How to Get Your Records Ready to Sell Your Business, With AI

Start 12 to 24 months before you plan to sell. Reconcile three years of accounts and your monthly management figures, list every one-off or personal cost a buyer should add back, gather contracts, leases, staff terms and supplier agreements into one indexed data room, and fix the gaps a buyer's checks would find. AI speeds up indexing, summarising and gap-spotting.

Buyers pay for profit they can verify. Every number you cannot back up becomes a reason to cut the price, hold back part of it, or walk away, and records are often weakest where the owner has done the books alone for years. Your accountant and a broker or lawyer still lead the valuation and the deal; AI's job is to get you ready for them.

Follow me on Instagram@sagnikteaches

What a buyer will ask to see

A buyer's due diligence (the checks they run before completing) arrives as a long request list. For a small business it usually covers six areas. Knowing the list in advance is the whole trick, because you can work through it at your own pace instead of in the middle of a negotiation.

Connect on LinkedInSagnik Bhattacharya

• Financial: annual accounts, monthly management accounts, tax filings, bank statements, debtor and creditor lists, loans and any personal guarantees.
• Commercial: sales by product, channel and customer, top suppliers, pricing and margins, marketing costs and results.
• Legal: leases, supplier and customer contracts, licences, trademarks and domain names, any disputes or claims.
• People: staff list with roles, pay, hours, contracts, holiday owed, pension or benefit arrangements, any grievances.
• Operations: stock records and valuation method, key processes, equipment and its condition, insurance policies and claims history.
• IT and data: systems and subscriptions, who owns each account, customer data and how it was collected, website and social accounts.

Build the data room index before collecting anything

A data room is simply a well-organised, access-controlled set of folders a buyer's advisers can work through. A shared drive with restricted access is enough for most small sales; specialist virtual data room services exist for bigger deals. Start with the structure, because it tells you what is missing.

Subscribe on YouTube@codingliquids

The illustrative business here is an independent toy shop with two shops and a website, whose owner plans to retire in about 18 months. Her files were spread across a laptop, a shared drive, an email archive and two filing cabinets. She exported a list of every file name and folder path from the shared drive and laptop (no contents, just names and dates) and gave it to an AI assistant with this prompt:

Below is a list of file paths from a small retail business preparing for sale.
Map each file to one folder in this data room index:
  01 Financial / 02 Commercial / 03 Legal / 04 People / 05 Operations / 06 IT and data
  (with the sub-folders listed below).
Return CSV: file_path, suggested_folder, confidence, note.
Then list every sub-folder that has NO matching file, as "Possibly missing".
Flag duplicates and files that look like old drafts.

Index: [paste index with sub-folders]
File list: [paste]

An illustrative part of the "possibly missing" list that came back:

Possibly missing:
- 01.3 Monthly management accounts: only 2024 and 2026 found, nothing for 2025
- 03.1 Leases: lease for Shop 2 found; no lease or licence found for Shop 1
- 03.4 Supplier agreements: 3 of the top 10 suppliers by name have no agreement file
- 04.2 Employment contracts: 5 contracts found; staff list shows 9 employees
- 05.4 Insurance: current policy found; no claims history
- 06.2 Domain and hosting: no record of who owns the domain name

This list became the project plan. Some items existed on paper (the Shop 1 lease was in a filing cabinet), some needed chasing, and some did not exist at all: four staff had never been given written contracts, which is the kind of thing that turns up in due diligence and makes a buyer nervous. Treat the AI's mapping as a first pass; it placed a supplier's price list in the Legal folder because the file was called "terms", which a quick review fixed.

Make the numbers stand up, 18 to 12 months out

Financial records are where most value is won or lost. Three jobs matter most.

Reconcile everything to the bank

Every month's figures should tie to bank statements, and the management accounts should add up to the annual accounts. If month-end has been a scramble, fix the routine now so there are a year of clean monthly figures before the sale starts; speeding up month-end close with AI covers the routine. AI can help find where things do not tie: give it the monthly totals and the annual accounts and ask which months or categories account for the difference.

Separate the personal from the business

Owner-run businesses often carry the owner's car, phone, a family member on the payroll or a personal subscription. Buyers will not pay for costs that disappear when you leave, but they also will not accept a claim that costs are personal without evidence. Start separating them now.

List your add-backs, with evidence

Add-backs are costs in the accounts that a buyer would not bear, added back to show the profit the business really makes for a new owner. The toy shop's illustrative schedule:

Reported profit of $96,000 plus $38,000 of add-backs gives an adjusted profit of $134,000. Expect every line to be challenged. A buyer's adviser might accept the car and the legal costs, but argue that shops need refitting regularly so the refit is really a recurring cost spread over years. If that add-back is rejected, adjusted profit drops to $123,000. The lesson for the schedule: only include add-backs you can evidence and defend, and ask your accountant which ones buyers in your sector usually accept. AI is useful for drafting the explanations and gathering evidence lists; it should not decide what counts.

Stock is the number buyers argue about in retail

In a toy shop, stock is often the biggest asset on the balance sheet and the most disputed. The buyer wants to know three things: how stock is valued, whether the count is right, and how much of it will actually sell.

The illustrative shop held about $210,000 of stock at cost. Running the sales history against the stock list showed that about $34,000 of it had not sold a single unit in 12 months: last year's licensed characters, discontinued board games, a line of kites. A buyer would either refuse to pay for it or knock it off the price anyway. Clearing it over the months before the sale, through a sale event and an end-of-line eBay shop, turned slow stock into cash rather than a price reduction. The approach is in finding dead stock and slow movers with AI.

Plan for a full stocktake close to completion, because the final price is often adjusted for the stock actually on hand. A stocktake a few months earlier, with the method written down, lets both sides agree how the completion count will work before it matters.

Contracts and leases: find the clauses that bite on a sale

A buyer's lawyer will read every material contract, looking for clauses that change or end when the business changes hands. You want to know about them first. AI is good at turning a stack of contracts into a comparable table, as long as you check its work. Upload each contract (on a business plan, or with training switched off) and ask:

Summarise this contract in one row with these columns:
parties; start date; term and renewal; notice period to end it; can it be
transferred to a buyer (assignment clause, quote it); does anything change if
ownership changes (change of control, quote it); exclusivity; personal guarantees;
anything unusual. Quote the clause number for every answer. If a point is not in the
contract, write "Not found" rather than guessing.

Illustrative rows for three of the toy shop's contracts:

The supplier row is the important one: the shop's exclusive right to sell a popular wooden toy brand in its area could disappear on a sale. That is worth a conversation with the supplier well before a buyer finds it. The personal guarantee on the lease matters to the owner too, because she will want to be released from it.

A realistic failure is worth knowing about. On the first pass, the AI reported "Not found" for change of control in the wooden toy agreement, because the relevant clause sat in a later amendment letter saved as a separate file. The owner only caught it because the index listed the amendment. Always give the AI every document that forms a contract, and have your lawyer confirm the summary for anything material. Summarising long documents without missing details has more on keeping these summaries honest.

People, and how much the business depends on you

Buyers look hard at owner dependency: if the owner does all the buying, knows every supplier and closes every big sale, the business is worth less without her. Two things reduce that risk.

• A complete staff file. A current staff list with roles, pay, hours, start dates, holiday owed and a signed contract for each person. The four missing contracts at the toy shop needed issuing properly, with advice, well before any sale conversation.
• Written processes. The owner recorded herself doing the seasonal buying, the supplier ordering and the website updates, and used AI to turn the recordings into step-by-step procedures. The shop manager then did the spring buying with the owner watching. A buyer can now see that the process lives in the business, not in one head. The method is in writing SOPs with AI.

Customers, suppliers and recurring revenue

Buyers also look at concentration: how much of your income depends on a few customers, and how much of your supply depends on a few suppliers. A toy shop has thousands of small customers, so its risk sits on the supplier side, which the contract table already showed.

A subscription box company is the opposite case and shows how different businesses need different evidence. Its value rests on recurring revenue, so a buyer will want monthly recurring revenue (the income from active subscriptions each month), churn (the share of subscribers who cancel each month) and cohort retention (how many subscribers from each sign-up month are still paying). An illustrative cohort table it might prepare:

The July row shows a weaker cohort, which a buyer will ask about (it followed a discounted first-box promotion). Having the explanation ready, with the data behind it, is far better than having it discovered. These figures must reconcile to payment processor reports and the bank; a buyer will check. Running a churn analysis with AI shows how to produce them.

Answering buyer questions quickly and accurately

Once a buyer's advisers are in the data room, questions arrive in batches. Keep a question log: question, date received, who answers, the answer, the documents it relies on, date sent. AI can draft answers from the data room documents, which saves hours, but every answer needs checking because buyers will rely on what you say, and wrong answers can come back later as claims under the sale agreement.

A good pattern is to ask the AI to answer only from the documents you give it and to cite the file for every statement. For example, to the question "Are there any disputes with landlords?" the draft might read: "The Shop 2 lease dispute over the service charge was settled last year (see 03.5 Settlement letter). There are no current disputes." The owner checks the settlement letter, confirms nothing new has arisen, and sends it. Where the AI cannot find a source, it should say so, and you answer from your own knowledge with your adviser.

You can rehearse this before any buyer appears. Ask your broker or accountant for a typical due diligence request list for your sector, give it to the AI with your data room index, and ask: "For each request, name the file that answers it, or mark it UNANSWERED." At the toy shop's rehearsal, six months before marketing the business, 11 of about 90 requests came back unanswered. Most were quick to fix, such as the last three years of insurance claims history and the domain registration record, but two (a missing licence agreement for the till software and an unwritten arrangement with a freelance delivery driver) took weeks. Finding them in a rehearsal cost nothing; finding them during a sale would have cost time and trust.

Keeping it confidential while using AI

• Use a business plan (such as ChatGPT Business or Claude Team, which do not train on business content by default) or switch off the model-training setting on an individual plan before uploading anything.
• Use a codename for the sale in file names, folder names and prompts, so a glimpse of a screen or a synced file does not start rumours.
• Strip personal data from staff and customer files before AI summarising where you can; redacting personal data before sharing documents covers the method. Buyers usually see anonymised staff lists until late in the process anyway.
• Give buyers data room access only after a confidentiality agreement is signed, and give view-only access with downloads switched off where your tools allow.

A timeline to work back from

For the toy shop, the preparation took about 60 hours of the owner's time spread over a year, much of it chasing paper. The AI work (mapping the file list, summarising 22 contracts, drafting procedures and answer drafts) took perhaps 10 of those hours and replaced several times that in reading and typing. The rest was decisions only she could make, which is exactly where her time should go.

Questions owners ask before selling

How many years of records will a buyer want?

Commonly three years of annual accounts plus monthly management figures for the current year, though buyers and lenders vary. Tax filings, payroll records and bank statements for the same period are usually requested too. Ask your accountant or broker what buyers in your sector typically ask for, and start filling gaps from the oldest year first, because old gaps are hardest to fix.

Can AI tell me what my business is worth?

It can explain how valuation methods work and help you prepare the figures they use, such as adjusted profit or recurring revenue. It cannot give a reliable price, because value depends on the buyers in your market, recent comparable sales, deal terms and how risky your business looks. Use a broker, an accountant with sale experience or a valuation specialist for the number.

When should I tell staff I plan to sell?

Usually later than you tell your advisers, and on advice. Early news can unsettle staff and customers, while some legal systems require you to inform or consult employees before a transfer of the business. Keep preparation work confidential by using codenames and restricted folders, and ask your lawyer when and how staff must be told where you operate.

Further reads

• How to Automate Monthly Management Reports With AI — Produce the monthly figures buyers ask for without the scramble.
• How to Check Your Margins Product by Product With AI — Show buyers margin by product, not just in total.
• Can AI Review a Contract? What Small Business Owners Should Know — What AI contract review can and can't be trusted with.
• Is Your Business Data Ready for AI? A Clean-Up Checklist — Clean up business data so the analysis stands up.
• How to Create SOPs From Screen Recordings With AI — Document the tasks only you know how to do.
• How to Track Contract Renewals and Deadlines With AI — Keep renewal dates visible while a sale is in progress.
• Reduce Owner Dependency: Use AI to Capture What Only You Know — Get what's in your head into a form staff can use: an interruption log, an AI interview prompt, decision rules with thresholds, and a shared assistant.
• How to Prepare a Business Loan Application With AI Help — Build a traceable loan evidence pack, test a weak trading month and use AI to draft explanations without inventing financial claims.
• Can AI Do My Business Taxes? What It Can Safely Prepare — What AI can safely prepare for your business tax return, where its tax answers go wrong, and a year-end pack you can build before the accountant sees it.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general due diligence practice for small business sales; ChatGPT and Claude business plan data settings. No valuation multiples are given because they depend on sector and market.