Amazon Go stopped. Sensei shows why the concept was never dead
For two days in Portugal, I had the opportunity to visit several stores equipped with the technology developed by Sensei. The programme included: a Continente supermarket, a convenience store located inside Gulp's offices and an automated pharmacy open almost 24/7.
The first conclusion is simple: we are no longer looking at a technology demonstrator. We are looking at a solution that works in real stores, with real customers, and addresses challenges that go far beyond simply replacing the checkout.
The most impressive concept is undoubtedly the Continente supermarket.
A real supermarket, not a laboratory
Unlike Amazon Go, which was often limited to small convenience formats, Sensei has deployed its technology in a supermarket of nearly 1,200 m² offering around 13,000 SKUs.
The store includes all the traditional departments:
- grocery
- fruit and vegetables
- meat
- fish
- bulk
- beverages
- large pallet displays
In other words, a genuine supermarket where all the usual complexities are present.
This is probably the most important demonstration: this technology is no longer limited to small-format stores.
The biggest difference with Amazon Go: the customer experience
One of the main reasons behind Amazon Go's failure was undoubtedly the friction created before customers even entered the store: a mobile app, a QR code, prior identification… All of these steps disrupted the shopping journey.
At Sensei, the experience feels much more natural. Customers simply walk into the store: no app, no smartphone, no QR code.
They pick up a product, put it back if they wish, continue shopping normally and then head to the checkout where their basket is automatically reconstructed before payment.
The shopping journey ultimately remains that of a traditional supermarket.
The technology almost completely disappears in favour of the customer experience.
How does the technology work?
The store ceiling is equipped with numerous cameras, while the shelves integrate smart weighing sensors.
As soon as customers enter, each of them is assigned a completely anonymous avatar within the system.
This is not facial recognition but rather a virtual identifier that tracks movements and interactions with products.
Every action is recorded:
- picking up a product
- putting a product back on the shelf
- placing an item in the basket
- behaviour in front of a shelf
When it is time to pay, the system automatically knows the exact contents of the basket.
Even when a family shops together, the different avatars are automatically grouped together to create a single shopping basket.
A solution designed to reduce errors... and theft
One of the most interesting aspects is the way Sensei handles potentially fraudulent situations.
During the visit, several demonstrations showed how the system detects suspicious behaviour.
For example:
- a customer trying to enter the wrong banana variety at the weighing station;
- swapping two wine bottles in an attempt to pay the price of the cheaper one.
- ...
In these situations, the system automatically generates an alert.
Depending on the situation, the system can:
- request the intervention of a staff member;
- add both product references to the basket until verification;
- flag an anomaly without unnecessarily interrupting the customer journey.
The objective is therefore not only to eliminate checkouts but also to strengthen loss prevention.
Much more than a payment system
This is probably what differentiates Sensei the most from the first generation of "Just Walk Out" solutions.
The technology does not only identify the products being purchased.
It also powers several essential store functions:
- inventory management;
- replenishment alerts;
- inventory support;
- planogram optimisation;
- shopper behaviour analysis;
- loss prevention.
In reality, several business cases are brought together within a single platform.
This integration is probably what makes the business model far more economically viable today.
And what about the staff?
Contrary to a common misconception, the objective is not to replace store employees.
On the contrary, the teams I met on site explained that the solution allows them to devote more time to customer advice and service.
Their interventions mainly concern:
- potential errors detected by the system;
- assisting customers;
- replenishment made easier thanks to automatic alerts.
Staff spend less time on repetitive tasks and more time on value-added activities.
The numbers that stand out
The Continente store I visited welcomes around 1,500 visitors per day.
Today:
- around 70% of all transactions already use the Sensei technology;
- these transactions account for nearly 82% of the store's total turnover.
These are particularly impressive figures for a technology that is still relatively new.
The real takeaway
What impressed me the most was not the technology itself.
It was the fact that it integrates naturally into the daily operation of a supermarket.
Every specific use case seems to have been addressed:
- weighted products;
- meat;
- fish;
- bulk products;
- large-volume items;
- group shopping;
- error prevention;
- loss prevention.
There is no alternative shopping journey, no exceptional procedure and no "workaround" to compensate for the solution's limitations.
Everything appears to have been designed to work in a real supermarket environment.
After this visit, my main impression was that we are no longer talking about a pilot project.
We are talking about a technology that is mature enough to support large-scale deployment.
And that is probably the biggest difference.
Podcast episode and more photos available via: https://vincentpanneels.substack.com/p/amazon-go-sest-arrete-sensei-montre
PortugalSenseiRetailRetailtechTechnologyAutomated StoresContinente