Due Diligence Checklist for Selling a Business: Build One With AI

A business broker can use AI to turn a seller's folder of documents into a due diligence checklist grouped by category, a data room index, a gap tracker and a log of buyer questions. Redact personal data before any upload, and have the seller's accountant and lawyer confirm the financial and legal items before buyers see them.

Buyers do due diligence to confirm what the listing claimed. Sellers who prepare the pack early face fewer surprises, and a broker who can show buyers an organised data room wins trust. The work is mostly sorting and chasing, which suits AI. The judgement about what a document means stays with people.

Follow me on Instagram@sagnikteaches

The checklist, category by category

Below is a seller-side checklist for a small business sale. Adapt it to the sector and the deal structure, and strike items that do not apply rather than leaving them blank. The "check" column says how you will know the item is ready.

Connect on LinkedInSagnik Bhattacharya

Financial records

Customers and revenue

• Top 20 customers by revenue, three years, with concentration percentages (anonymised for the early data room).
• Contracts or terms of business for the top customers, with renewal dates and change-of-control clauses.
• Revenue by product or service line, by month.
• Pipeline and order book, with dates.
• Returns, refunds and complaints log for the last 12 months.

Contracts and legal

• Supplier agreements, particularly any with exclusivity or minimum purchase terms.
• Commercial lease or property documents, with expiry dates and assignment rules.
• Licences, permits and certifications, with expiry dates.
• Litigation, claims and disputes: current, threatened and settled in the last five years.
• Insurance policies and claims history.
• Intellectual property: trade marks, domain names, software licences, ownership of any custom work.

People

• Organisation chart and headcount by role (names removed in the first release).
• Employment contracts for key staff, with notice periods and restrictive covenants.
• Pay, bonus and benefit summary, with accrued holiday liability.
• Contractor agreements, and any worker classification risk.
• Pending grievances, disciplinary matters or claims.

Operations, assets and compliance

• Fixed asset register, with condition and any assets under finance.
• Inventory count and valuation method.
• Key systems and software, with licence terms and who holds the admin logins.
• Health, safety and environmental records relevant to the sector.
• Data protection records: what personal data is held, where and why.

That is more than 40 line items for even a small company, which is why AI is useful for turning it into something trackable. For a sector-specific twist, selling an accounting practice and preparing its due diligence pack shows how the list changes when the asset is a client book.

Subscribe on YouTube@sagnikteaches

The main example: a craft brewery supplier

An illustrative case. A business broker is selling a craft brewery supplier: eight staff, a leased warehouse, about $3 million in annual revenue and one distribution contract that accounts for 22 per cent of sales. The seller has documents in three places: an accounting system, a shared drive with no consistent naming, and a former manager's email. The broker has six weeks before buyers get access.

The job: build the checklist, find what exists, index it, list what is missing and set up a way to log buyer questions. The steps below are in the order the broker would do them.

Step one: turn the generic list into this seller's list

Give the model the category list above and a short, anonymised business profile, and ask it to tailor the list. It is a drafting task, not a research task, so it should not invent items the business does not need.

Here is a generic seller-side due diligence checklist and a profile of the business
(a craft brewery supplier with 8 staff, a leased warehouse, one distributor providing
22% of sales, equipment held on finance). Produce a tailored checklist as a table with
columns: category, item, why a buyer will ask, document types that would satisfy it.
Add items specific to this business and mark them NEW. Remove items that clearly do not
apply and list them separately with a reason. Do not invent facts about the business.

A useful output adds rows such as "distribution agreement: term, exclusivity, termination and change-of-control clauses" (a buyer will go straight to the 22 per cent customer) and "equipment finance schedule: balance, early settlement terms". It might also add something generic, like "brand trade mark registration", that the broker confirms with the seller. Treat every NEW row as a question, not a fact.

Step two: the data room index and naming convention

A data room is a secure folder structure, usually in a purpose-built platform, where buyers review documents. The index is its table of contents. Decide a numbering scheme before uploading anything, so that every file name tells a buyer what it is.

01 Financial
   01.01 Annual accounts
   01.02 Tax returns
   01.03 Management accounts
02 Customers
   02.01 Top customer schedule
   02.02 Customer contracts
03 Contracts and legal
04 People
05 Operations and assets
06 Compliance

File name pattern: 01.02_TaxReturn_FY2024_v2_2026-10-20.pdf
Pattern: [index]_[Description]_[Period]_[version]_[date]

Ask the model to propose the index from your tailored checklist, and then to rename a batch of messy files. Give it filenames and short descriptions only, not the documents themselves:

Here is a list of 40 file names from the seller's shared drive. Map each to an index
number from the structure above and propose a new file name using the pattern. If a file
does not fit, mark it UNSORTED. If two files look like versions of the same document,
flag them as a possible duplicate. Output a table: old name, proposed index, new name, flag.

An illustrative result: "Final accts FINAL (2).pdf" becomes "01.01_AnnualAccounts_FY2024_v1_2026-10-20.pdf", flagged "possible duplicate of 'Accounts 2024 signed.pdf'". A person opens both files before deleting either, because the "final" one is sometimes the draft.

Step three: the gap tracker

Once files are mapped, compare the index against the tailored checklist. Each row becomes a tracker line with a status. The table lives in a spreadsheet the broker and seller both use.

Use AI to produce the first draft of this table from the index and the checklist, and to write the chasing messages. A sample message, with the "[first name]" placeholder left for the broker:

Hi [first name], we're building the buyer data room and still need three things from you:
1. The full warehouse lease (we only have page 1).
2. Signed contracts for two members of staff.
3. The lender statement for the equipment finance.
Could you send them by Friday 7 November? If any of these don't exist, tell me and we'll
note it rather than leave a gap. Thanks.

The last sentence matters. A document that does not exist is a finding to explain, not a hole to hide. Buyers forgive honest gaps more readily than missing papers they discover later.

Step four: redact before any AI tool or data room sees a file

Due diligence documents hold the most sensitive information a business owns: employee pay, customer terms, bank details, personal data. Three rules apply.

1. Sequence. No buyer sees anything before a signed confidentiality agreement. The early data room holds anonymised summaries; detail is released in stages as the buyer proves seriousness. The NDA and buyer qualification process covers that gate.
2. Redaction. Remove or mask names, national identifiers, home addresses, bank account numbers and salary details of individual employees before processing with AI or sharing in the early data room. Use summaries in place of full records: "Employee 3, operations manager, tenure 6 years, notice period 3 months".
3. Tooling. Use only an AI tool under a business agreement, and prefer working on file names, extracted fields and anonymised summaries over uploading the originals. See how to redact personal data from documents with AI for a safe routine, and check the result by eye: automated redaction misses things such as a name in a scanned signature block.

Step five: the buyer question log

Once buyers start reading, questions arrive by email, call and in the data room's own Q&A tool. A single log keeps answers consistent so that two buyers do not get different answers to the same question.

Ask AI to draft answers from documents already in the data room, to cite the file and clause behind each one, and to suggest the category. The seller or their adviser approves every answer. Add a column for "shared with", because fair dealing between competing buyers is easier to show with a log, and because an answer given to one buyer may need to reach all.

Using only the documents indexed below, draft an answer to this buyer question.
Quote the document reference and clause for each statement. If the documents do not
answer it, say so and list what we would need to find out. Do not estimate or assume.

Question: [paste]
Indexed documents: [paste index + extracted relevant clauses]

An illustrative good result states: "Clause 14 of the distribution agreement (02.02) requires the distributor's written consent to assignment. No consent is on file (gap 02.04). We need the seller to confirm whether it has been requested." That is the output of a tool kept to the documents, and it exposes a gap instead of covering one.

Scoring what buyers will push on, with a worked sum

A tracker tells you what is missing; it does not tell you what will cost the seller money. A second pass ranks the items buyers will treat as risks, so the seller can prepare an explanation before the question arrives. Ask the model to propose a risk score for each tracker row, then have the broker and the seller's accountant overrule it.

To see why this ranking matters, run an illustrative sum. Suppose the asking price is four times annual earnings of $450,000, which is $1.8 million. A buyer who discovers the distributor's consent clause late may argue that 22 per cent of revenue is at risk and propose cutting the price by a notional $150,000 or asking for part of it to be paid later, contingent on the consent. If the seller already holds the distributor's written consent in the data room on day one, the argument has far less to stand on. The same hour spent chasing that one letter protects more value than a week spent polishing the folder names.

Those numbers are made up to show the logic. Your accountant will know which figures a buyer applies and how deal terms work in your market, so involve them in deciding which risks to settle early.

Staged release: what each buyer sees and when

Buyers should not see everything at once. A staged release protects the seller if a buyer walks away, and AI can prepare each stage's version of a document.

Ask the model to prepare the stage-two version of each document, for example turning a customer list with names into "Customer 1: drinks distributor, 22 per cent of revenue, written agreement with a change-of-control clause". A person verifies that the anonymisation does not leave an identifying detail such as a unique location or product. The anonymous teaser template applies the same thinking to the first stage.

The same checklist on a different business: a specialist food importer

To show how the list bends, consider an illustrative specialist food importer with 14 staff selling to restaurants and shops. Several categories change.

• Compliance becomes central. Buyers will ask for food safety audits, supplier approval files, product traceability records, import documentation and any recall history. These are rows the craft brewery example barely needed, so the model's tailoring step should add them.
• Inventory is large and perishable. The stock valuation method, shelf-life policy and write-off history matter. Ask AI to draft questions a buyer would ask about stock ageing, then the seller's accountant answers them with figures.
• Supplier concentration replaces customer concentration. If one overseas producer supplies 40 per cent of goods, a buyer will want the supply agreement, the relationship history and any exclusivity terms.
• Currency exposure. Buying in one currency and selling in another produces margin swings. A buyer will want a schedule of hedging or pricing policy, and a clear explanation of any year where margins moved sharply.

The lesson for brokers is that a template is a start. Spend the first hour of each engagement asking the owner what keeps the business running and where it would break, then tailor the list to those points before AI touches it.

A disclosure note for known problems

Every business has issues. A lease close to expiry, a customer dispute, a missing licence renewal. Good practice is to disclose known problems with the supporting documents rather than wait for a buyer to find them. Ask the model to draft a short, neutral disclosure note per issue from facts the seller provides, and have the lawyer settle the final wording, since what is said in diligence can affect later claims.

Draft a neutral disclosure note (80 words maximum) from these facts. State what the
issue is, what the current position is and what documents are in the data room. No
opinion, no minimising words like "minor", no promises about outcome.

Facts: [seller's facts, anonymised]

An illustrative result: "The warehouse lease expires in January 2028. The landlord has not yet confirmed renewal terms. The lease (03.04) and the correspondence of 14 October (03.05) are in the data room." It is dull, which is the point. The reviewer's job is to cut anything that sounds like persuasion, because disclosure notes work best when they read as plain records.

Measuring whether the data room works

Three simple figures tell you if the preparation paid off. Count the buyer questions that could have been answered from a document that was already in the data room: that number should fall to near zero after the dry run. Track the days between each buyer's access date and their first request for something missing. And measure how long it takes to answer a question, from receipt to approved answer; if it is longer than two working days, the approval step is the bottleneck, not the drafting.

What AI misses or gets wrong in this work

• Silent assumptions. It may state that a lease "can be assigned" when the clause says assignment needs the landlord's consent. Always read the clause.
• Numbers that do not tie. It will happily summarise accounts that disagree with the tax return. Reconciliation is the accountant's job; ask AI to list differences, not to explain them away.
• Wrong version. It indexes the draft contract because the file was named "final". Check dates and signatures.
• Over-disclosure. It suggests putting full employee schedules in the first release. The staging rule overrides it.

Timing the six weeks

1. Week 1: tailor the checklist, agree the index, collect everything into one holding folder.
2. Week 2: rename and index, run the first gap tracker, send the first chasing messages.
3. Week 3: redact and prepare the early release, accountant reconciles the financial schedules.
4. Week 4: lawyer reviews contracts and the lease, flags change-of-control and assignment clauses.
5. Week 5: dry run. Someone not involved opens the data room as a buyer and tries to answer ten set questions from it alone.
6. Week 6: open to the first cleared buyer, start the question log.

The dry run in week five finds the real problems, such as a missing exhibit, a broken folder permission or a confusing file name. Fix them before a buyer does. A confidential information memorandum workflow sits upstream of this and should state nothing that the data room cannot back up. If you would rather set this up with someone, that is what my AI implementation consultation covers.

Questions sellers and brokers ask about the checklist

When should a seller start preparing due diligence documents?

Before the business is listed, ideally three to six months ahead. Gathering contracts, tax returns and records takes longer than expected, and a prepared data room shortens the buyer's review and reduces the chance of a price chip late in the deal.

Can I upload the whole data room to an AI tool?

Not as it stands. Leases, employee records, customer lists and tax files hold personal and commercial secrets. Redact personal data first, use a tool covered by a business agreement, and only upload what the task needs. Never upload before the buyer has signed a confidentiality agreement.

What is a gap tracker?

A table listing every checklist item, its status (received, partial, missing, not applicable), the owner who must supply it and a due date. It shows the broker and seller at a glance what is still holding up the buyer's review.

Does AI replace the accountant and lawyer?

No. AI sorts, indexes and drafts. The accountant confirms the financial schedules and the lawyer reviews contracts, leases and disclosures. Treat AI output as a working draft that a qualified person checks before anything reaches a buyer.

Further reads

• Seller Discretionary Earnings With AI: Recast and Find Add-Backs — Recast the financials and document every add-back.
• AI Due Diligence on Real Estate: Read Leases, Surveys, Searches — The buyer side: reading leases, surveys and searches.
• Estate Agent AML Checks: Organise Due Diligence Files With AI — Organising due diligence files with AI in another trade.
• AI Readiness Checklist for Accountants, Solicitors, Consultants — Check your firm is ready to use AI on client files.
• Business Broker CRM With AI: What to Compare Before You Buy — Tracking buyers and deals in a CRM.
• Business Broker Marketing: AI Buyer Outreach and Anonymous Listings — Buyer outreach and anonymous listings.
• Do AI Business Valuation Tools Work? What They Miss for Brokers — Test valuation software against three closed deals and identify the facts a quick model cannot see.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general M&A practice for small business sales; no vendor pricing is quoted. Have the seller's accountant and lawyer confirm what each category must contain for the deal.