AI Franchise Due Diligence: A Step-by-Step Workflow

Good franchise due diligence is mostly reading and phone calls. AI can take over the reading, pulling fees, earnings, and closure data across dozens of brands in the time it used to take to open one PDF. What it can't take over is the judgment and the conversations. Here's a workflow that uses AI for the grunt work and keeps you in charge of the decision.

The structure is simple: screen wide, then go deep, then verify with humans, then have a lawyer check the contract. AI carries the first two stages and assists the third.

Stage 1: Screen wide

The goal here is to kill bad options fast. Most franchises you look at won't survive a basic financial screen, and you want to find that out in minutes, not after a 300-page read.

Pull these for every brand on your list:

• Total investment (Item 7, low and high)
• Franchise fee and royalty (Items 5 and 6)
• Whether Item 19 exists and the median figure if it does
• Net unit growth (Item 20: did more units open or close last year?)
• Litigation and bankruptcy (Items 3 and 4)

With AI connected to a franchise database, this is one question: "Compare these brands by investment, royalty, median Item 19 revenue, net unit growth, and litigation count." You get a table. Anything that flunks (no earnings disclosure, shrinking unit count, a stack of franchisee lawsuits) comes off the list.

A blunt rule that holds up: cheaper isn't safer. We've seen it in the loan data. When we analyzed SBA outcomes, the high-default franchises had lower average investment than the low-default ones. A small franchise fee tells you nothing about whether the business works.

Stage 2: Go deep on the survivors

Now you've got three or four brands worth real attention. Spend it.

For each, look harder at:

• Item 19, average vs. median. A franchise can advertise a big average that a handful of top units inflate. The median is the typical owner's reality. If the gap is wide, ask why.
• Item 20 trends over three years. One bad year happens. Three years of net closures is a pattern. The FDD gives you three fiscal years of opened, closed, transferred, and ceased counts, so use all of it.
• SBA loan performance. Real loans, real outcomes. A high charge-off rate means lenders have watched a lot of these businesses fail. That's market evidence the franchisor can't spin.
• Fee stack. Item 6 hides the recurring costs: ad fund, tech fees, required purchases. A modest royalty plus a heavy fee stack can quietly change the math.

This is where a database connected to AI earns its keep. You ask follow-ups in plain language and get numbers back, instead of cross-referencing tables by hand. Comparing franchises with AI covers this stage in more detail.

Stage 3: Talk to real franchisees

No AI replaces this. The most honest picture of a franchise comes from people running one.

Item 20 exists partly for this reason: the FDD must list current franchisees with contact info, plus former franchisees who left. Call both. The franchisor will happily point you to its three happiest owners. The names the franchisor didn't hand you, and the ones who quit, are where the truth lives.

What to ask: Did your real costs match Item 7? Does your revenue look like Item 19? How responsive is corporate? Would you do it again? The former franchisees especially tend to tell you things no disclosure document will.

If you've got the franchisee contact list pulled already, this stage is a series of phone calls instead of a research project. Finding franchisee contacts for validation calls walks through it.

Stage 4: The legal review

You've found a brand the numbers like and the owners back. Now, before you sign or pay anything, a franchise attorney reads the agreement.

This is non-negotiable, and it's the one stage where AI's "mostly right" isn't good enough. The clauses that decide your downside risk are legal, not numerical:

• Item 17: renewal, termination, transfer, dispute resolution. Can you sell the business? Can they end your contract, and on what terms?
• Item 12: territory. Is it protected, or can the franchisor open a unit down the street?
• Item 8: sourcing. Are you required to buy from the franchisor at prices it sets?

Use AI to flag what's in these sections so you walk into the lawyer's office informed. Don't use it to skip the lawyer.

The division of labor

The whole point of AI in due diligence is to move the bottleneck. The old bottleneck was reading: hours per FDD, multiplied by every brand you considered, which meant most people only seriously looked at one or two. The new bottleneck is judgment and conversations, which is where it should be.

Let the machine read. Call the owners yourself. Pay the lawyer. That's the workflow, and it scales to as many franchises as you're willing to consider.

Want the data layer for all of this inside ChatGPT or Claude? See how the AI integration works.