TAP. OPEN. GRAB. GO
BevTM is an AI Cooler—customers tap their card to unlock the door, take what they want, walk out, and get a receipt—no scanning.
Generates 30%+ more revenue than traditional vending machines.
Transactions settle instantly
Anti-theft software
99%+ accuracy in field
Why replace your vending machines with BevTM AI Coolers?
Average transaction increases by 30%+, more repeat customers, and less maintenance—no motors or coils.
How BevTM Coolers Work (Step by Step Walkthrough)
Payment Pre-Authorization
The customer taps (e.g., Apple Pay), inserts, or swipes their card to place a pre-authorization. If funds meet your configurable minimum, the transaction proceeds.
Door Unlock & Prompts
BevTM receives the authorization and unlocks the door. An audio/voice prompt instructs the customer to pull the handle and begin shopping.
Shopping
The customer opens the door and takes desired items. The system verbally confirms they may “take whatever you like.”
Door Close
The hinged door closes automatically (or the customer closes it).
Item Detection & Pricing
After the door closes, AI cameras and weight sensors identify which items were taken and pass that data to BevTM to compute the retail total.
Auto-Checkout (Closed Loop)
The pre-authorization is released and the customer is charged the final calculated amount. Unlike setups using standard PAX readers that may hold funds until settlement, BevTM voids the pre-auth and captures only the final amount—closed loop.
Frequently Asked Questions
1. What makes the machine AI?
Our machine uses learned models and makes inferences on what was took. In short: it’s AI because the machine learns from data and infers what happened using computer vision + sensor fusion, rather than relying on static, brittle rules.
2. How does the AI initially know what is being sold? Do you have to train it?
A: Operators select products from our master library—or add new SKUs—and set the machine’s initial quantities. When a customer removes an item, computer vision identifies exactly which product was taken. No, you do not need to train the machine. Our kiosks and software were uniquely developed for existing vending operators that don't have AI expertise. The problem with other smart coolers is that they require pre-existing knowledge in AI application and machine learning. Our machines bridge the gap and we developed roll-out ready software to help you scale from day 1!
3. This AI stuff sounds super complicated, am I going to need training or have to train my staff on how to work the machines?
A: This is the #1 question and we agree that AI and its applications are intimidating. Because of our operating knowledge and experience (Founder grew up in the coin-operated amusement and ATM space), we developed a program that is operator friendly, roll-out ready, and made the learning curve extremely simple and dummy proof! We are selling to vending companies and being able to do this was a crucial part of our development plan. So to best-answer the question: No, we developed the technology for traditional vending operators that aren't AI-experts. Hence, we developed the AI to be ready-to-go out the door and not need training or expertise.
4. What are the recurring fees?
Every transaction has credit card fees plus there are software fees and AI fees. A full breakdown on recurring fees are as follows:
- Credit Card Fees: 4% + 10 cents per transaction
- AI Recognition Fee: 4 cents per transaction
- Software Fee: $39 per month per machine
- Communication Fee: $20/month for wireless or free for Wi-Fi.
5. Can't you just steal from it?
A: You can try to steal from the machine and block the cameras or be super sneaky and quick, but the end result is this: the machine will catch you! Here is how it works: Our computer vision sees what you took and in the event it can't see, it can see before and after you leave the machine. It will run an auto-audit comparing inventory in the machine before and after it was closed and in the event of a camera blockage or quick trick, the software will flag the transaction and charge the customer what is missing. You also have the ability to be alerted and manually charge the cardholder.
6. How accurate is the AI in identifying the correct product?
Typical end-to-end recognition accuracy is ~95–99% but is closer to 95% in practice due to packaging similarity, mixed planograms, and not always ideal lighting. BevTM performs >99% under inconsistent facings, poor lighting environments, and similar products including identical products in different sizes (i.e.. 12 oz coke and 20 oz Coke)!
7. Does the cooler use weight sensors, computer vision, or both?
Many machines out there are vision-only (camera but not sensors or weight measures). This fact is irrelevant in itself, it's the results that matter in practice--in the field. A well-oiled machine performs >99% accuracy like BevTM--we uses field data from mixed lighting scenarios, similar packaging, and real-life targeted theft attempts plus child-like shenanigans and tricks. Hybrid models and new gadgets often suggest improved performance, but this is one of the biggest myths. How a machine makes inferences to detect what is took is what matters and different technologies can achieve the same thing. How the technology learns and behaves on a software platform is what drives results.