Judgment vs. Production: The Only AI Framework a Boutique Bank Needs

The one-question test for any task

Managing directors do not need a maturity model or a vendor's capability matrix. They need a question they can ask about any item on an associate's task list: if two good associates did this independently, would the results be interchangeable?

If yes, it is production. A buyer list built from the rule "sponsors and strategics that closed a comparable acquisition within two years" will come out nearly the same no matter who builds it. A financial summary that pulls three years of P&L into a standard layout has one correct form. An NDA tracker has one correct state at any moment. These tasks have a right answer, and a person reaching that answer is spending hours on something a machine can reach faster and more consistently.

If no, it is judgment. Which of the names on that list the managing director calls personally, and in what order, depends on relationships and reads that no database holds. Whether the add-back for the owner's brother-in-law's salary is defensible is a judgment call. Whether to run a broad auction or a targeted process of twelve buyers is the single most important decision on the mandate, and it belongs to the banker.

The test is not perfect at the edges, and the next sections deal with those edges. But applied with care across a mandate, it sorts most of the work on the first pass, and it does so without any technical knowledge.

The production list, by phase

It helps to see the sorted list concretely, phase by phase. In discovery and readiness, production includes collecting documents through a single portal, normalizing financial statements into a consistent format, and flagging gaps against a standard readiness checklist. In valuation, production is the workbook itself: the comparable set drawn from closed transactions in the vertical, the adjusted EBITDA bridge, the sensitivity table. In marketing materials, production is the first-draft teaser and CIM, assembled from the questionnaire and the data room into the structure every buyer expects.

In buyer matching and outreach, production is the long list, filtered by sector, size, and documented acquisition criteria, and the log of every touch. In the offers phase, production is the NDA queue, data room permissions, and the IOI comparison table drawn from the letters. In negotiation through close, production shrinks to tracking: the diligence request list, the open-items log, the LOI milestones.

Notice that production exists in every phase but is concentrated in the middle four. A boutique that moves production to software in valuation, marketing materials, buyer matching, and offers has addressed where many junior hours go. The first and last phases remain mostly human, and that is correct.

The judgment list, and why it is shorter than it feels

Now the other column. In discovery, judgment is whether to take the mandate and what to tell the owner about his odds. In valuation, it is the view on multiple and which comparables actually apply. In marketing materials, it is the narrative: which growth story is credible, what to say about customer concentration, how to position a management team that will not stay. In buyer matching, it is the short list and the call order. In offers, it is which IOIs to advance and what to tell each buyer about where they stand. In negotiation, it is almost everything.

When a managing director writes this list out, two things usually surprise her. The first is that the judgment list is shorter than the production list. The second is that judgment is concentrated at the beginning and end of the mandate, and in a handful of moments in between. The weeks between the kickoff and the first buyer call are mostly production. So are the weeks between launch and the IOI deadline.

This matters because it tells the bank where recovered time could come from and where it will not. A bank will not save hours in negotiation. It may save them in the production-heavy stretches, which is where the associate and analyst live.

The gray zone and how to handle it

Some tasks resist the test. The CIM narrative is the clearest example. The structure and the figures are production. The three paragraphs that explain why this business will be worth more to a buyer than its EBITDA suggests are judgment. The same document holds both.

The rule for mixed tasks: software produces, a named person edits, and the edit is where judgment enters. The associate does not write the CIM from a blank page; she corrects a draft that already has the right sections and the right numbers, and she rewrites the narrative paragraphs with what she heard from the owner. The time saved is the time she would have spent on structure and figures. The judgment she applies is the same as before, applied to a better starting point.

Buyer selection works the same way. The database produces a list of names that fit the rule. The managing director applies judgment to it: adding the family office she met at a conference, removing the sponsor that just lost its platform CEO, moving three names to the top because she can get them on the phone this week. The list is production. The edits are judgment. The output is better than either alone.

A second gray-zone rule: when in doubt, treat it as judgment for the first mandate and reclassify after. The cost of wrongly automating a judgment task is a bad document in front of a buyer. The cost of wrongly keeping a production task manual is a few hours. The asymmetry favors caution on the first pass.

What the framework says about tool selection

Once the work is sorted, tool evaluation gets simple. A tool is worth considering if it takes a production task off a person's desk completely, not partially. A buyer database that still requires the analyst to verify every entry against another source has not moved the task. A CIM generator that produces prose but no figures has moved half of it, and the half that remains is the slower half.

The framework also says what a tool must not do. It must not make judgment calls on the bank's behalf. A platform that recommends a multiple, picks the buyers to contact, or drafts the counter to an IOI is doing the bank's work with the bank's license on the line. The bank should want software that produces drafts, keeps records, and traces every figure to its source, and that stops there. FINRA's 2026 guidance similarly emphasizes supervision, documentation, testing, monitoring, and human review for firms using generative AI, while NIST describes its AI RMF as a voluntary framework for managing AI risk and incorporating trustworthiness into system use and evaluation [1] [2]. The managing director decides, the associate contacts the buyer, and the software holds the record of what was decided and who was contacted.

Applying it this week

The practical step is a one-hour exercise. Take the task list from the mandate currently closest to launch and sort every item into one of the two columns. Do not assume that most items are production without measuring your own list. Mark the items that are repetitive, source-based, and reviewable, then note which person currently owns each one. That is the capacity a bank can investigate for recovery.

Then pick the production task that costs the most hours and ask what it would take to have software produce the first draft. Usually that task is either the CIM or the buyer list. Either is a sensible place to start, because both have a visible before-and-after and both touch every subsequent phase. Record the baseline hours, rework, and review owner before the pilot so the bank can distinguish a real improvement from a faster first draft.

How Acquivo Pro fits

Acquivo Pro is built around the production column. On one login the bank runs deterministic valuations across 21 industry verticals, generates teasers and CIMs from the client questionnaire and data room, builds a prospect list using documented buyer criteria, and tracks NDA, IOI, and LOI flow for every client in a single pipeline. Every figure traces to its source so the banker's edit is informed. The judgment column stays with the bank, which makes every decision and every buyer contact. To see it on a live mandate, request a demo through the Contact page.