Startup Miracle logo
Startup Miracle
← Back to all posts

Blog

Franklynn Automotive's AI Agent Drafted Its Card Fee Ask

Javier Aguilera·Oct 28, 2026auto repairback officeAI agentsvendor bills
Franklynn Automotive's AI Agent Drafted Its Card Fee Ask

Franklynn Automotive's AI Agent Drafted Its Card Fee Ask

On October 21, 2026, Dan Mallatt, owner of Franklynn Automotive, posted on X that his AI agent had negotiated his shop's credit card processing fees. Processing fees are what a business pays to accept card payments. In Mallatt's post on X, he says it's saving about $5,000 a year on about $2 million a year in card payments, his card volume. Card volume is the total dollars customers pay the shop by card. Those are his own numbers, self-reported, from one shop.

Our view, plainly: this is a back-office win on a bill that's easy to pay without reading, and the AI did the homework that usually keeps an owner from pushing back. If you run a repair shop, the lesson isn't that you'll save the same amount. It's that a recurring bill can be worth a careful look, and that you should get a few things ready before you hand that look to an AI agent.

The tool and how the work was split

According to a write-up by AI Business Plays on how Mallatt negotiated his fees, the tool he used was Grok Bot. In plain terms, an AI agent is a program that works through a chain of steps by itself and passes what it produced back to you for a decision.

Per the write-up, Mallatt gave the agent 24 months of processing statements. A card processing statement is the monthly bill that lists what the shop paid to accept cards. The agent found the markup, the charge the processor (the company that handles the shop's card payments) adds on top of the fees set by card networks such as Visa and Mastercard, and built the case on his volume. It also drafted every email and the counteroffer.

Mallatt reviewed each email before it went out, according to the write-up. That detail matters as much as the savings.

Notice what the agent didn't do in this account. It didn't pick a new processor, sign anything, or change how the shop takes payments. It read, it found one line worth asking about, and it wrote drafts. The decisions stayed with the owner, which is the split we'd want to see on any money task.

Where the 0.35% sat on the statement

The write-up describes the change in terms of markup over interchange, so it helps to know the two parts. Interchange is the fee set by the card networks on each card payment. The write-up notes it isn't negotiable.

The processor also sends the monthly statement. Markup over interchange is what the processor charges on top of interchange for its own service. That's the part a shop can ask about.

According to the AI Business Plays write-up, Mallatt's processor markup over interchange dropped from 0.35% to 0.10%. The write-up says that figure isn't in the text of the X post itself, so we attribute it to the write-up's account.

Why a bill like this goes unread

Card processing gets set up once, usually when a shop opens or switches systems, and then it runs in the background. The statement arrives every month, but it's long and full of line items that are hard to read.

That's not a knock on processors. They provide a real service, and their pricing is a negotiation like any other vendor's. The point is that a negotiation needs someone who understands the bill well enough to ask a clear question.

That's where the agent helped, per the write-up. It read two years of statements, found the part that could be discussed, and turned that into emails an owner could send. In our read, that's work a busy owner tends to put off, not work he can't do.

Mallatt read every email before it went out

The part of this story we'd underline is the review step. Mallatt didn't let the agent send messages under his name on its own. According to the write-up, he reviewed every email and the counteroffer before anything went out.

That's how we think owners should start with any AI agent touching money or vendors. The agent does the reading, the math, and the first draft. The owner reads it, changes what needs changing, and decides whether to send it.

It also keeps the relationship with the processor in the owner's hands. A vendor negotiation is still a conversation between two businesses, and the owner should sound like himself in it.

Where the gap sits

What we'd look at in a shop like his is the setup.

Start with the statements. The agent worked from 24 months of them, per the write-up. Can you pull two years of processing statements today? If they're scattered across email, a processor portal, and a filing cabinet, gathering them is the first job.

Then read one statement yourself, even if an agent will read the rest. You don't need to master it. You only need to recognize the line items well enough to check the agent's summary against the paper. If the agent says it found a markup, you should be able to point to where it appears.

Next, know your own card volume. You don't need an exact figure to start, but a rough annual total by card helps any conversation about pricing. Your bookkeeping system or processor reports should have it.

Two years of statements and one rule

Once you have the statements and the volume, decide who approves. In a small shop, that's usually the owner. Write down the rule: nothing goes to the vendor until you've read it.

Then decide what a good outcome looks like before the first email goes out. We'd get final pricing in writing. Whatever you agree to should be on paper, not only in a call or an email thread.

Finally, pick one bill. Card processing is the one in this story, but the same approach could apply to other recurring vendor bills an owner rarely reads. Start with one, finish it, and see what you learn.

Keep a short record as you go: what the agent found, what you sent, what came back, and what you agreed to. That record becomes the starting point the next time the contract renews, and it shows you whether the agent's reading held up.

What we wouldn't read into it

It's one owner's account of one negotiation. His shop, volume, and processor are his own. Another shop could get a different answer, or no change at all.

We also wouldn't treat the dollar figure as something to expect. Mallatt reports about $5,000 a year on about $2 million in card payments. A shop with different volume, a different processor, or a different starting markup will see a different result.

What the story does show is a pattern worth copying: give the AI the documents, let it find the question, and keep the owner as the person who signs off. We've written about a similar handoff in what to hand over first in QuickBooks upkeep.

Card processing fee FAQ for repair shops

What did Franklynn Automotive's AI agent do?

According to an AI Business Plays write-up, Dan Mallatt gave Grok Bot 24 months of processing statements. It found the markup. It also drafted every email and the counteroffer. Mallatt reviewed each one before it went out, and he says on X the result is saving about $5,000 a year.

How do I get my shop ready to review card processing fees with AI?

Gather two years of processing statements and a rough figure for your annual card volume. Then set one rule before you start: nothing goes to the processor until you've read and approved it.

How much can a shop save on card processing fees?

Mallatt reports about $5,000 a year on about $2 million in annual card payments, his own self-reported figure from one shop. Your result depends on your volume, your processor, and your current markup, so treat his number as one example rather than a target.

If you want the long version of how we think about readiness, Start with a FREE quiz and Get Your AI Score.

BOOK YOUR STRATEGY SESSION

Get a clear plan for your first AI win in 30 days.

Join Lead Consultant Javier Aguilera. We’ll skip the generic pitch, identify your costliest manual bottleneck, and score your readiness for AI automation.

Conversational AI
Custom agents designed to capture and convert revenue.
Unified Intelligence
AI-powered operations across sales, marketing, legal, finance & ops.
Zero-Friction Execution
Loom-based updates; no endless meetings.