Tamara Ashworth
All postsAI x Business Acquisition

How I Use AI to Pressure-Test a Seller's Add-Backs Before I Make an Offer

How I use AI to pressure-test a seller's add-backs and seller's discretionary earnings before I make an offer: the add-back categories, the evidence I demand for each, the red flags AI surfaces, and where the final call stays mine.

· 14 minute read · By Tamara Ashworth
How I Use AI to Pressure-Test a Seller's Add-Backs Before I Make an Offer feature image

Short answer: Before I make an offer on a small business, I use AI to sort every add-back the seller claims into categories, match each one to a document that proves it, and recalculate seller's discretionary earnings (SDE) using only the add-backs that survive. AI does the sorting, the matching, and the arithmetic across three years of statements. I decide which add-backs are real, which are wishful, and what the business is actually worth to me. A listing might advertise $600,000 of SDE while the defensible number is $485,000, and at a 3x multiple that gap is a $345,000 swing in price. The add-back review is where I find it, and I run it before I ever put a number in a letter of intent.

Key Takeaways

  • SDE is a seller-built number. Add-backs are the adjustments that turn reported profit into the figure on the listing, and every dollar of them gets multiplied by the asking multiple.
  • An add-back is only as good as its evidence. If I cannot tie it to a bank statement, a payroll record, a contract, or a receipt, it does not count until the seller proves it.
  • AI is excellent at categorizing add-backs, matching them to documents, and flagging patterns that repeat year after year. It is not the one who decides whether an add-back is reasonable.
  • Recurring one-time expenses are the most common trap. A cost that shows up every year is a cost of doing business, whatever the seller calls it.
  • Running the review before the LOI lets the corrected number shape the price and the structure, instead of renegotiating after you have anchored.

Why Add-Backs Are the First Place I Look

When a broker sends over a listing, the number that sells it is almost always seller's discretionary earnings. It is a reasonable idea. A small business is usually run by an owner who pays themselves in a mix of salary, perks, and profit, so reported net income understates what a new owner could take home. SDE tries to fix that by adding back the owner's pay and any expenses that would not continue under new ownership.

FREE CHECKLIST

Set up an AI chief of staff that does real work for you.

Get the AI Chief of Staff Checklist, plus Tuesday tutorials for running real work with AI agents.

Unsubscribe anytime.

The problem is that every add-back is a judgment call, and the person making the judgment is the one selling. Nobody lies, usually. The seller honestly believes the family cell phone plan is a business expense they will not need to keep paying, or that last year's roof repair will never happen again. But each of those beliefs gets multiplied by the asking multiple, so a $40,000 stack of generous assumptions becomes $120,000 or more in purchase price.

That is why I look at add-backs before I look at almost anything else. The revenue is what it is. The add-backs are where the story gets told, and a story deserves to be checked.

What an Add-Back Actually Is

An add-back is an adjustment that increases reported earnings by removing an expense that supposedly will not continue, or by normalizing an expense that was unusually high or low. In plain terms, the seller is saying: this cost is in the books, but it will not be there for you, so the business earns more than it looks.

Most add-backs fall into a handful of families, and recognizing the family tells you how hard to push.

The family matters because the evidence changes. A non-cash item is arithmetic. A personal expense is a claim. A one-time cost is a prediction. I trust them in that order.

How I Use AI to Sort and Match Every Add-Back

The first pass is mechanical, and this is where AI saves me the most time. I ask the seller for three years of profit and loss statements, the general ledger or a transaction export, bank statements, payroll summaries, and the add-back schedule the broker prepared. Then I give the AI the raw material and ask it to do four things in order.

  1. Build one list. Pull every claimed add-back from the schedule, the broker memo, and the P&L notes into a single table with the year, the amount, the stated reason, and the source document.
  2. Assign a family. Label each item as owner compensation, personal expense, one-time cost, non-cash, or related-party arrangement.
  3. Find the matching transaction. Search the ledger and bank data for the actual entries behind each claim. Flag any add-back that has no matching transaction, or where the amount does not agree.
  4. Check for repeats. Compare the three years and flag any add-back that appears in more than one, or that grows over time.

This used to be a long evening with a spreadsheet and a pot of coffee. With AI it is a few minutes of setup and then a review of the output. The point is not speed for its own sake. The point is that I can run it every time, on every serious deal, so I am never tempted to skip the step because I am tired or excited about the business.

One rule I hold to: I never let the AI decide that an add-back is valid. I ask it to show me what supports the claim and what contradicts it. The conclusion is mine.

The Evidence I Require for Each Category

An add-back with no evidence is an opinion. I keep a short standard for what counts as proof, and I give that standard to the AI so it checks every item the same way.

Add-back familyEvidence I requireWhat makes me discount it
Owner pay and benefitsPayroll records, tax forms, benefit invoicesNo allowance for the cost of a replacement manager
Personal expensesStatements showing the charge, a clear business-versus-personal splitRound numbers, no receipts, a story that changes between calls
One-time costsInvoice, contract, explanation of what happened and why it will not repeatSimilar costs in other years, or a cause that still exists
Non-cash itemsDepreciation schedule, loan statementsRarely an issue, but confirm they tie to the tax return
Related-party dealsLease, market rent comparison, payroll for the relativeRent or pay far from market, no written agreement

When the AI finds an add-back that does not meet the standard, it goes on a separate list I call the open items. Those are not rejected. They are simply not counted until the seller closes them with documents. That framing keeps the conversation cooperative, because I am not accusing anyone of anything. I am asking for the page that proves the number.

The Red Flags AI Catches That I Would Miss

The real value of running this with AI is not the sorting. It is the pattern recognition across a lot of boring data. A few things come up again and again.

None of these automatically means the seller is dishonest. Small business books are often messy, and bookkeeping shortcuts are common. What each one means is that the SDE on the listing is less certain than it looks, and I should price that uncertainty.

The Replacement Cost Question Most Buyers Skip

This is the add-back mistake I see most often, and it has nothing to do with documents. A seller adds back the full owner salary and calls it a day. But if the owner is actually running the business, someone has to do that job after closing. Either you do it, or you hire someone to do it.

If I plan to run the business myself, I still want to know what the job is worth, because my time has a price. If I plan to hire a manager, the cost of that manager is a real expense that has to come back out of the earnings. Either way, the owner's full compensation is not free money to a new buyer.

I ask AI to help me build the replacement picture. I give it the owner's description of their week, the org chart, and the job responsibilities, and have it draft a role description with a realistic market salary range for the area. Then I subtract that cost from SDE and see what the business earns before an owner takes anything. That number, sometimes called adjusted EBITDA, is often the honest one, and it is usually lower than the headline.

A Worked Example

Here is a made-up example, not a real deal, to show how the math moves. Say a listing claims $600,000 of SDE with an asking multiple of 3x, so an asking price of $1,800,000.

The AI builds the add-back table and finds the following.

The recalculated figure lands closer to $485,000 before the replacement-cost adjustment, and lower after it. At the same 3x multiple, the supportable price is about $1,455,000 before replacement cost. That is the gap between a deal that works and one that quietly does not.

Notice that none of this is dramatic. No fraud, no villain. Just a stack of reasonable-sounding assumptions that each need a document. That is exactly why I check them.

How the Corrected Number Changes the Deal

Finding a lower SDE does not mean I walk away. It means I now have a more honest basis for the conversation, and I can use it three ways.

This is also why I run the review before I draft the letter of intent. Once a number is written down and signed, every change feels like a retrade. Done earlier, the same adjustment feels like diligence.

Where the Judgment Stays Mine

AI makes this review faster and more consistent, but it cannot do the parts that matter most.

I treat AI as the analyst who prepares the work, and myself as the person who signs the offer.

A Practical Add-Back Review Checklist

This is the order I follow. You can copy it and adjust it to your own deals.

  1. Collect three years of P&Ls, the general ledger export, bank statements, payroll summaries, tax returns, and the seller's add-back schedule.
  2. Have AI build a single table of every claimed add-back with year, amount, stated reason, and source.
  3. Label each item by family: owner pay, personal, one-time, non-cash, or related-party.
  4. Match each item to ledger and bank entries. Flag anything with no match or a different amount.
  5. Compare across years. Flag any add-back that repeats, grows, or changes description.
  6. Move unsupported items to an open list and request specific documents for each.
  7. Reverse related-party arrangements to market rates, in both directions.
  8. Build a replacement cost for the owner's role and subtract it from SDE.
  9. Recalculate SDE and adjusted earnings using only supported items.
  10. Compare the corrected figure to the listing and decide how it should change price or structure.
  11. Note anything that needs a professional, such as a quality of earnings review.

The whole review, from raw documents to a corrected number, now takes me a day or two instead of the week it used to. That speed means I can do it before I get attached to a deal, which is exactly when it is most useful.

FAQ

What are add-backs in a small business sale?

Add-backs are adjustments a seller makes to reported profit to show what the business would earn for a new owner. They remove expenses that supposedly will not continue, such as the owner's salary or personal costs run through the company, and they normalize unusual items. The resulting figure is usually called seller's discretionary earnings, and it is the number the asking price is built on.

Which add-backs do buyers usually dispute?

The most disputed are personal expenses with no receipts, one-time costs that appear in more than one year, family members on payroll, and the full owner salary with no allowance for replacing the owner's work. Related-party rent that is below market is another common one, because the new owner will likely pay more.

How does AI help review add-backs?

AI does the assembly and the cross-checking. It builds one table from several documents, categorizes each add-back, searches the ledger and bank data for matching transactions, and flags repeats and mismatches across years. It does not decide whether an add-back is valid. That judgment stays with the buyer and their advisers.

Should I still hire an accountant for a quality of earnings review?

On larger deals, yes. AI helps you arrive prepared, spot the weak points, and ask sharper questions, but a qualified accountant can verify the underlying records and give lenders and sellers a credible third-party view. I use AI to make that review cheaper and more focused, not to skip it.

When should I run the add-back review?

Before you sign a letter of intent. The corrected earnings should shape your price and structure from the start. Once an offer is signed, changing the number feels like a retrade, and the seller is far less receptive.

Add-backs are not a trick, and most sellers are not trying to fool you. But the earnings number on a listing is a story told by the person who benefits from it, and a story deserves to be checked before you pay a multiple of it. A careful review turns an uncertain figure into one you can defend, and it turns a nervous negotiation into a calm one.

If the file is still a pile of exports, start with how I turn messy seller financials into a clean underwriting file. Once the corrected SDE is in hand, the next page I open is how I use AI to value a small business before I make an offer, and then how I structure the LOI so the price matches the number that survived. The full pass after a signed LOI is the AI due diligence checklist.

If you are working through an acquisition and want help building the AI workflows that make this kind of diligence fast and repeatable, that is the kind of work I do with a small number of clients. You can request a strategic AI consulting conversation and we can talk through what fits your deal.

Current Search Intent Check

Recent Search Console data shows people arriving through "how can small business owners use ai to get practical guidance when planning their next steps?". That changes the bar for this post: it needs to answer the operator question directly, name the workflow being improved, and give the reader a practical decision rule instead of another broad AI opinion.

Recent Search Console data shows people arriving through "small business acquisition due diligence checklist". That changes the bar for this post: it needs to answer the operator question directly, name the workflow being improved, and give the reader a practical decision rule instead of another broad AI opinion.

Want more of my writing in your Google results? Add Tamara Ashworth as a preferred source.

FREE CHECKLIST

Set up an AI chief of staff that does real work for you.

Get the AI Chief of Staff Checklist, plus Tuesday tutorials for running real work with AI agents.

Unsubscribe anytime.

WANT IT BUILT?

Tell me where growth is getting stuck.

Book a conversation →