The regulatory line for AI in a boutique investment bank is clear in practice: software prepares documents, lists, and comparisons, while the bank makes every decision and every buyer contact. A platform can draft a CIM, rank a buyer list, and lay IOIs side by side. It cannot advise the owner, contact a buyer, negotiate a term, or represent anyone. The exact obligations depend on the bank’s role, jurisdiction, and applicable securities laws, and regulators continue to apply existing obligations when firms use AI tools. [2] Keeping that line visible in the firm’s documents, workflow, and client communications helps the bank show who performed the activity.
Why the line matters more for a boutique than for a large bank
A large bank has a compliance department whose job is to document who did what. A boutique has a managing director who does the work and a part-time compliance consultant who reviews it quarterly. When a regulator, an owner’s counsel, or a buyer’s counsel asks who prepared the valuation and who decided which buyers to contact, the boutique’s answer has to be immediate and unambiguous.
The risk with AI is not automatically that a regulator will object to software. FINRA states that its rules and the securities laws continue to apply when member firms use generative AI, just as they apply when firms use other tools. [2] The practical risk is a blurred record in which it is unclear whether a responsible professional performed the activity and whether the firm’s supervisory process covered the tool. If a buyer list went out and no one at the bank can say who approved it, the firm has a problem regardless of how the list was built.
The line also matters for liability. If the bank describes its software to an owner as something that finds buyers and negotiates on the owner’s behalf, the bank has told the client that the tool is doing the bank’s job. When a deal goes badly, that description can become part of the record. If the bank describes the software as something it operates to prepare materials faster, the record more accurately identifies the bank’s role.
Where the line sits in each phase of the sell-side process
The line is easiest to hold when the firm knows where it falls in each phase.
In discovery and readiness, software normalizes the owner’s financials, flags gaps in the questionnaire, and produces a readiness summary. The bank decides whether to take the mandate and what the engagement letter says.
In valuation, software runs the model and produces a range with traced inputs. The bank decides what range to present to the owner and how to explain it.
In marketing materials, software drafts the teaser and CIM. The bank decides what the positioning is, edits every claim, and approves what goes out.
In buyer matching and outreach, software produces a ranked list from documented acquisition criteria and fit data. The bank decides which buyers to contact and makes every contact. No outreach email, call, or teaser transmission originates from the software without a banker choosing to send it.
In offers, software tracks NDAs, manages data room access, and lays out IOIs for comparison. The bank decides who gets access, what the comparison means, and what to tell the owner.
In negotiation through close, software organizes the LOI terms, the exclusivity timeline, and the diligence request list. The bank negotiates. There is no phase in which the software acts on anyone’s behalf.
The common thread is that software produces an artifact and a person produces a decision. Writing that sentence into the firm’s operating procedure, once per phase, is most of the work.
The language test: how the bank describes its tools
The line is held or lost in language. Three audiences hear how the bank describes its software: owners during the pitch, buyers during the process, and the firm’s own team in daily use. All three should hear the same thing.
The acceptable formulation is that the bank uses a platform to prepare valuations, marketing materials, buyer lists, and offer comparisons, and that the bank’s professionals make every judgment and every contact. The unacceptable formulation is anything that gives the software agency: the platform recommends, advises, selects, reaches out, negotiates, or represents. Those verbs can assign the tool an activity that may carry regulatory obligations or require professional judgment. The SEC notes that whether an activity makes a person a broker depends on what the person actually does, including participation in solicitation, negotiation, or execution. [1]
This is not pedantry. Pitch decks, websites, and engagement letters get read later by people looking for a problem. A managing director can audit the firm’s materials in an afternoon by searching for those verbs and replacing them. The platform recommends buyers becomes the platform produces a ranked list from which our team selects buyers. The meaning for the owner is the same. The record is different.
Something to do this week: read the bank’s current pitch deck and website for any sentence in which software is the subject of a verb that implies judgment or contact. Rewrite each one.
The record: proving who decided
A clear line in language has to be matched by a clear line in the record. For each decision point in the process, the firm should be able to show who made the decision and when.
The practical mechanism is approval. A buyer list is not a buyer list until a named banker approves it. A CIM is not sent until a named banker approves it. An IOI comparison is not shared with the owner until a named banker approves it. The approval is logged with the person and the timestamp. This is the same mechanism the output review rule uses for numerical accuracy, applied here to the question of agency.
The log serves a second purpose. It is evidence that the firm’s professionals exercised judgment, which is what the owner is paying for and what a regulator or compliance reviewer may ask to see. FINRA says a member firm using GenAI must maintain a reasonably designed supervisory system tailored to its business, including attention to technology governance, data privacy and integrity, reliability, and accuracy. [2] A platform that generates a list and sends it without a human approval step has removed evidence of that review. A platform that generates the list and waits for approval has created it.
The firm’s compliance reviewer, whether in-house or outsourced, should be able to pull the approval log for any mandate and see a complete chain of human decisions from engagement to close. If that chain has gaps, the gap is the finding, and it is better to find it internally.
Handling the owner who wants the software to do more
Some owners, especially those who have used consumer AI, will ask why the bank does not simply let the software run the process. The answer is the same line, explained from the client’s side.
The owner is paying for the bank’s judgment, relationships, and accountability. Software has none of those. A ranked buyer list is a starting point; knowing that a buyer re-traded its last three deals, if the bank has verified that information, is the bank’s contribution. A drafted CIM is a starting point; deciding how to position customer concentration is the bank’s contribution. An IOI comparison is a starting point; advising the owner which offer is actually best is the bank’s contribution and may involve regulated or otherwise professional activity that requires the appropriate role and advice. The bank should confirm the applicable requirements with qualified counsel. [1]
The owner should also understand that the line protects them. A process in which a human is accountable for every contact gives the owner a clear person to ask when something goes wrong. Software that acts on its own can create a gap in accountability that lands on the owner.
Put this in the engagement letter in one paragraph: the firm uses software to prepare materials and manage process information, and all advice, buyer contact, and negotiation is performed by the firm’s professionals, subject to the engagement scope and applicable law. The paragraph takes two minutes to add and answers the question before it is asked.
How Acquivo Pro fits
Acquivo Pro is designed on the preparation side of the line. The firm operates it to run valuations, generate teasers and CIMs from the questionnaire and data room, build a prospect list using documented acquisition criteria and fit information, and manage NDA, IOI, and LOI flow in a single pipeline. Every output is an artifact the bank reviews and approves; the bank makes every decision and every buyer contact. Figures trace to their source, which supports the approval record. To see it on a live mandate, request a demo through the Contact page.
Key Points
- The regulatory line for AI in a boutique investment bank is that software prepares artifacts and the bank’s professionals make every decision and every buyer contact.
- A boutique bank holds the line in each sell-side phase by writing one sentence per phase stating what the software produces and what the banker decides.
- Describing software with verbs like recommends, advises, or reaches out can assign it an activity that may carry regulatory obligations, so those sentences should be rewritten.
- A logged human approval on every buyer list, CIM, and offer comparison is evidence that the bank exercised the judgment its client is paying for.
- The engagement letter should state that software prepares materials and the firm’s professionals perform all advice, contact, and negotiation within the engagement scope and applicable law.
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.