, it establishes the role the AI is playing, the transaction context, and the boundaries of the task. At the section level, it specifies what each section must contain, what it must not contain, and what cross-references it must maintain. At the provision level, it directs the AI to make specific drafting choices rather than defaulting to generic forms.

Consider a hypothetical template for an early-stage stock purchase agreement. A document-level prompt might read: “You are drafting a stock purchase agreement for a seed-stage preferred stock financing. The company has fewer than ten employees and no general counsel. The transaction involves a single lead investor and three smaller participants. Draft for this specific context. Do not include provisions appropriate for larger or more complex transactions.” This single directive, embedded as a comment or instruction block at the top of the template, eliminates the most common category of AI drafting errors: provisions that don’t belong in this type of agreement.

Five Directives That Change the Output

Through iterative testing with legal document templates, five specific directive categories have proven effective at producing AI output that approximates what experienced counsel would draft.

First, the evidentiary-control directive. This tells the AI to treat every row in a schedule, every entry in a disclosure, and every item in a closing checklist as an evidentiary record. The practical effect is that the AI stops producing generic placeholder language (“such items as may be reasonably requested by Purchaser”) and instead produces specific, verifiable entries. A closing-conditions schedule generated under this directive will list specific deliverables rather than catch-all categories.

Second, the named-owner directive. This tells the AI to assign a named human owner to every action item, deliverable, and approval in the document. Instead of “the Company shall deliver,” the output becomes “the Company, through [Name/Title], shall deliver.” This directive prevents the common AI failure of producing documents where every obligation is assigned to an abstraction (“the Company,” “the Purchaser”) without identifying the person responsible for performance. In a transaction context, knowing that the CFO is responsible for delivering the capitalization certificate and the CEO is responsible for the officer’s certificate is operationally critical.

Third, the defined-terms reconciliation directive. This tells the AI to use a defined term only if the term appears at least three times in the operative provisions, and to expand any term used fewer than three times into its full definition at the point of use. This single directive eliminates the bloated definitions sections that are the most visible artifact of generic AI drafting. It also forces the AI to evaluate whether each defined term earns its place in the document, which is exactly the analysis an experienced drafter performs instinctively.

Fourth, the failure-simulation directive. This tells the AI to evaluate each covenant, condition, and deliverable by asking: “What happens if this obligation is not performed? Who notices, what is the consequence, and how is it enforced?” This directive produces documents with more thoughtful remedies provisions and, more importantly, surfaces obligations that the AI included by default but that serve no functional purpose. If the answer to “what happens if this isn’t performed?” is “nothing, because nobody would notice or care,” the provision probably doesn’t belong in the agreement.

Fifth, the counsel-sign-off directive. This tells the AI to never mark a legal conclusion, a representation, or a compliance determination as complete without flagging it for human counsel review. The practical effect is that the AI’s output includes inline markers (brackets, comments, highlight) at every point where a legal judgment is required. Instead of producing a representation that “the Company is in compliance with all applicable laws” as final text, the AI produces “[COUNSEL: confirm scope of compliance representation, including environmental, employment, and data-privacy carve-outs].” This transforms the document from a finished draft that might contain errors into a working draft that explicitly identifies where human judgment is needed.

Why Embedding the Prompt in the Document Matters

The reason to embed these directives in the template rather than in a separate prompt has to do with durability and portability. A prompt typed into a chat interface exists for one session. A directive embedded in a document template persists across every use of that template, across every AI tool that processes it, and across every user who works with it. When a firm maintains a library of transaction document templates, embedding AI completion prompts in those templates means that every AI-assisted draft produced from the library inherits the firm’s quality standards automatically.

This approach also addresses a governance concern that has become increasingly relevant as AI tools proliferate in legal practice. When an AI completion prompt is embedded in the document, it becomes part of the document’s version history. It can be reviewed, updated, and approved through the same process the firm uses for template maintenance. It creates a record of the instructions that governed the AI’s drafting decisions, which is relevant to both malpractice risk management and to the that practitioners owe their clients.

The broader insight is that AI-assisted legal drafting will improve not primarily through better models, though models will certainly improve, but through better prompts. And the most impactful prompts aren’t the ones typed into a chat window. They’re the ones built into the documents themselves, traveling with the template, enforcing quality constraints at the point of generation, and producing output that reflects the judgment of the practitioners who designed the template rather than the statistical average of the model’s training data.

For practitioners evaluating their own AI workflows, the question isn’t whether to use AI for drafting. The question is whether the AI is receiving instructions that produce work product you’d be willing to send under your name. If the answer is no, the fix isn’t to stop using the tool. The fix is to give it better instructions, embedded in the document, where they’ll do the most good.

If you are building or refining AI-assisted legal drafting workflows and want to discuss how embedded completion prompts can improve your corporate transaction documents, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

Download: Buy-Side Confidentiality Agreement (FMA-014) (.docx) — a companion template for this post. Adapt with counsel before use.

Buy-side M&A counsel sees the pattern often enough that it has become its own category. A strategic buyer or a private equity sponsor signs a confidentiality agreement drafted by the seller’s banker, assumes the document is boilerplate because it is three pages long and uses the usual words, and then discovers at the auction’s second round that it has agreed to a pure standstill, no clean-team mechanics, a one-way non-solicit with no carve-outs for general solicitation, and a Delaware-forum specific performance clause that converts every arguable breach into an injunction fight. The seller-friendly NDA is a thoroughly standard document in the lower middle market. It is not written to protect the buyer.

The buy-side confidentiality agreement is the fix. It is not a wholesale rewrite — most of the information-use architecture in the seller’s draft is uncontroversial — but there are eight provisions where a buyer should hold a hard line, and where sophisticated buy-side counsel will red-line even on a tight auction timeline. The ten minutes it takes to flag these at the NDA stage saves a six-figure strategic argument later.

Standstill carve-outs — pure standstills kill the auction

The standstill is the seller’s defensive provision. In a typical draft, the buyer agrees that for a specified period (12 to 24 months is standard) it will not acquire the target’s securities, propose a transaction, solicit proxies, join a group, or publicly request a waiver of the standstill. The provision sounds reasonable in the abstract. In practice, a pure standstill — one with no carve-outs — can trap a buyer who later learns, through perfectly innocent channels, that a competing bidder has emerged and that the seller is now negotiating a definitive agreement with that bidder on terms the buyer could beat.

The Delaware Chancery Court’s decision in Martin Marietta Materials v. Vulcan Materials Co., 56 A.3d 1072 (Del. Ch. 2012), stapled a two-year injunction onto a buyer that attempted a public tender offer while a standstill was still live, and the decision clarified that Delaware will enforce standstills vigorously where their terms are clear. The lesson for buyers is not to avoid signing them — strong NDAs with real standstills are a market feature — but to negotiate four specific carve-outs. First, a fiduciary-out or termination carve-out that allows the buyer to request a waiver privately if the target enters a definitive agreement with a third party. Second, a termination-upon-signing carve-out so the standstill falls away once the target signs a merger agreement with anyone. Third, a response-to-third-party-standstill-waiver carve-out, so the buyer is not disadvantaged if the seller grants a waiver to a competing bidder. Fourth, a response-to-public-tender carve-out, so the buyer can respond if an unsolicited public bid for the target emerges from outside the auction.

The Delaware Supreme Court’s decision in RAA Management v. Savvis, 45 A.3d 174 (Del. 2012), is the companion authority on the standstill-plus-no-reliance combination. Buyers who agree to both without carve-outs often find themselves without recourse even if the seller’s diligence disclosures turn out to be misleading. The pairing is a seller’s drafting signature, and buy-side counsel should either negotiate the no-reliance language down or trade for broader standstill carve-outs on the way out.

Clean-team provisions — who actually sees the sensitive commercial data

Strategic buyers that compete with the target in overlapping product lines need clean-team mechanics or the diligence process creates antitrust exposure that outlasts the deal. Even for private equity sponsors with portfolio-company overlap, clean teams are a market expectation on anything commercially sensitive. The buy-side NDA should specify a restricted-access data room tier, limit the viewers of competitively sensitive data to outside counsel and designated economic consultants, and either exclude or redact the categories that create the most antitrust friction — customer-level pricing, customer lists, forward-looking commercial plans, bid data, and cost structure.

The clean-team provision also governs which inside personnel at the buyer may see any of the sell-side data. The default a seller will offer — all buyer personnel subject to the NDA — is too broad for a strategic buyer. The negotiated version names a short list of inside personnel (frequently the deal team only, with a specified head of corporate development, a specified in-house counsel, and a specified CFO delegate) and excludes commercial, pricing, and sales personnel. Downstream, this protects the buyer against the argument that it used competitive-decision-making personnel’s exposure to seller data in post-deal pricing or bidding.

Non-solicit windows — employee non-solicit versus no-hire, with the right carve-outs

A seller-drafted NDA often sweeps the non-solicit broadly: for 18 or 24 months after signing, the buyer may not solicit or hire any employee of the target. That broad version is overbroad and, in some jurisdictions, unenforceable, but the real problem is strategic — a buyer that loses the deal has effectively been locked out of the target’s labor market, which can hamstring its own hiring for two years. The buy-side red-line narrows the restriction to a true non-solicit (not a no-hire), covers employees above a specified seniority threshold, runs for 12 months, and carries three carve-outs that matter.

First, general solicitation — advertisements, LinkedIn posts, recruiter searches not targeted at the target’s employees — must be excluded. Second, broker-sourced hires through independent search firms, provided the broker’s search was not directed at the target, should be permitted. Third, employees who approach the buyer unsolicited, or who respond to a non-targeted general advertisement, should be eligible to be hired. Without these carve-outs, a routine recruiter outreach can breach the NDA. With them, the restriction remains meaningful for its intended purpose — stopping the buyer from using the diligence process as a recruiting pipeline — without foreclosing ordinary hiring.

Residual rights — unaided memory and the recurring-industry buyer

A residuals clause states that the receiving party may use ideas, concepts, and know-how retained in the unaided memory of its personnel after the engagement, so long as no documents or embodied confidential information is used. The clause is standard in software and technology NDAs and is a market expectation for strategic buyers in recurring industries. It matters because without it, every subsequent transaction the buyer evaluates in the same sector raises the question whether the buyer’s thinking has been “informed by” information from the earlier target — a question with no clean answer and with real litigation potential.

Sellers resist residuals clauses because they correctly see them as cutting against the information-use architecture of the NDA. The negotiated middle is a narrow residuals clause: unaided memory only, no documents or physical embodiments, no identification of the target by name, no use for the purpose of competing against the target in a specific transaction, and often an explicit statement that trade secrets are excluded from the residuals carve-out. For a strategic buyer that signs five or ten NDAs a year in the same industry, that narrow clause is worth a hard negotiation.

Employee contact restrictions — the single point of contact

Seller-drafted NDAs often include an employee contact provision that reads simply: the buyer may not contact any employee of the target without the written consent of a designated representative. The buy-side version should accept the restriction but negotiate the mechanics. First, name the single point of contact at the seller — frequently the banker or the general counsel — and require that approvals be acted on within a specified business-day window. Second, carve out contacts that are not in connection with the transaction (an existing commercial relationship, for example, between the buyer and the target in the ordinary course). Third, carve out approaches initiated by the target’s employees themselves. The broad language a seller will offer can otherwise create a factual dispute every time a sales team at the buyer has a routine commercial call with the target.

Privilege protection — common interest and the diligence-analysis trap

The Delaware Chancery decision in Depomed, Inc. v. Horizon Pharma plc and the Southern District of New York line that follows the Teleglobe framework both underscore that counsel’s diligence analysis — the summary emails, the issues lists, the risk memos circulated among the buyer’s counsel, outside counsel, and the buyer’s deal team — is privileged when prepared in anticipation of litigation but can lose privilege protection when shared with the seller or the seller’s counsel without a common-interest framework. The buy-side NDA should include a common-interest provision that explicitly contemplates the exchange of privileged materials where the parties have aligned interests on specific matters (regulatory analysis, pending litigation at the target, tax structuring), coupled with language that no waiver of privilege occurs by virtue of the exchange.

The provision should also address the fate of counsel’s work product if the deal does not close. The default — return or destruction of all confidential information — would, read strictly, require destruction of outside counsel’s own diligence files, which creates malpractice-record retention problems. The buy-side red-line carves out counsel’s internal work product and litigation files from the destruction obligation, subject to the ongoing confidentiality covenant.

Return or destruction — electronic data, metadata, backup copies

The return-or-destruction obligation is standard in every NDA. The buy-side mechanics worth negotiating are three. First, destruction (not return) should be the buyer’s option — physical return of hard drives and paper files is operationally expensive, and the buyer should have the option to certify destruction instead. Second, the obligation should apply to materially-identifiable copies, not metadata, cache, backup tapes, or disaster-recovery images, which cannot practically be purged on a short timeline. Third, the certification should be signed by a designated executive or by outside counsel on behalf of the buyer, not by every individual who accessed the data room. A seller-drafted “each individual recipient” certification is operationally unworkable at any buyer of meaningful size.

Governing law, forum, and specific performance — the enforcement architecture

The last substantive negotiation on a buy-side NDA is the enforcement architecture. Sellers want Delaware law, Delaware Chancery, specific performance available without posting bond, and a prevailing-party fee-shifting clause. Each of those terms is defensible on its own, but the combination tilts the enforcement economics heavily toward the seller — Chancery issues preliminary injunctions quickly, bond-free specific performance means the seller faces no downside for seeking relief, and fee-shifting means the buyer bears the cost even of a procedural dispute.

The buy-side red-line usually accepts Delaware law and Chancery jurisdiction (the forum is a market feature for Delaware-chartered targets), negotiates specific performance as available only after a factual showing of irreparable harm (not stipulated), requires bond-posting consistent with Court of Chancery Rule 65, and either removes the fee-shifting clause or converts it to mutual-prevailing-party language. On an NDA governed by Florida law, Florida’s enforcement framework under Fla. Stat. § 542.335 for restrictive covenants is the parallel analysis, and the Florida Business Court Division is the forum sophisticated counsel will specify.

Where to go from here

The buy-side NDA attached to this post is a buyer-protective master. It marks binding versus nonbinding obligations explicitly, builds in each of the eight carve-outs discussed above, and includes placeholder fields for party names, state of governing law, and the designated single point of contact. Adapt with counsel before use.

Download: Buy-Side Confidentiality Agreement (FMA-014) (.docx) — a companion template for this post. Adapt with counsel before use.

For related buy-side and sell-side M&A drafting on this site, see our buyer-protective letter of intent, the founder/sell-side M&A NDA template, and our exclusivity agreement template. For the Florida restrictive-covenant backdrop, see the .

If you are negotiating a confidentiality agreement from the buy side and want a second set of eyes on standstill carve-outs, clean-team mechanics, or the non-solicit language before you sign, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

This post uses hypothetical scenarios for illustrative purposes only. It does not describe any actual client, transaction, or representation, and is not legal advice.

A pattern has started showing up in transaction documents over the past eighteen months, and it’s visible within the first two pages. The agreement arrives as a Word document with default blue hyperlinks in the table of contents. The definitions section runs three pages long, defining terms that appear exactly once in the operative provisions. There’s a mutual indemnity section even though only one party has any meaningful indemnification obligation. The notice provisions include a paragraph about notices sent by “nationally recognized overnight courier,” copied verbatim from a form that was last updated when FedEx was the default delivery mechanism for signature pages. And scattered throughout the document are phrases like “it is important to note that” and “for the avoidance of doubt,” deployed so frequently that they’ve become structural filler rather than clarifying language.

None of these are wrong in the way a misspelled party name is wrong. They’re wrong in the way a suit that doesn’t fit is wrong: they signal that nobody with judgment reviewed the output. And the people reading your documents, investors’ counsel, diligence teams, opposing transaction lawyers, transfer agents, can see it immediately.

The Formatting Tells Are Specific and Fixable

The first and most obvious tell is hyperlink formatting. Microsoft Word’s default style for hyperlinks is blue, underlined text. When a document is drafted natively in Word by someone who works in legal documents regularly, the hyperlinks in the table of contents and cross-references are formatted to match the surrounding text: black, same font, no underline. When a document is generated by an AI tool and pasted into Word, the default blue hyperlink style persists. This is a five-second fix (select all, change the hyperlink style), but leaving it in place tells the reader that the document went from generator to send without a human formatting pass.

The second tell is excessive bullet-point formatting. Legal documents have used numbered paragraphs, lettered sub-clauses, and Roman numeral sub-sub-clauses for decades, and for good reason: enumerated provisions create unambiguous cross-reference targets. “Section 3.2(a)(iii)” points to exactly one place. A bullet point, by contrast, has no inherent reference identifier. AI drafting tools default to bullet points because they’re optimized for readability in a chat interface, not for legal operability. When those bullets survive into a transaction document, they create provisions that cannot be precisely referenced in a bring-down certificate, a closing condition, or a side letter.

The third tell is the presence of a defined-terms section that defines terms used only once. Defined terms exist for a reason: they allow a complex concept to be stated precisely once and referenced concisely thereafter. If “Material Adverse Effect” appears fourteen times in an agreement, a defined-terms section earns its space. But when an AI-generated draft includes a definitions section that defines “Business Day” (used once, in the notice provision), “Governmental Authority” (used once, in a representation), and “Knowledge of the Company” (used once, in a single disclosure qualifier), the definitions section is performing no useful function. It’s adding two pages of text that a reviewer must cross-reference against the operative provisions, finding each time that the defined term could have been stated in plain English at the point of use without any loss of precision.

The Substantive Tells Are More Damaging

Formatting issues can be fixed in an afternoon. Substantive tells take longer to identify and create real problems when they survive into executed documents.

The most common substantive tell is the mutual indemnity in a transaction where indemnification should run in only one direction. In a typical stock purchase agreement, the seller indemnifies the buyer against breaches of the seller’s representations. The buyer might indemnify the seller against breaches of the buyer’s representations, but the buyer’s representations are usually limited to authority, organization, and investment intent, making the buyer’s indemnification obligation functionally narrow. An AI-generated draft will often produce a fully symmetric indemnification section, with identical baskets, caps, and survival periods for both parties, because the training data includes forms where symmetry was appropriate. In a seed-stage acquisition where the founders are selling and the acquirer is buying, a symmetric indemnity is structurally wrong, and its presence tells the acquirer’s counsel that nobody analyzed the indemnification economics for this specific deal.

Another substantive tell is the inclusion of notice provisions that don’t match the transaction’s actual communication patterns. A hypothetical early-stage company with three employees doesn’t need a notice provision requiring delivery “to the attention of the General Counsel, with a copy to the Chief Financial Officer, at the following address.” The company doesn’t have a general counsel or a CFO. It has a founder who reads email on a phone. A notice provision should reflect how the parties actually communicate, and an AI-generated provision that includes a multi-officer notice chain signals that the drafter didn’t tailor the document to the actual parties.

Generic hedging language represents another category. Phrases like “it is important to note that,” “for the avoidance of doubt,” and “notwithstanding the foregoing” have legitimate uses, but AI drafting tools deploy them as transitions between paragraphs, the way a student uses “moreover” in an essay. In a well-drafted legal document, “notwithstanding the foregoing” creates a specific exception to a specific preceding provision. When it appears at the start of a paragraph that doesn’t actually conflict with anything above it, a careful reader knows the drafter didn’t trace the logical relationship between provisions.

What to Actually Remove

General advice to “keep it lean” or “tighten up the drafting” doesn’t help because it doesn’t tell you what to cut. Here is what to cut, specifically.

First, remove every defined term that is used fewer than three times in the operative provisions. Replace it with the substantive language at the point of use. If “Business Day” appears once, write “any day other than a Saturday, Sunday, or federal holiday” in the sentence where it matters, and delete the definition.

Second, remove or rewrite every notice provision to reflect the actual officers and communication methods of the actual parties. If the company’s founder is the only person who will receive notices, the notice provision should say so. Email notice should be permitted if that is how the parties will actually communicate, which for early-stage transactions it almost always is.

Third, examine every indemnification, representation, and covenant provision for inappropriate symmetry. If the transaction’s risk profile runs primarily in one direction, the document’s protective provisions should run primarily in that direction. Delete the mirror-image provisions that the AI added for structural completeness.

Fourth, search the document for “it is important to note that,” “for the avoidance of doubt,” “it should be noted that,” and “as discussed above.” Delete each instance. Read the sentence without the phrase. If it still makes sense (it almost always does), leave the phrase out.

Fifth, update all hyperlink formatting to match the document’s body text style. This takes less than a minute in Word’s styles pane and eliminates the single most visible formatting tell.

Why This Matters Beyond Aesthetics

The concern here isn’t that AI-assisted drafting is inherently problematic. Used well, AI tools can accelerate first-draft production significantly, and the guardrails available for AI-assisted contract drafting are improving. The concern is that an unreviewed AI draft communicates something about the drafter’s process, and what it communicates is unflattering.

When an investor’s counsel receives a stock purchase agreement with symmetric indemnities, default-blue hyperlinks, and a three-page definitions section for a five-page agreement, the message received is: this document was generated and sent without substantive review. That assessment, whether or not it’s fair, colors the entire negotiation. Requested changes get scrutinized more carefully. Representations get pushed back on more aggressively. The transaction’s overall friction increases because the counterparty has lost confidence in the document’s quality.

In a diligence context, the effect is even more direct. A potential acquirer’s counsel reviewing a data room full of corporate documents can distinguish between documents drafted by experienced counsel and documents generated by AI tools. As the make clear, competent representation requires the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. A document that contains structural tells of unreviewed AI output raises questions about whether that standard was met.

The fix isn’t to stop using AI tools. The fix is to review AI output with the same rigor you’d apply to a first draft from a junior associate: check the formatting, verify the substantive provisions against the actual transaction, remove provisions that don’t belong, and make sure the document reads like it was prepared for this deal, by someone who understood it.

Counterparties can tell the difference. So can courts reviewing corporate governance documents. The goal is a document that reflects judgment, not just generation.

If you are reviewing transaction documents and want to ensure they reflect the professionalism your deal requires, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

Download: Fintech M&A Regulatory Diligence Checklist — FMA-017 (.docx) — a companion template for this post. Adapt with counsel before use.

The 2026 fintech M&A story plays out the same way more often than anyone wants to admit. A buyer signs an LOI on a payments or consumer-lending target with a clean CFPB complaint-portal history, a BSA/AML program on paper, and a comfortable sense that the restructured federal CFPB is not going to be the gating item. The regulatory diligence team then pulls the state side: a New York DFS exam cycle with an open MOU, a California DFPI complaint pattern that implicates UCL § 17200, a Texas OCCC subpoena from a multistate AG coalition, and a Florida OFR Chapter 560 renewal filing that is six weeks late. None of those shows up on the federal portal. All of them are the diligence story.

The practical takeaway from the 2025 restructuring of the Consumer Financial Protection Bureau is not that consumer-financial regulation went away — the governing statutes are intact, and enforcement actors have proliferated. It is that the diligence center of gravity moved. Federal supervisory posture has receded; state financial regulators, state attorneys general, and private UDAAP litigation have moved up the priority stack. For buy-side counsel in fintech, payments, and consumer-finance deals, that means the diligence checklist has to look different in 2026 than it did in 2023.

Where Consumer-Finance Regulation Sits in 2026

The CFPB’s posture is reduced but the Bureau is not gone. Dodd-Frank’s UDAAP standard at 12 U.S.C. § 5531 remains the governing federal unfairness, deceptiveness, and abusiveness framework for covered persons. Section 1033 at 12 U.S.C. § 5533 still sits in the statute book. What has changed is the frequency, scope, and intensity of CFPB enforcement, examination, and rule-making — and the slack is being taken up by state financial regulators and state attorneys general.

The 2026 regulatory map therefore has a different shape. For a national fintech, the practical risk inventory now looks less like one federal supervisor and more like a mosaic of state regulators: New York DFS (money transmission and BitLicense), California DFPI (consolidated banking, CFL licensees, CCFPL), Texas OCCC (consumer-finance lending), Florida OFR (Chapter 560 money services businesses), and parallel structures in Illinois, Massachusetts, and Washington. Each of them has independent authority, independent examination rhythms, and independent enforcement history.

State AG Enforcement Gap-Fill

The second big shift is enforcement by state attorneys general, often in multistate coalitions. State AG offices have moved into consumer-financial enforcement territory that was CFPB-dominant two or three years ago. Multistate coalitions have filed consumer-protection actions touching bank-partner lending programs, buy-now-pay-later disclosures, overdraft practices, fee transparency, and crypto-adjacent payments. The posture is familiar from the Dodd-Frank preservation of state enforcement authority at 12 U.S.C. § 5552, which was always there and is now the primary driver.

The diligence consequence is that a target’s state AG subpoena history — including CIDs, investigative demands, and multistate coalition inquiries — is now a diligence priority, not an afterthought. A clean CFPB record does not substitute. Buy-side regulatory diligence has to pull the state AG log in each state where the target holds a license or does meaningful business, and read it like a federal exam report.

UDAAP in 2026: The State Analogues Matter Most

UDAAP cases in 2026 increasingly land under state analogues, not federal § 1031. The Florida FDUTPA framework at Fla. Stat. § 501.201 and the California Unfair Competition Law at Cal. Bus. & Prof. Code § 17200 are the two most-cited state engines, with Massachusetts Chapter 93A, Illinois CFA, and New York GBL § 349 following. Each has its own standard, its own private right of action, and its own damages framework. California’s UCL in particular carries a broad “unlawful, unfair, or fraudulent” standard that reaches conduct the federal § 5531 analysis might not.

Diligence has to look at the state UDAAP history — FDUTPA class filings, UCL private attorney general actions, state AG § 349 letters — and the state consumer-complaint data behind them. The federal CFPB complaint portal is still a useful signal, but it is one data stream among six or seven.

Rule-Rescission Posture and Valuation Implications

Several targets built their 2024-2025 commercial model around specific CFPB rules that have since been rescinded, paused, or stayed. The Payday Lending Rule, the Section 1033 open-banking rule, and portions of the small-business lending data collection rule have each seen that treatment. The diligence question is not just “was the target compliant.” It is whether the historical rule structure that drove the target’s pricing, product mix, or data-access architecture has been removed, and what that does to the valuation.

A target whose unit economics depended on a particular rule structure that no longer exists has a repricing problem, not just a compliance one. The rep framework has to reach historical compliance with rules in effect at the time of the conduct, with a schedule exception for the rescinded provisions and a buyer-friendly knowledge definition. The knowledge-rep definition is the hinge; a tight knowledge group lets the buyer rely on the rep without putting founders in a position of representing what they could not have known.

Section 1033 and Open Banking

Section 1033’s open-banking framework at 12 U.S.C. § 5533 is still on the books, and the rule-making posture has shifted. For data-aggregation targets, open-banking-access-dependent fintechs, and personal-finance-management platforms, the diligence question is whether the target’s data-partner contracts assume a rule that has been paused and whether the contractual data-access architecture still works without rule-based mandatory access. If the business model relies on screen-scraping fallback, the fallback has to be legally substantiated and operationally viable.

Download: Fintech M&A Regulatory Diligence Checklist — FMA-017 (.docx) — a companion template for this post. Adapt with counsel before use.

Residual Federal Supervisor Exposure

The CFPB’s reduced posture does not mean no federal supervisor. The OCC retains authority over national banks and federal savings associations, including the bank partners that sit behind many fintech programs. The FDIC supervises state non-member banks and has primary posture on bank-fintech partnership guidance. The Federal Reserve supervises state member banks and holding companies, with the full BHCA framework intact. The FTC retains Section 5 unfair-and-deceptive authority over non-bank fintechs under 15 U.S.C. § 45 and has been more active in the 2026 fintech space than many expected.

Where the target’s product sits under a bank partnership, the diligence has to pull the partner bank’s MOUs and consent orders, not just the fintech’s own enforcement history. A bank partner under a formal action often creates program-level restrictions that flow through to the fintech target by contract.

BSA/AML: Unchanged, But Now the Baseline

Bank Secrecy Act obligations under 31 U.S.C. § 5311 are unchanged by the CFPB’s posture. MSB registration, state money transmitter licensing (including Florida OFR Chapter 560 and the parallel change-of-control framework at Fla. Stat. § 560.141), BSA officer continuity, CIP and CDD/EDD files, SAR filing history, and OFAC screening logs are all baseline diligence. In the 2026 posture, they are also more likely to be the primary federal exposure item on the deal, since the other federal supervisory streams have receded.

The change-of-control filing stack is where many fintech deals discover timing exposure. State money-transmitter statutes require pre-closing or contemporaneous change-of-control filings, and NY DFS Part 200 for BitLicense targets runs through its own approval cycle. The Florida consumer-data regime adds another layer where personal consumer information is in scope.

The Priority-Stack Shift

The practical consequence for the diligence checklist is a rotation of priorities. Items that have moved up include: state AG enforcement history (including subpoenas and CIDs), state regulatory exam history and open MOUs, UDAAP state-case review (FDUTPA, UCL § 17200, Chapter 93A, GBL § 349), consumer-complaint data from state portals and BBB, bank-partner supervisory posture, and BSA/AML licensing continuity.

Items that have moved down include: CFPB consent-decree inventory review, CFPB complaint-portal volume as a standalone signal, and CFPB enforcement-pipeline questions. Those items still belong in the diligence file; they are no longer the headline.

The rep architecture tracks the same rotation. The compliance rep should expressly reach the state UDAAP analogues (FDUTPA, UCL § 17200) rather than hiding behind a generic “consumer-financial laws” formulation. The regulatory-MAC definition should capture new state AG enforcement actions between signing and closing, not just federal actions. The specified-indemnity bucket should pull out known state AG investigations and open consent orders into a six-year carve-out, matched to the specified-indemnity architecture that the four-cap framework is designed to support.

What the Diligence File Should Contain at Closing

By closing, the fintech regulatory diligence file should hold the state license stack (money transmitter by jurisdiction, BitLicense, lender licenses, sales-finance licenses), the full BSA/AML package (written program, risk assessment, independent audit, CIP, SAR history, OFAC logs), the state AG enforcement history, the state exam reports and MOUs, the UDAAP litigation history (federal § 1031 and state analogues), the consumer-complaint data across portals, the Section 1033-era compliance posture and data-partner agreements, the consumer-arbitration-agreement review with class-action waiver analysis, the TCPA/FCRA/FDCPA class-action history, and the bank-partner supervisory file where applicable. The companion checklist above is the version of that file we hand to clients to populate.

The posture to carry into the 2026 fintech LOI is also simple. Federal CFPB posture is still relevant but no longer the headline. State regulators and state AGs are. UDAAP cases will come under state statutes more often than under § 1031. BSA/AML has not moved and is now carrying more of the federal exposure. And the diligence file either tells the state story at closing or the deal reprices.

If you are a buy-side counsel evaluating a fintech, payments, or consumer-finance target in the 2026 regulatory environment, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

Here is what a legal template looks like from the outside: a Word document with blanks where the names go. The founder downloads it, fills in the company name, swaps the date, and sends it to a counterparty. It looks right. It reads like a real agreement. And in the narrow sense of containing enforceable language, it probably is a real agreement.

Here is the problem: a template that contains the right words in the right order is doing roughly ten percent of what a well-built practitioner form actually does. The other ninety percent is the infrastructure around the words. The control sheet that forces the drafter to identify the governing documents, the conflict checks, the securities-law pathway, and the approval chain before changing a single bracket. The completion checklist that prevents someone from skipping the exhibits or forgetting the joinder. The three-pass review protocol that catches the cross-reference that now points to the wrong section because someone deleted a paragraph in the middle. The attorney-review warning that says, in plain language, this document is a starting point and not a legal opinion.

Strip all of that away and what remains is a contract that looks finished but has no quality infrastructure behind it. That is what most free template libraries deliver. It is also what most AI-generated contracts deliver, for the same reason: the output looks polished, but the process that produced it had no checkpoints.

The difference between a template and a practitioner master

A practitioner master is what a transactional lawyer actually works from. It is not a blank form. It is a controlled document with its own version number, its own completion discipline, and its own review protocol baked into the file itself. The distinction matters because the quality of a legal document is not determined by the quality of the final prose. It is determined by the quality of the process that produced the final prose.

Consider what a well-built practitioner master for a stock purchase agreement contains beyond the agreement text itself. First, a matter completion sheet at the top of the file that requires the drafter to identify the parties, the governing documents, the approval path, the securities-law basis, the closing mechanics, and the deal team contacts before drafting begins. Every one of those fields is a forcing function. A drafter who has to write down the federal exemption basis before touching Section 3 will catch the accredited-investor issue before it becomes a closing problem. A drafter who skips that step will not.

Second, an AI completion prompt embedded in the document that tells any language model exactly how to approach the form: treat every row as an evidentiary control, not a generic task; assign a named human owner and objective evidence to each open item; reconcile defined terms across all schedules; simulate the failure cases. This is not a generic instruction to “fill in the blanks.” It is a role-specific directive that constrains the model’s output to the document’s actual requirements. The º£½ÇºÚÁÏ Entrepreneur Forms Library includes these prompts in every form precisely because the prompt is part of the document’s quality infrastructure, not an afterthought.

Third, a completion checklist at the end that enumerates every step the drafter must verify before the document goes out: version control, source documents used, defined-term reconciliation, factual versus legal sign-off separation, and archival. A document that ships without running the checklist has not been completed. It has merely been drafted.

Fourth, a three-pass quality review protocol. Pass one checks completeness: every workstream, every owner, every deadline, every document, every approval. Pass two checks consistency: names, dates, defined terms, cross-references, consideration amounts, and share counts must match across every section and every exhibit. Pass three is operational proof: walk the closing script with the actual people and systems that will execute it. A document that passes all three is ready to go out. A document that has only been proofread for typos has passed zero of three.

Why the infrastructure matters more than the prose

The prose in a well-drafted agreement is important, but it is also the part that gets the most attention and needs it the least. Lawyers have been writing purchase-price adjustment clauses and indemnification baskets for decades. The language is mature. The formulations are well understood. The risk is not that the language will be bad. The risk is that the language will be perfectly good and applied to the wrong facts.

A stock purchase agreement with a flawless indemnification section and the wrong share count in the recitals is a dangerous document. A board consent with beautiful “RESOLVED” paragraphs that references the wrong certificate of incorporation is a dangerous document. A closing checklist that lists every document but assigns no owner and no deadline to any of them is a dangerous document. In each case, the words are fine. The process failed.

This is the gap that template libraries cannot close by adding more templates. The problem is not the supply of contract language. The problem is the absence of control infrastructure. A founder who downloads a stock purchase agreement template, fills in the names, and sends it to a buyer has a document that contains all of the right sections and none of the right controls. The sections will look correct to someone who does not know what to look for. The missing controls will be invisible until something goes wrong at closing or, worse, after closing.

What AI gets wrong about legal documents

The rise of AI-assisted drafting has made this problem more visible, not less. A language model asked to draft a stock purchase agreement will produce one. It will be well-organized, clearly written, and formatted professionally. It will also, with near certainty, include provisions nobody asked for: a defined-terms section that defines terms used only once, a mutual indemnity in a transaction that calls for one-way protection, a notice provision copied from a form that does not match the parties’ actual addresses, and three sentences of hedging in the representations that dilute the very protections the buyer needs.

The model does this because each addition is individually defensible and collectively invisible. No single extra clause looks wrong. The document as a whole looks thorough. The problem is that thoroughness optimized for appearance rather than function produces a document that is harder to negotiate, harder to close, and harder to enforce, because every unnecessary provision is a potential dispute point that did not need to exist.

The fix is not to ask the model to “keep it lean.” That instruction gets absorbed into the model’s general prior and changes nothing. The fix is to embed the constraints in the document itself, the way a practitioner master does: name the specific provisions that belong, name the ones that do not, require the drafter to justify every addition against the completion checklist, and run the three-pass review before anything goes out. The guardrails that actually work are structural, not stylistic.

What founders should look for in a template

A template worth using has five characteristics that distinguish it from a blank form.

First, it has a matter completion sheet or equivalent control header that forces the drafter to identify the transaction-specific facts before changing any bracketed text. If the template lets someone start editing the purchase price section without first confirming the cap table, the governing documents, and the approval authority, the template is missing its most important feature.

Second, it has version control. The document should state its version number, its preparation date, and whether it is an original practitioner master or a working draft. The fundraising and investor forms in the Entrepreneur Forms Library carry version tags for exactly this reason: a drafter needs to know whether the form reflects the current state of the law and the current state of the firm’s practice.

Third, it has a completion checklist that is specific to the document type. A stock purchase agreement checklist is different from a board consent checklist is different from a closing checklist. Generic “review before sending” instructions are not a checklist. A checklist names specific items, requires specific evidence, and assigns specific responsibility.

Fourth, it has an attorney-review warning that states plainly that the document is a starting point and not a finished product. This matters for two reasons. It protects the founder from the false confidence that comes with a professional-looking document. And it protects the lawyer who built the form from the implication that the form was intended to be used without legal review.

Fifth, it separates factual completion from legal sign-off. A drafter can fill in the company name, the share count, and the purchase price. A drafter cannot confirm the securities-law exemption, the board-approval sufficiency, or the tax treatment. The template should make that boundary visible, not bury it. A well-built secondary stock purchase agreement form separates these workstreams explicitly, with counsel sign-off fields that cannot be completed by the drafter.

The bottom line

A legal template is a container. What matters is what the container requires of the person using it. A blank form requires nothing: fill in the names, save the file, send it out. A practitioner master requires everything: identify the facts, confirm the law, run the checklist, review in three passes, get counsel sign-off, and only then send it out. The difference between those two processes is the difference between a document that looks right and a document that is right.

The good news is that the infrastructure is not hard to build or hard to use. It is a matter completion sheet, a checklist, a review protocol, and a clear warning about the boundary between drafting and legal advice. Any firm that builds its forms this way gives its clients something far more valuable than contract language. It gives them a process that catches the mistakes the language alone cannot prevent.

If you are building or reviewing transaction documents and want access to practitioner-quality forms with built-in completion discipline, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

Download: Tariff Diligence Checklist — FMA-016 (.docx) — a companion template for this post. Adapt with counsel before use.

A common 2026 cross-border M&A pattern looks like this. Buyer signs an LOI on a mid-market consumer-products target with a tidy EBITDA, a landed-cost model built in 2022, and a supply chain that leans heavily on three or four Asian suppliers. By the time quality-of-earnings closes, the finance diligence team has flagged four tariff reclassifications, a First Sale pricing structure that will not survive a CBP look-back, and an open prior-disclosure that the target’s general counsel described as “routine.” The deal does not die, but the purchase price drops seven figures, the indemnity architecture rewrites itself, and the rep & warranty insurance underwriter walks away from everything with the word “customs” in it.

Three years ago, a cross-border M&A deal cared about country-of-origin reps mostly to the extent that products were already sitting in a bonded warehouse. In 2026, the country-of-origin rep is a standalone diligence module. Reciprocal tariffs — layered under the IEEPA authorities at 50 U.S.C. §§ 1701-1707, Section 232 under 19 U.S.C. § 1862, and the Section 301 regime at 19 U.S.C. § 2411 — have moved the duty rate on a given SKU from “look it up in the HTS” to “depends on the current political cycle.” That is the diligence shift buyers who closed in 2022 did not run, and the one no buyer in 2026 can skip.

The 2026 Tariff Environment, Briefly

The 2026 trade posture is defined less by any single rate and more by the layering of authorities. IEEPA, which Congress passed in 1977 as an emergency statute, now underwrites a cycle of reciprocal duties that respond to foreign measures on a rolling basis. Section 232 national-security actions on steel, aluminum, and derivative products continue to compound. Section 301 investigations and tariff adjustments have moved beyond China to a handful of other jurisdictions. Court challenges to each authority remain live — the Court of International Trade and the Federal Circuit are the venues to watch — but a buyer sitting at an LOI table today has to assume the current rates are the rates, with volatility layered on top.

The structural consequence for M&A is that the landed-cost model in the target’s financials is only as reliable as the HTS classifications, country-of-origin claims, and First Sale documentation behind it. A target that has not refreshed those files since 2022 is working from a tariff environment that no longer exists.

The Country-of-Origin Rep as a New SPA Module

Three years ago, country-of-origin language lived in supply agreements. In 2026, it belongs in the stock purchase agreement — typically a three-pronged rep. First, that no products sold or imported by the target originate from any country subject to additional-duty treatment under IEEPA, Section 232, or Section 301, except as scheduled. Second, that all HTS classifications used by the target are accurate and supported by documented classification work (binding rulings, classification memoranda, or outside-counsel opinions). Third, that any First Sale pricing claims used to reduce the dutiable value under 19 C.F.R. § 152.103 are supported by invoice-level documentation of a bona fide sale for export to the United States, with the manufacturer’s margin substantiated.

Each of those prongs comes with its own disclosure schedule and its own diligence workpaper. Buyers who put the country-of-origin rep into a single generic sentence lose their recovery vector when the exposure shows up post-close.

Transfer Pricing as a Tariff Vector

Related-party cross-border transfer pricing used to be a tax-diligence item. In 2026 it is also a customs-diligence item. The CBP First Sale rule, IRC § 482 transfer-pricing substantiation, and IRC § 1059A’s related-party inventory limitation now intersect in ways that matter at close. A target that moves goods through an affiliated foreign manufacturer and claims First Sale pricing without a current § 482 study is running an open customs-tax exposure that is one audit letter away from a reassessment. The diligence question is whether the § 482 documentation, the intercompany agreements, and the CBP valuation disclosures tell the same story. When they do not, the delta is the exposure.

The BOM-Level Country-of-Origin Audit

Serious buy-side diligence in 2026 includes a bill-of-materials audit on the target’s top twenty input components — the inputs that drive the majority of landed cost. For each, the diligence team pulls the current country of origin, the HTS classification used at import, the FTA preference claim (if any), and the current reciprocal-tariff exposure. A target that cannot produce that map inside the data room has not run the audit itself, which usually means it has not reclassified anything since the current tariff regime took hold. A buyer who closes without running the map inherits the gap.

This is not a boutique exercise. Consumer-products, industrial-components, electronics, apparel, and even food-and-beverage targets — anything with a non-trivial import footprint — needs the audit. The exercise lives in the same workstream as a quality-of-earnings pull, and often ends up driving a quality-of-earnings reprice at the LOI-to-purchase-agreement step.

FCPA and Sanctions Overlay

Reciprocal tariffs have raised the incentive for importers to misroute goods through third-country trans-shipment, mis-mark country of origin, or misclassify under a lower-duty HTS code. Each of those is a customs violation on its own; several of them also create FCPA exposure (if a foreign customs broker is paid to look the other way) or sanctions exposure (if the transshipment point implicates OFAC-restricted jurisdictions). A diligence team that has run the country-of-origin audit but not the sanctions and FCPA overlay has done half the work. The right posture is a joint review of CBP filings, OFAC screening logs, and payment histories to foreign brokers and freight forwarders.

Download: Tariff Diligence Checklist — FMA-016 (.docx) — a companion template for this post. Adapt with counsel before use.

Indemnification Architecture

Standard-form indemnities do not catch tariff exposure well. The right 2026 structure splits pre-closing customs liability out of the general indemnity bucket into a specified indemnity, uncapped by the general cap, surviving six years to match the CBP statute of limitations at 19 U.S.C. § 1621. A specified escrow or holdback — typically two to five percent of purchase price — carries the exposure through the first full look-back audit cycle. Where the target has an open CBP binding ruling that materially drives valuation, a separate carve-out holdback tied to the ruling’s survival makes sense.

Buyers who leave tariff exposure inside the general indemnity cap discover at the first demand letter that the cap is exhausted before the customs liability lands. This is the architecture the four-cap indemnity structure was built for; tariff exposure is now the paradigm specified-indemnity bucket.

Rep & Warranty Insurance Posture

RWI underwriters have aggressively tightened their customs and tariff posture in 2026 policies. Many 2026-vintage quote forms carry a customs-and-tariff exclusion that reaches both pre-closing customs liability and any post-closing reclassification exposure. The negotiation posture matters. A blanket “customs and tariff” exclusion should become a tailored exclusion limited to matters disclosed in diligence, with everything else covered under the policy. When the broader exclusion is non-negotiable, the specified indemnity above becomes the sole recovery vehicle, and the dollar size of that escrow needs to reflect the shift.

The parallel exercise shows up in other 2026 diligence modules where RWI underwriters have carved out emerging risk categories; tariff follows that same pattern.

Deal-Killer Scenarios

Discovered tariff exposure is not always a price adjustment. There is a scenario arc where the exposure is a walk-away. In a representative pattern, a buyer discovers that the target’s entire landed-cost model assumes a First Sale claim the invoice chain cannot support, that the target has been using a lower-duty HTS code without a binding ruling, and that an open prior disclosure is being treated internally as closed. Each exposure on its own might be a price adjustment; together they imply that the diligence team cannot bound the pre-closing liability. That is a walk-away, and the LOI should have a clean termination mechanism for it — not a reverse-break fee, but a diligence-contingency termination that releases both sides. Buyers who did not negotiate that termination at the LOI step find themselves renegotiating on the back foot.

The right LOI drafting posture is a diligence contingency specifically keyed to customs, tariff, and trade-controls matters, with a cash-free, debt-free, and tariff-exposure-free closing condition that forces the finding into the open before signing. A full M&A diligence checklist is the normal starting point; tariff diligence is now the module most buyers underweight.

What the Diligence File Should Look Like at Closing

By closing, the diligence file should contain the twenty-item BOM country-of-origin audit; HTS classification memoranda or binding rulings for every material SKU; First Sale documentation for any value-reduction claim; a current § 482 transfer-pricing study with intercompany-agreement cross-references; CBP binding-ruling pulls from the ; prior-disclosure files; AD/CVD scope-ruling pulls on any product category under order; FTA preference-claim substantiation (USMCA, CAFTA-DR, KORUS); OFAC and BIS screening logs; UFLPA supply-chain mapping under 19 U.S.C. § 1307; and a tariff-reclassification risk scoring on the top SKUs by revenue. The companion checklist above is the version of that file we hand to clients to populate.

The posture to carry into the 2026 LOI is simple. Country of origin is now a rep, not a schedule. Tariff exposure is a specified indemnity, not a general bucket. RWI will push back on customs; the counter-move is a tailored exclusion. And the diligence file either tells the full story at closing or the deal reprices — sometimes to zero.

If you are evaluating a cross-border or import-heavy target in the 2026 tariff environment, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

Download: Founder M&A Letter of Intent — Buyer-Protective (FMA-013, .docx) — a companion template for this post. Bilateral master with binding/nonbinding provisions marked. Adapt with counsel before use.

Buy-side M&A counsel sees the same problem often enough that it has become a category. A strategic buyer walks into an auction, receives the sell-side letter of intent, red-lines it under time pressure, and signs a document that hands away three or four negotiating levers before the definitive agreement is ever drafted. The sell-side LOI is a genuinely useful document — it is designed to move fast, protect founder economics, and lock the buyer into a workable exclusivity window. It is not written to protect the buyer.

The buyer-protective LOI is the answer to that problem. It is a bilateral document, structured with clear binding and nonbinding provisions, that lets a buyer set the terms of the process before the definitive agreement negotiation begins. Every section is a negotiating position. The document below walks through the ten provisions that actually move value on a mid-market deal, in the order they typically get contested.

Binding versus nonbinding — the boundary that decides everything else

The first drafting move is not about price. It is about which sections of the LOI create enforceable obligations and which do not. A buyer-protective LOI marks each provision explicitly as binding or nonbinding, and the binding sections are usually a short list: exclusivity, confidentiality, expense reimbursement, governing law, and the mechanics of termination. Everything else — price, structure, escrow, indemnification — is nonbinding, subject in every respect to the parties’ successful negotiation and execution of a definitive agreement.

This distinction matters because Delaware and Florida both recognize a doctrine sometimes called Type II preliminary agreement liability. If the parties agree to a set of major terms and then agree to negotiate the remaining details in good faith, a court can hold the reneging party liable for reliance damages, even where the LOI purported to be nonbinding. The Second Circuit’s Teachers Insurance & Annuity v. Tribune line is the doctrinal anchor, and Delaware has cited it favorably. The buyer-side fix is to state, expressly and repeatedly, that no duty to negotiate exists and that either party may walk away at any time for any reason or no reason.

The price bridge — enterprise value is not equity value

The proposed enterprise value is the headline number the founder repeats to friends. Equity value is the number that hits the founder’s wire. The difference is the price bridge: enterprise value plus unrestricted closing cash, minus indebtedness, minus transaction expenses, adjusted for working capital target versus actual, adjusted for any rollover, escrow, holdback, or seller note. A buyer-protective LOI states the bridge in the LOI itself, defines each adjustment with reference to the target’s own balance sheet, and reserves the buyer’s right to adjust price at signing based on updated diligence.

The mistake founders make is treating “cash-free, debt-free” as a synonym for “we agreed on price.” It is not. Cash-free, debt-free is a definition, and the definition is negotiated. Deferred revenue, capital leases, drawn-but-unused revolver, accrued vacation, unpaid bonuses — each of these lands on one side of the bridge based on drafting, not accounting. A buyer-protective LOI names each category and where it lands.

Purchase-price adjustment — the estimate is not the number

Buyers get an estimated closing statement from the seller two or three days before closing. It is an estimate. The definitive agreement then provides for a post-closing true-up, usually within 60 to 120 days after closing, based on actual working capital, actual cash, actual debt. The buyer-protective LOI puts the true-up on the table at the LOI stage: sample calculation, methodology consistency, buyer inspection rights, dispute resolution through an independent accounting arbiter, and a specified escrow or holdback to fund any post-closing adjustment in the buyer’s favor.

Due diligence — the veto that the buyer keeps

A well-drafted buyer-protective LOI treats due diligence as a condition to signing the definitive agreement, not merely a step on the path. The clause reads, in substance: “Buyer’s proposal is subject to the satisfactory completion, in Buyer’s sole discretion, of confirmatory legal, financial, tax, commercial, technical, intellectual-property, open-source, AI, privacy, cybersecurity, environmental, regulatory, and industry-specific due diligence.” “Satisfactory in Buyer’s sole discretion” is the operative phrase. Without it, a founder can argue that the buyer is committed and that only material diligence issues justify a walk. With it, the buyer retains the option value that a bilateral term sheet is supposed to preserve.

The exclusivity window — narrower than the seller wants, wider than the seller thinks

Exclusivity is the single most contested binding provision in an LOI. The founder wants 30 days. The buyer wants 90. A negotiated middle is 45 to 60 days with an automatic extension of an additional 30 days if the buyer has, by the initial expiration, delivered a first draft of the definitive agreement and continued to pursue diligence in good faith. The buyer-protective version also names the persons covered by exclusivity (the target, its subsidiaries, its officers, its directors, its equityholders, and their respective representatives), the prohibited activities (soliciting, encouraging, entertaining, discussing, or negotiating any competing proposal), and the notice obligation if a competing inbound is received.

The exclusivity clause interacts with any fiduciary-out the target’s board has in its charter or in the LOI itself. For a target with public-company-adjacent governance (a private equity portfolio company with a sophisticated board, for example), the fiduciary-out is negotiated in the LOI itself. For a founder-controlled private target, the buyer typically negotiates the fiduciary-out down to a narrow safety valve — the target may respond to an unsolicited written bona fide competing proposal that the board reasonably determines is likely to result in a superior offer — and requires immediate notice of any contact.

Rep and warranty insurance — the retention allocation

On a mid-market deal in 2026, RWI is priced in at the LOI stage. Premium is roughly 2.5 to 3.5% of the policy limit. The retention (the buyer’s out-of-pocket exposure before RWI pays) is typically 0.5% of enterprise value, dropping to 0.25% after 12 months. The buyer-protective LOI states the split: the seller funds a specified portion of the retention through the indemnity escrow, and the buyer covers the balance. It also states the exclusions the buyer expects the policy to carry — known matters, purchase-price-adjustment items, forward-looking reps, and taxes for periods after closing — so the founder is not surprised at the underwriting stage when the RWI carrier declines to cover items the founder assumed were insured.

Founder and employee arrangements — the retention pool as a bridge

In a founder-led target, the buyer is buying the business and the founder in roughly equal parts. A buyer-protective LOI addresses the founder’s post-closing role explicitly: rollover percentage, cash bonus at closing, cash retention over 12 or 24 months, equity grants in the buyer or in NewCo, and the definition of “cause” that would forfeit any of it. It also identifies the key employees the buyer requires and reserves a retention pool — typically 2 to 5% of enterprise value — funded from the purchase price and paid over time to those key employees, subject to continued employment. The founder rarely likes this line item, but it is the buyer’s mechanism for ensuring that the business the buyer is paying for is still there in year two.

Regulatory allocation — who pays for what

HSR, CFIUS, state licensure, foreign regulatory clearances, industry-specific approvals — the buyer-protective LOI names each one applicable to the transaction, allocates the filing fees and outside-counsel costs between the parties, and states the cooperation obligation. The most contested piece is the divestiture obligation. On any transaction with a plausible antitrust theory, the buyer typically resists a “hell or high water” divestiture covenant and asks for a divestiture cap (a specified dollar amount of divestitures the buyer will accept as a condition of clearance). The seller typically pushes for the widest cap possible. The LOI stage is where this negotiation begins.

Expenses and break fees — putting the buyer’s costs into the deal

A buyer-protective LOI often includes a reverse expense reimbursement — if the seller breaches exclusivity or terminates the LOI in favor of an alternative transaction, the seller reimburses the buyer’s documented out-of-pocket costs up to a specified cap, typically 0.5 to 1.5% of enterprise value. On larger deals, a break fee replaces or supplements the expense reimbursement — a fixed liquidated-damages amount payable on defined breach events. Delaware caselaw is highly protective of break fees at reasonable levels (typically 3 to 4% of equity value is well within the enforceable range), and Florida courts follow substantially the same approach on liquidated-damages enforceability.

Governing law, forum, and equitable relief — the enforcement layer

The final binding provision is the enforcement architecture. Delaware law and Delaware Chancery are the strong buyer-side default for any transaction involving a Delaware target — Chancery is fast, sophisticated, and predictable, and Chancery will grant specific performance on exclusivity breaches, no-shop breaches, and definitive-agreement covenants where the standard is met. For Florida targets, Florida law and the applicable Florida judicial circuit (with a jury trial waiver) is the corresponding default, and the Florida Business Court Division in the appropriate circuit is the forum most sophisticated deal counsel select. The buyer-protective LOI states the choice, acknowledges that money damages would be inadequate, and expressly authorizes the buyer to seek injunctive relief without posting bond.

Where to go from here

The buyer-protective LOI attached to this post is a bilateral master. It marks binding and nonbinding provisions explicitly, includes a matter completion sheet that walks counsel through every economic and process term, and separates the substantive letter from an internal AI completion prompt and a website posting README. Adapt with counsel before use.

Download: Founder M&A Letter of Intent — Buyer-Protective (FMA-013, .docx) — a companion template for this post. Bilateral master with binding/nonbinding provisions marked. Adapt with counsel before use.

For related M&A drafting practice on this site, see our posts on the founder / sell-side LOI, our exclusivity agreement template, and the broader seller-friendly vs. buyer-friendly deal terms discussion.

If you are structuring a buy-side proposal for a private-company acquisition and want a second set of eyes on the LOI before it goes out, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

The 2026 version of this story plays out the same way in businesses and firms of every size. Someone is asked to update an agreement for a new arrangement. The AI tool produces a polished rebuild in four minutes — relocated clauses, new protective provisions, modernized structure, market-standard insurance, a tidy schedule. It reads like the work of a careful senior associate. The polish is the problem. Nobody asked for most of it, one of the “new” clauses is the original document’s own language moved to a different section, and the redline the tool generated is colored text rather than tracked changes, so nobody can accept or reject a single mark. The draft gets forwarded as “the updated version” by someone who has not read past page two.

Generative AI has made contract drafting faster than at any point in the history of the profession. It has not changed who answers for the result. The person who sends the draft owns the draft — every relocated clause, every invented cross-reference, every confident summary of a provision that says something else. Here are the guardrails that make AI-assisted drafting reliable, for the person whose name is on the work and the client whose deal is on the page.

Know the failure modes before you trust the output

AI drafting tools fail in characteristic ways, and none of them look like failure on the screen. They relocate provisions to where provisions conventionally belong, which reads as tidiness and functions as change. They add every protective term they can justify, because thoroughness is what their training rewards. They produce comparisons that are imitations — text colored red and blue — rather than genuine tracked changes a reader can accept, reject, and verify. They describe their own output with total confidence, including the parts they got wrong. And they occasionally invent things outright: a cross-reference to a section that does not exist, a defined term that was never defined, a summary of a clause that says the opposite. Public court dockets already carry examples of professionals who filed machine-invented material without checking it; the pattern is not hypothetical. None of this means the tools are unusable. It means the workflow around them has to be built for exactly these failures, the way good deal hygiene is built for human ones.

Guardrail one: real tracked changes, and tests decide

A contract turn from an AI tool should be delivered exactly as a turn from a careful associate would be: a clean draft and a redline against the version the client last saw, with the redline made of real tracked insertions and deletions. Then two mechanical tests get run before anyone reads a word. Rejecting every tracked change must reproduce the base document exactly. Accepting every tracked change must produce exactly the clean draft. If either test fails, the pair is misrepresenting what changed, and it does not go anywhere.

The point of the tests is that they replace trust with proof. A tool will describe every output as “the updated version” with perfect confidence, and a reader moving fast will believe it. The tests do not care about confidence. The same goes for the structural sweep — numbering, cross-references, brackets, leftover internal notes — which is precisely the layer where machine errors hide, because a document can be grammatically flawless and structurally broken at the same time. The full protocol is the one laid out in the version-control discipline most deals skip, and an AI-heavy workflow needs it more, not less.

Guardrail two: the instruction is less, not more

Left to its defaults, a language model drafts the way a nervous first-year drafts: it adds every provision it can justify, because each addition is individually defensible and thoroughness reads as competence. On a working form a business has used for years, that instinct produces exactly the wrong document — one the owner no longer recognizes, with three load-bearing changes buried under thirty that are not. The corrective is a discipline the tool has to be told, explicitly and every time: the form is the baseline; every mark must be required by the structure, requested by the client, or fixing a true defect; a clause that looks like an addition gets traced to the source before it is touched, because it may be the original document’s own term relocated; suggestions go in as bracketed notes to confirm, never as silent operative text; and rates, limits, and other economics do not move without a decision by the person who owns them. The reasoning behind that approach is laid out in a companion piece on what a good redline looks like, and it applies with double force when the drafter is a machine that never gets tired of adding.

Guardrail three: data handling is a configuration decision, made once

The question “is it safe to put this document in the AI” has to be answered before the matter starts, not paragraph by paragraph. Consumer-grade tools that train on what users type are a different category from enterprise deployments with contractual no-training and retention commitments, and anyone using these tools on other people’s documents should be able to say which category they are in and why. Client documents, deal terms, and anything that could identify a transaction do not go into a tool whose data handling you cannot describe. Frameworks like the formalize this instinct — map where the data goes before you rely on the system — and the mapping takes an hour, once, instead of a judgment call every afternoon. The same care extends to what the tool leaves behind: generated drafts and intermediate output live in a clearly separated working area, not commingled with reviewed final documents, so that six months later nobody has to guess which version a human actually adopted.

Guardrail four: every correction becomes doctrine

This is the guardrail with the highest return and the one most users skip. When a reviewer corrects an AI draft — cut the unrequested provision, put the original clause back where it was, restore the coverage requirement the client had chosen, turn the silent addition into a bracketed note — that correction is a rule surfacing. Left in the chat, it evaporates, and the next session makes the same mistake with the same confidence. Written down once, in plain language the tool loads before it touches the next document, it compounds. The operations getting real leverage from these tools keep those rules somewhere they own — a repository under their control, versioned like any other document, readable by every tool and every person in the practice — rather than locked inside a single vendor’s product. The result over a year is a body of drafting doctrine that reflects how the practice actually works, applied consistently by a tool that no longer needs to be re-taught, and improved every time a human catches something new.

Guardrail five: a human reads every page

The last guardrail is the oldest one. Someone reads the document — not the summary of the document, not the tool’s description of its changes, the pages. That reading is where the garbled clause that passed every structural test gets caught, where the relocated fee clause gets recognized as the original document’s own, and where the judgment a client is actually paying for happens. The division of labor that works is the same one that works in agentic M&A diligence: machines for the volume, humans for the decisions. A tool can produce five turns in an afternoon; only a reader can decide which one should exist.

What a client should ask

Clients are entitled to ask about all of this, and increasingly do. The useful questions are concrete. Does the firm or team use AI tools on my documents, and which ones? Where does my information go, and does the tool train on it? How is an AI-produced draft verified before I see it — is there a real redline, and what tests does it pass? Who owns the playbook the tool follows, and does it reflect how my documents should be handled? Anyone who has done the work will answer all four without hesitation. Anyone who cannot is telling you something important about the four-minute draft you are about to sign.

The technology is not going away, and it should not. Used with these guardrails, it lets a small team deliver turns at a pace that used to require a floor of people, without lowering the standard the work has always demanded. Used without them, it produces beautifully formatted documents that nobody read — and in most cases, the market eventually finds out which kind of document it was handed.

If you are evaluating how AI is used on your contracts, or building a drafting workflow that has to survive scrutiny, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

Here is the redline story that plays out more often than anyone in the profession likes to admit. A business owner sends her lawyer the subcontractor agreement the company has used for years and asks for it to be refreshed before the next round of hires. What comes back is a document that is forty percent blue and red. Provisions have been relocated to where they “belong.” A fee-adjustment clause has grown bracketed alternatives. There is a new representations section with its own disclosure schedule, a compliance rider, and coverage requirements rewritten to something more realistic. Every change is individually defensible. The owner reads it and asks the only question that matters: why would we have this when I didn’t ask for it?

That question is the whole subject of this post. A redline is not just a document; it is a negotiating signal and a claim about who owns the paper. The best drafting I know of changes less, and the discipline behind changing less is more rigorous than the instinct to improve everything.

The form is the baseline, not raw material

The first principle is a posture. When a client hands over a working form, that form is the baseline against which every mark is judged, not a starting point to be brought up to some external standard. It contains years of business logic in the client’s own voice, and it has been signed, repeatedly, by counterparties who accepted it. A reviser who treats it as a template to be modernized has changed the client’s document into the lawyer’s document, and the client now has to re-review their own form provision by provision to find out what they still own.

This is not an argument against changing anything. New parties, new defined terms, and a genuinely broken clause all require ink. It is an argument about where the burden sits. The default answer to any given mark is no, and the mark has to earn its way onto the page.

Every mark has to answer one question

The test for each change is necessity, and it comes in three flavors. Does the structure require it — a new signatory, a term that must now be defined, a statutory requirement? Did the client actually ask for it? Does it fix a true defect — a garbled sentence, a dropped word, a cross-reference to a section that no longer exists? If a proposed change cannot answer yes to one of those, it does not go in, however well drafted it is.

The reason is not aesthetic. Every provision in a redline is a negotiation surface. A term the other side has to read is a term they can object to, counter, or trade against something that matters. A form that comes back forty percent marked gives the counterparty forty percent more places to push, and it buries the three changes that are load-bearing under thirty that are not. There is also a credibility cost: when a reader recognizes boilerplate overreach in one place, every other demand in the document becomes suspect. Asking for what you do not need damages your standing on what you do. The same logic drives good negotiating on the money terms, which is why it recurs in the cash-free, debt-free traps that catch founders — the side that asks precisely tends to be the side that gets taken seriously.

Trace before you touch

The most embarrassing red ink in practice is the client’s own term, restyled until it looks like the lawyer’s addition. Picture a form in which a late-payment charge sits in an odd location — under the invoicing procedures rather than with the payment terms. A thorough reviser moves it to where it belongs, tidies the language, and brackets the alternatives. The client opens the redline and sees what appears to be a brand-new fee that nobody requested. Moved furniture reads as new furniture.

So before any clause is treated as an addition, or restructured, trace it to the source document. If it is the client’s own language, conform the entity names and leave it where it sits, in the order the client wrote it. Relocation is a change, and it is almost never a necessary one.

Suggestions are notes; instructions are text

Lawyers see risks clients have not thought about, and they should say so. The question is how. Terms the client has actually decided get drafted tersely, as operative text, exactly as instructed: a non-compete limited to the county where the contractor works, a sixty-day termination right for either side, an annual price adjustment tied to a published index. A suggestion the client has not yet accepted goes into the draft as bracketed text with a short note — “confirm whether to include” — so that it is visible, discussable, and removable in one keystroke. It never becomes operative silently. A client who finds an unrequested restriction quietly in force in their own form has learned something about their lawyer, and it is not the lesson the lawyer intended.

Economics and risk terms do not move on their own

Commission rates, thresholds, insurance limits, indemnities, term and termination economics — these belong to the client. In a representative scenario, a form requires a subcontractor to carry a level of coverage that counsel doubts a small operator can obtain, and counsel quietly lowers it. The reasoning may be sound. The change is still wrong to make unprompted, because it altered the client’s allocation of risk without a decision by the client. When a term looks commercially unrealistic, it becomes a question on the call list, not an edit on the page.

Keep third parties out unless the law needs them in

A related instinct is to pull affiliated companies into an agreement to make protections broader. Sometimes the law genuinely requires it. A parent company that is a party to the old contract being replaced has to sign to terminate it — a two-party document cannot end a contract one of the parties never signed. And in Florida, lets a third-party beneficiary enforce a restrictive covenant only if the contract expressly identifies that person as an intended beneficiary — so an affiliate that needs to enforce a non-solicit has to be named, or the protection is illusory. Those are reasons to bring a third party in, and the drafting should say exactly how far: a joinder “solely for purposes of Section X,” a beneficiary designation for one covenant. Absent a reason like that, the affiliate stays out. Every additional party is additional exposure, additional signature logistics, and, for a client trying to keep entities legally separate, an argument the other side did not have before.

Fix the defects, keep the voice

Minimal ink does not mean tolerating errors. A dropped verb in a waiver clause, a defined term that appears in two spellings, an “in the event that” that never says what event — these get fixed, because a court will read them and a counterparty will exploit them. What does not get fixed is the client’s phrasing where the phrasing is merely informal. The line is defect versus style, and a good reviser can articulate which side of it every mark falls on. This is also where operating agreements and other governance documents reward restraint; the drafting in a Florida LLC deadlock provision works because it does one thing precisely, not because it does everything.

How to read a redline as the client

If you are on the receiving end, here is the pass to run. Check the header first: it should state exactly which version is being compared against which, and the base should be the last document you saw. Then read every insertion and ask whether you asked for it; anything you did not request should either fix a defect you can see or come with a note explaining itself. Look at the economics — rates, caps, limits, dates — and confirm none of them moved without your decision. Look for brackets; every one should be a blank you will fill or a question you will answer, never a decision someone made for you. Look for new parties and schedules and ask what work each is doing. And remember that the most consequential change in a document is often a single word — the holdback that turned on the word “final” is the standing reminder — which is exactly why a redline with less noise is a safer redline. The fewer marks there are, the more likely you are to catch the one that matters.

Less red ink is not laziness. It is the visible result of a reviser who traced every clause, tested every change against necessity, kept the client’s decisions with the client, and left the client’s voice where they found it. In most cases, that document negotiates faster, signs cleaner, and reads, years later, as still belonging to the business that uses it.

If you are updating a form agreement or reviewing a redline you did not expect, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.

A common Florida deal pattern looks like this: three weeks into negotiating a services agreement, both sides are working from a file named something like Agreement_FINAL_v3_clean_edits(2).docx. The buyer’s counsel sends over “the latest” with a redline attached. The redline shows a change to the indemnity cap that the clean copy does not contain. Nobody can say which version the business owner approved on the phone last Thursday, because the file she approved has since been saved over. Signing is Friday.

Nothing in that scene is exotic. It is what happens when a document moves through a negotiation without version control — the discipline software teams treat as table stakes and most deals still skip. Here is what the discipline actually consists of, why each piece exists, and what it looks like now that AI tools are producing more drafts, faster, than any associate ever could.

Version zero is the document you received, not the one you wrote

Every negotiation starts with a document someone else produced — the counterparty’s form, the client’s decade-old template, the prior deal’s purchase agreement. That document is version zero, and the first rule is that it is never edited. It is saved exactly as received, in its own folder, and every later version is measured against it.

The rule sounds trivial until the base arrives as a PDF. The instinct is to have someone retype it or run it through a converter and start marking up the result. The disciplined move is to convert it to an editable file and then verify the conversion word for word against the PDF before a single change is made. A redline is only as trustworthy as its base; if the base silently dropped a sentence in conversion, every comparison downstream inherits the error and reports it as your change or, worse, hides it entirely.

Nothing gets overwritten, because history is evidence

The second rule is that no version is ever saved over. Each revision that leaves the working session gets a number — v01, v02, v03 — and the prior one is archived, not deleted. Iterating inside a single working session does not require a new number for every save; the boundary is exposure. The moment a draft has been seen by the client, the other side, or a reviewing partner, it is frozen forever.

This matters because a negotiation is a sequence of positions, and disputes are fought over the sequence. A post-closing fight routinely turns on when a word entered the document and who put it there. Consider the eight-figure holdback that turned on the single word “final” — the kind of dispute where the ability to reconstruct exactly which draft introduced a term, and whose redline carried it, is worth more than any argument about what the parties must have meant. If your file history is one overwritten document, you have no history, and you are litigating from memory against someone who kept theirs.

Every version is a pair

A version is not a file. It is two files: the clean draft and a redline showing every change against the base the recipient already knows. Sending a clean copy with the changes described in an email is not version control; it asks the reader to reconstruct the diff by hand and to trust that the description is complete. Sending only a redline is not much better, because the recipient has to accept every change themselves to see what they would be signing.

Three things make the pair trustworthy. First, the redline has to consist of real tracked changes — insertions and deletions the document itself records — not text colored red and blue to look like one. A cosmetic redline cannot be accepted or rejected, cannot be verified, and cannot be relied on. Second, the redline runs against the base the other side last saw. A comparison against your own intermediate draft tells the counterparty nothing useful and, when the two documents disagree, invites the suspicion that something was slipped in between. Third, the header of the redline states exactly which two versions are being compared, so that a reader six months later does not have to guess.

Verification is a test, not a feeling

The clean and the redline are produced separately and can diverge. That divergence is the most dangerous error in document work, because the counterparty reads both and will find the discrepancy at the worst possible moment. So before any version leaves, two mechanical tests get run, and they are tests in the literal sense: they either pass or they do not.

The first test is reject-all. Rejecting every tracked change in the redline must reproduce the base version word for word. If it does not, the redline is misrepresenting what was changed. The second test is accept-all. Accepting every tracked change must produce exactly the clean draft. If it does not, the clean contains something the redline never disclosed, or the redline promises something the clean lacks. The clean, for this reason, is never retyped; it is generated by accepting the tracked file, so that the two cannot drift.

Around those two tests sits a shorter sweep: section numbering is an unbroken sequence, every internal cross-reference points at a section that still exists after the edits, brackets balance and each remaining bracket is a deliberate blank, and nothing intended as an internal note to draft survives in a document headed outside. Then someone reads the rendered redline page by page. Not skims — reads. Software catches structure; only a reader catches a garbled clause that passed every structural check.

The label on the outbound draft is a promise

Internal version numbers are working history and can run high; a heavily negotiated form may pass through a dozen internal turns before the client sees the second one. What the recipient sees should be a clean external counter — the first draft sent is their v1, dated and labeled, whatever it was internally. What actually went out is frozen in its own folder, so that there is never a question about which document a recipient has in hand, and the internal history stays intact behind it.

The label is a promise because Florida law does not wait for signature pages to treat words as binding. Under Florida’s Uniform Electronic Transaction Act, , an email exchange agreeing to a change can be an enforceable record with an enforceable signature. In a representative scenario, an operations manager emails the other side that “the revised exhibit works for us” and the reply says “agreed” — that exchange is a version event whether or not anyone updates a document, and a version log that does not capture it has a hole in it. The discipline is to record, in one line per version, what changed, on whose instruction, and by what medium, including the emails.

What changes when the AI is in the room

Generative AI tools now produce full contract turns in minutes, and the temptation is to treat speed as a reason to loosen the process. It is the opposite. A tool that can generate five drafts in an afternoon can also overwrite, retype, and cosmetically color five drafts in an afternoon, and it will describe each as “the updated version” with perfect confidence. The rules above are exactly the rules that keep a fast tool honest: a frozen base, real tracked changes, the reject-all and accept-all tests reported in plain terms with every delivery, and a human who reads the pages. The firms getting real leverage from these tools are the ones that made the tool follow the discipline, not the ones that let the tool replace it. That is the same conclusion reached in a different setting in what agentic M&A execution changes and what still requires a human, and it holds even more firmly for the document a client will actually sign.

None of this is glamorous, and none of it shows up in a term sheet. It shows up when a deal goes sideways and one side can say, with a file to prove it, exactly what every draft said and when — and the other side cannot. The likely outcome in that fight favors the side that kept the versions.

If you are negotiating an agreement and want the drafting process run with this kind of discipline, feel free to reach out to our firm manager, Magda, at Magda@montague.law, or fill out our contact form. Mention you read this post.