The Next Frontier: The Karp Question, AI Loops, and What Franchisors Need to Own
The Next Frontier explores how emerging technology and AI are reshaping franchising in real time. Pulled from our monthly insider newsletter, Franchise Unfiltered, this section focuses on practical applications, not hype, showing how forward-thinking franchise systems are using tech to remove friction, scale operations, and build durable competitive advantage.
This month, I want to pose two simple ideas that sit in slight juxtaposition with each other, by design.
One is a macro debate playing out among Fortune 500 executives. The other is the skill I believe you should be pushing your team to get familiar with right now.
First, the debate
What would happen if you woke up tomorrow morning and one of the big frontier labs, Anthropic or OpenAI for example, decided to compete directly or indirectly in your line of business?
This is a burning question that large enterprise executive teams are grappling with.
Last week, Palantir CEO Alex Karp went on CNBC and had what many in the media described as a “crash out.” Here is the full interview outside of the CNBC paywall.
As Karp explains, technical customers want “control over their computers, their models, their data stack, and their alpha. They want to know they own the means of production, and it’s not being transferred to someone else.”
This is a conversation you will want to keep a pulse on.
What does this mean for franchisors?
Every smart software supplier your brand relies on is asking themselves the Karp question right now. Some will get steamrolled, while others will leverage these tools to become far more valuable to you than they are today.
Your job is not to own a data center. You are not building your own models, although some at the bleeding edge are figuring out how to leverage open-source models. But you do need to own your data, your workflows, and your candidate and customer relationships. You also need technology partners who are not one model release away from extinction.
This is why I believe franchising will have one or two main companies emerge as leaders at the app layer: model agnostic, secure, purpose-built for the franchise hierarchy, and a far better on-ramp and infrastructure for franchise organizations than anything the labs will ship for the enterprise masses.
One of these companies is in our FSN portfolio. Book a call to learn more.
Now for the skill your team needs to learn
At the same time, the release of Claude Fable 5 is mind-bending.
When you combine these frontier models with what the brightest builders describe as “loops,” things get very interesting.
Think about it like this. A prompt is one instruction. You enter a question or request, receive an answer, and then decide what to do next.
A loop is a goal the AI keeps working toward until it gets there, without you sitting in the chair prompting it at every step.
Three parts make or break a loop:
A verifier: This makes sure you do not have an agent agreeing with itself on repeat. Without a real check on the result, you do not have a loop. The check can be a test that passes or fails, a metric that goes up or down, or, if you are using coding tools, a build that compiles or crashes. You need to separate the verifier role so the same agent is not grading its own homework.
State: This is what makes the loop learn. With each pass, the AI has to know what it already tried. Without that, it repeats the same mistake every cycle. A small file on the side records what is done, what failed, and what is next. Every day, the run resumes with this knowledge instead of starting from zero.
A stop condition: This keeps it sane. Every working loop has two ways to stop: the goal is met, or a hard limit says, “After X tries, stop and report.”
The skill of the decade is building the loop the model runs inside.
One prompt gives you one answer. One good loop gives you a worker that keeps executing toward an actual goal.
I am personally deep in the technicalities of agent and app deployment, with a specific focus on how to do so in the franchise hierarchy while keeping token costs from exploding and maintaining security and IP ownership.
We are watching the results inside real franchise systems right now, and the implications are enormous.
For homework this month, I would recommend asking your team to read up on AI loops. Search for “The Karpathy Method” and spend some time with it. Andrej Karpathy is widely regarded as one of the top AI minds in the world, and he has published excellent material on his learnings.
Want More Tech & AI Insights?
This conversation is just one example of what we share in Franchise Unfiltered, our monthly newsletter for franchisors who want real perspective on what’s changing, what’s working, and what actually matters in franchising right now, including:
Unfiltered perspectives from certified franchise experts embedded in franchise systems every day
Sharp insights and fresh perspective you won’t find anywhere else
No fluff. No vendor spin. Just practical insights you won’t find in press releases or panel soundbites
If you’re building, advising, or supporting franchise brands and want a sharper lens on where the industry is headed, subscribe to get Franchise Unfiltered straight to your inbox.