Tell us about yourself and your company.
I grew up in a family business in restaurant technology, designing POS terminals. So I came into AI with a grounding in how restaurants actually deploy and use technology.
Berry AI uses computer vision to give QSRs operational visibility. Our core product is an AI camera drive-thru timer, integrated with POS data, that tracks the full customer journey from the moment a car enters the lot to the moment it leaves, not just two checkpoints in the pavement.
Today Berry is the most deployed camera drive-thru timer globally, working brand-wide with brands like Zaxby’s and Culver’s. Across deployments, customers have seen 20 to 40 percent reductions in drive-thru service times.
How did you approach product-market fit and early customer growth?
For us, product-market fit came down to use case selection. Vision AI can theoretically do almost anything, which is exactly the trap. Most vision AI companies pick a use case that demos well but can never survive real-world deployment, and end up stuck in pilot limbo for anywhere from months to over a year.
We evaluated use cases against four criteria: how accurate does the system need to be, how often will the camera view get blocked, does it need to work in real time, and how visually varied are the objects. Drive-thru timing scored well on all four. Operators act on aggregated data, so occasional misses on individual cars don’t break the system.
There was also a natural opening in the market. Loop timers were aging out, and many brands with complex drive-thrus were ready to skip loops entirely and go straight to cameras.
From there, growth came incrementally. A few stores, then 30, then hundreds. We designed pilots to prove scalability, not to look good, testing in both representative stores and the most complex layouts a brand had.
What is one lesson or piece of advice you would share with other leaders or founders?
One idea that has stayed with me: you can recover from a failure in competency, but you cannot recover from a failure in character.
This matters even more in QSR. When a brand signs with a vendor, they aren’t buying a product, they’re choosing a partner for the next five to ten years. Their operators will live with your system every day. What brands are really evaluating is whether they can trust you over that horizon.
In practice, trust means three things. Meet your promises, don’t overpromise. Be transparent, share the ugly along with the pretty, including what your technology can and cannot do today. And be accountable. When something breaks, own the mistake and correct it swiftly.
Competency problems can be fixed. But once a customer catches you shading the truth, that trust is very hard to rebuild.
What are the biggest trends or shifts you are seeing in your industry right now?
The biggest shift is that vision AI is moving beyond pilots into real production adoption.
For years, the pattern in QSR AI was an exciting demo, a five-store pilot, a press release, then quiet. The technology was good enough to show, but not good enough to run a business on.
That has changed. Zaxby’s and Culver’s have adopted camera-based drive-thru timing across every restaurant in their systems. Vision AI has become core operational infrastructure for these brands.
I expect this to accelerate. Once one brand proves a use case at full scale, the risk calculation changes for everyone else. Executives have moved past asking whether the technology works, and now ask which use cases are mature enough to deploy today.
What are the most common mistakes people make in your area of expertise?
A few I see often:
1. Choosing the wrong use case. Vision AI use cases vary wildly in maturity. Pick the wrong one and you end up in pilot limbo, where the system is good enough for a demo but never good enough to deploy, wasting months to over a year.
2. Underestimating the last mile, and running pilots that hide it. Getting a vision system to 95 percent accuracy is easy. Getting to 99.9 percent is brutally difficult. Small pilots in friendly, standard stores hide this gap, then the rollout hits stores with bad lighting or unusual layouts and the system falls apart. My rule of thumb is to pilot in 20+ locations that reflect the true diversity of your system, or deliberately pick your hardest stores to find the limits before you commit.
3. Ignoring time cost. Operators scrutinize the dollar price but rarely ask about the time price. I once heard a prospective customer with about 100 stores mention they had two dedicated people just maintaining their vision AI solution. Always ask a vendor how much of your team’s time the system will consume at launch and ongoing. A vague answer is a red flag.
What advice would you give to someone looking to build real authority in this space?
Two things: customer understanding and focus.
Customer understanding means spending time in stores. Credibility comes from understanding what makes each brand’s operation unique, and from seeing the same technology through the eyes of a restaurant GM, a district manager, a franchise owner, and a corporate team, because each experiences it differently.
Focus means going deep and perfecting one thing before broadening. The temptation in AI is to chase every use case, and the result is usually a company that does five things at a demo level and nothing at a deployment level. Depth earns you the right to an opinion. Breadth without depth just makes you another vendor with a roadmap.
