Treat every missing, undefined, conflicting, or conditional offer field as an exception, not as an answer to infer. Capture the bidder’s exact wording and document location, classify the issue, state why it matters, draft a neutral clarification question, and assign an internal owner such as the deal lead, finance reviewer, or counsel. An AI tool may help locate passages, compare versions, and propose a question, but a designated human must verify the source and approve any internal use. Keep the register internal until the deal team decides what, if anything, can be shared. Do not let software contact a bidder, interpret legal effect, approve a term, or silently normalize figures.
Start with an exception, not an inferred answer
An offer comparison becomes unreliable when an empty field is filled with a plausible assumption. The safer unit of work is an exception record. Create one when a term is absent, a definition is missing, a condition is unclear, or two passages do not reconcile. This applies to illustrative fields such as purchase price, consideration mix, assumed debt, working-capital treatment, earn-out mechanics, timing, financing conditions, exclusivity, or stated binding effect. The register records uncertainty rather than converting it into a clean-looking number.
Use the offer, email, attachment, version label, and page or section locator as the source of truth for each record. A secondary spreadsheet or model can summarize the issue, but it should not replace the underlying passage. ABA guidance describes an LOI as a blueprint for primary proposed terms and emphasizes tracking open questions, negotiation points, and document versions during a transaction. [3]
Define the fields every exception needs
A useful register should be specific enough that another reviewer can reproduce the concern. Include an exception ID, offer or bidder label, received date, document name and version, exact source passage, location, affected comparison field, issue type, and current status. Add a plain-language explanation of what is unclear without stating a legal conclusion. Record the clarification question, proposed internal owner, reviewer, priority, dependencies, and next review date.
Separate source text from analyst commentary. Quote only the short passage needed to locate the issue, then label any paraphrase as a paraphrase. If a figure is copied, preserve its unit, currency, sign, period, and whether it is stated as cash, stock, debt assumed, contingent consideration, or another component. If the document does not say, write “not stated” rather than selecting a default.
Classify ambiguity before routing it
Classification makes escalation more consistent. Use a small controlled vocabulary such as missing, undefined, internally inconsistent, conditional, unit or period mismatch, version conflict, and source unreadable. A missing definition is different from a conflicting figure. A condition that depends on financing is different from a term that simply has no stated date. Those distinctions help the owner ask a focused question instead of reopening the entire offer.
Add a materiality rationale without pretending to calculate a result that the source does not support. For example, say that an undefined debt assumption may prevent like-for-like enterprise-value comparison, or that an unstated closing condition may affect process planning. Do not describe the legal effect of a clause unless qualified counsel makes that determination for the relevant jurisdiction and transaction documents. The register is an internal triage instrument, not a legal opinion.
Use AI for locating and comparing, not deciding
An AI-assisted workflow can search a permitted document set for offer-term labels, extract candidate passages, flag repeated terms with different values, and draft a neutral clarification question. The prompt should require citations to the supplied document and should return “not found” when the passage is absent. It should not be asked to complete a term from market convention, guess the bidder’s intent, or choose the better interpretation.
NIST describes its AI Risk Management Framework as voluntary guidance for managing AI risks and incorporating trustworthiness considerations into the design, development, use, and evaluation of AI systems. [1] Apply that principle operationally: document the task, source set, output date, reviewer, and known limitations. If an output cannot point back to a passage, treat it as a lead for review, not as evidence.
Assign the right human escalation owner
Ownership should follow the question, not the software. A deal lead or banker can usually own a commercial clarification about process timing or consideration presentation. A finance reviewer can check arithmetic, units, periods, and consistency with the comparison model. Counsel should review questions involving binding language, exclusivity, conditions, representations, regulatory matters, or other legal interpretation. A process owner can manage status, version control, and the internal queue.
Name one accountable owner and, where needed, one consulted reviewer. Avoid assigning an issue to a broad group with no decision path. The owner is responsible for verifying the passage, refining the question, and recording the human disposition. That disposition can be “question approved,” “source reconciled,” “hold for counsel,” “not material,” or “unresolved.” It should never be “AI accepted.”
Write clarification questions that do not lead
A good internal question describes the gap and asks what must be confirmed. For an absent definition, ask: “The offer states a purchase price of [quoted amount], but does not define whether assumed debt is included. What treatment should the team request clarification on?” For an inconsistent figure, identify both passages and ask which value, unit, or period is intended. For a condition, quote the condition and ask the owner to identify the factual dependency that must be confirmed.
Avoid questions that embed an answer, characterize a term as binding, or invite software to negotiate. The register can prepare a draft for human review. A person decides whether the question is appropriate, whether it should be sent, and who may send it. No AI workflow should contact the bidder, change the offer, make a concession, approve a term, or disclose a comparison against another bidder.
Set permissions, checks, and stopping conditions
Before using an AI tool, identify the permitted corpus and the people authorized to access it. Confirm the team’s confidentiality policy, the tool’s handling settings, retention terms, and whether counsel’s approval or client consent is required for the proposed use. Do not treat a generic product statement as proof that a particular deal is secure or privileged. ABA commentary on Formal Opinion 512 highlights confidentiality safeguards, informed consent considerations, knowledge of tool limitations, and the need for human review and supervision. [2] Requirements can vary by jurisdiction, engagement, and firm policy.
Stop the workflow when the source cannot be found, OCR is unreliable, versions cannot be reconciled, a prompt exposes restricted information, the output cites a passage that does not say what it claims, or the issue could change valuation, structure, legal rights, or process strategy. Escalate those cases to the designated human reviewer. A pilot should be discontinued if reviewers cannot reproduce the source trail or if repeated tests produce unsupported answers.
Review the register before it drives comparison
The reviewer should check every populated field against the original offer and confirm that unresolved items remain visibly unresolved. Recalculate only from verified figures, preserving units and periods. Compare versions before deciding that a discrepancy is real. Check that each question is neutral, each owner is named, and each high-impact item has an appropriate finance or legal review path. Keep a record of corrections so the comparison model and narrative use the same current source.
A useful output is an exception register with source passages, clarification questions, named escalation owners, status, and a clear boundary around what has not been determined. That output supports an organized discussion without claiming that the software has interpreted the offer. It also gives the deal team a defensible stopping point when evidence is incomplete.
Keep the workflow internal and user-operated
The practical objective is not to make an ambiguous offer look complete. It is to make uncertainty visible, route it to the person who can resolve it, and preserve the evidence used in the discussion. Acquivo is a software platform that delivers the entire sell-side M&A process, valuation, marketing materials, buyer matching, and deal management, on one platform for business owners, M&A attorneys, and boutique investment banks. The workflow described here remains user-operated: the deal team controls documents, permissions, review, outreach, and decisions.
Before closing the review cycle, confirm that the register contains no unsupported inferred terms, no unreviewed AI passages, and no external action taken by software. Then carry only human-approved clarifications into the next internal comparison or counsel discussion.
Key Points
- Create an exception for every absent, undefined, conflicting, conditional, or unreadable offer term instead of inferring an answer.
- Preserve the exact source passage, document version, location, units, and period so a reviewer can reproduce the issue.
- Use AI to locate passages, compare versions, and draft questions, while a named human verifies every output.
- Route commercial, financial, process, and legal questions to accountable owners with the right expertise.
- Stop when the source trail fails, confidentiality permissions are unclear, or the issue could affect valuation, structure, rights, or strategy.
- Keep clarification drafts internal until an authorized human decides whether and how to communicate externally.
Sources and references
Source retrieval dates record research access, not article publication. Read our editorial and date policy.
Choose your next step
Acquivo is user-operated software. Owners and their professionals remain responsible for decisions, communications, and professional review.
Educational information, not individual legal, tax, or investment advice. Consult qualified professionals for your circumstances. Acquivo provides software, not representation or negotiation on your behalf.