AI in Sell-Side Discovery: Intake, Normalization, and Readiness

What discovery actually costs a small bank today

Discovery is the phase most boutique banks underestimate because nobody bills for it in their heads. A signed engagement letter arrives, the owner sends a zip file of tax returns, QuickBooks exports, a customer list in whatever format the bookkeeper used, and a handful of contracts. An analyst spends the first week figuring out what is there and what is not. The second week goes to rebuilding the income statement in the bank's template. The third week is spent emailing the owner for the things that turned out to be missing: the lease amendment, the customer concentration by year, the owner's actual compensation versus what the books show.

As an illustrative scenario, a ten-person bank running six mandates could have roughly one analyst's full attention tied up for the better part of a month on each mandate before valuation work starts. The hidden cost is not just hours. It is the delay. Every week spent in discovery is a week the market can move, the owner can lose conviction, or a buyer who was interested in the space can close something else. A managing director who signed the mandate may want to be in front of buyers in sixty to ninety days. Discovery done by hand can consume a material part of that window.

Intake: one request list, one upload, one place

The first thing AI-era tooling fixes is the request list. Most banks keep a master diligence checklist in a spreadsheet and copy it for each new client. The analyst sends it, the owner ignores half of it, and the next three weeks are a chase. The structural problem is that the request list and the document storage live in different places, so nobody can see at a glance what is outstanding.

A structured questionnaire changes this. The owner answers questions in sequence (ownership, revenue by line, customer concentration, employee count, facilities, debt, litigation, related-party transactions) and uploads supporting documents against each answer. The software knows which items are answered and which are not, so the analyst's chase becomes a status view rather than a memory exercise. The owner uploads a document once, and it is tagged to the question it answers. When the CIM needs a customer concentration table in week six, the data is already structured rather than buried in an email thread.

What the analyst does this week: take your current request list and rewrite it as questions the owner can answer in plain language, grouped by topic. Even without software, a request list framed as questions can produce more complete answers than one framed as document names.

Normalization: recasting the books without reinventing them

Lower-middle-market financials are rarely clean. Owner compensation runs through multiple lines. Personal vehicles, family members on payroll, one-time legal settlements, and a 2021 PPP forgiveness sit in the operating numbers. Normalization is the process of adjusting reported EBITDA to a figure a buyer can underwrite, with every adjustment documented. If an adjusted EBITDA measure is presented, the SEC describes EBITDA as earnings before interest, taxes, depreciation, and amortization and says it should be reconciled to GAAP net income, with enough detail for a reader to understand the reconciling items. [1]

AI can assist with the first pass. Given a trial balance and a general ledger export, software can map accounts to a standard chart, flag line items that look like add-back candidates (anything labeled ‘owner,’ ‘personal,’ or ‘one-time,’ unusual spikes in a single month, related-party vendor names that match the owner's last name), and produce a draft adjustment schedule. What it produces is a list of candidates with the source transaction attached, not a final answer.

The banker's job in normalization does not change. The associate still has to decide whether the $40,000 in ‘consulting’ paid to the owner's brother-in-law is a legitimate add-back or a cost the buyer will actually incur. The difference is that the associate starts from a schedule where every candidate is listed with its source, rather than from a blank spreadsheet and a 400-page ledger. That is the shift: from finding to deciding.

Readiness: knowing what is missing before you promise a timeline

The readiness question is simple. Can this company go to market in the next sixty days, or does it need ninety days of cleanup first? Most banks answer it by feel. The MD has seen enough companies to sense when the books are a mess, but the sense arrives late, usually after the engagement letter is signed and the timeline has been promised.

A readiness view makes the gaps explicit. It shows which questionnaire sections are complete, which documents are missing, whether the financials reconcile across tax returns and internal statements, and whether the customer, vendor, and employee data are in a state that will support buyer diligence. The output is not a score to be trusted blindly. It is a punch list the MD can review in ten minutes and use to have the hard conversation with the owner: ‘We can go to market in March if you get me reviewed financials by January, or we go in May with what you have.’

The data-room plan also needs to account for sensitivity. The FTC recommends tailoring information to the diligence need and stage, masking customer identities, and aggregating competitively sensitive information when appropriate, especially when a bidder could be a competitor. [2] That does not replace the MD's judgment or counsel's advice. It gives the bank another readiness question to answer before promising a timeline.

That conversation, had early, is the single highest-value thing in discovery. It sets the owner's expectations, protects the bank's credibility, and prevents the process from stalling in week ten when a buyer asks for something the bank should have known was missing in week one.

Where the analyst's time goes instead

When intake, normalization, and readiness are structured, the analyst's month in discovery may compress to something closer to a week of real work. That is a possible workflow outcome, not a measured result. The question is what to do with the rest.

The answer is analysis the bank never had time for. Cohort revenue by customer acquisition year. Gross margin by product line, not just in aggregate. The working capital trend that a sponsor's associate will ask about in the first management meeting. The honest read on whether the growth story holds up when you strip out the one large contract. These are the things that make a CIM credible and an IOI defensible, and they are exactly what gets cut when the analyst spends three weeks on data entry.

A managing director can test this in one mandate. Track where the analyst's hours go in discovery for the next engagement. If more than half are collection and recasting rather than analysis, that is the hour budget software should be absorbing.

How Acquivo Pro fits

Acquivo Pro structures discovery around a client questionnaire and a branded document portal where the owner sees requests and uploads each document once. The firm runs valuations across 21 industry verticals from that intake, and the same questionnaire and data room later feed the teaser and CIM, so nothing collected in discovery has to be re-entered in a later phase. Every client sits in a single pipeline on one login, which means the managing director can see readiness status across active mandates rather than asking analysts for separate updates. To see it on a live mandate, request a demo through the Contact page.