Our Take
Smart carts are moving from pilot to deployment, but the story is not the hardware—it's that grocery chains now have edge-compute infrastructure to serve personalized offers and data at checkout, which matters far more than marginal usage gains.
Why it matters
Grocery retailers are consolidating checkout and customer data into a single in-store touchpoint. This shifts how they capture loyalty signals and test promotions, and it signals the practical limits of what grocer AI can do without reshaping store economics.
Do this week
Retail ops teams: audit your cart maintenance and power-management plan before deployment; Caper Carts require infrastructure that many stores underestimate.
Weis deploys Instacart Caper Carts across Pennsylvania
Weis Markets, a Pennsylvania-based grocer with 199 locations, is adding Instacart's Caper Carts to select stores. The carts include basket-facing and outward-facing cameras, certified scales, location-tracking systems, and a touchscreen. Shoppers can track spending in real time, apply digital coupons at the cart, access Buy It Again recommendations based on purchase history, and sign up for Weis Rewards loyalty programs directly from the device.
The carts use edge computing on the hardware combined with cloud AI trained on more than 1.6 billion online grocery orders (per Instacart). Weis and Instacart already operate a same-day delivery partnership across 133 locations in Pennsylvania, New York, and Delaware, established in 2023.
This rollout expands Instacart's broader Caper Cart footprint. The company now operates carts in more than 100 cities across 15 states, including deployments at Kroger, Schnucks, ShopRite, and Fairway Market banners.
Usage data from existing deployments
At a Schnucks location with 10 Caper Carts and approximately 160 traditional carts, the smart carts handled more than 10% of sales on busy days (per Retail Dive reporting). That metric—derived from a single store—remains the most concrete usage figure in published reports and shows adoption is real but niche.
Weis is layering AI across the checkout stack
Beyond smart carts, Weis completed a chainwide rollout of Toshiba Global Commerce Solutions' ELERA Security Suite across all 199 self-checkout lanes. The system includes produce recognition and loss prevention using edge AI. At deployment completion in December 2025 (per Toshiba's announcement), more than 94% of customers selected the produce recognition feature at self-checkout.
Separately, Albertsons Companies deployed an in-house computer vision tool for produce quality control in distribution centers, focusing initially on strawberries and grapes. The system runs on Google Cloud's Gemini Enterprise platform and Vision AI and is designed to catch moldy or damaged fruit before shelving and improve consistency in quality inspections.
Both moves reflect a broader pattern: grocers are embedding AI into the supply chain and checkout, not primarily to reduce labor but to capture data and enforce consistency at scale across large store networks.
What this means for retail operations
Smart carts are no longer vaporware. They are now operating hardware in stores with measurable—if modest—adoption. For retailers considering deployment, the real payoff is not checkout speed but the unified customer data layer: a single device that tracks what customers buy, which offers they redeem, and which recommendations they act on. This solves a problem that omnichannel grocers have fought for years: siloed loyalty data across online and in-store channels.
The catch: smart carts require sustained infrastructure investment (power, network, maintenance, staff training) and will remain a niche checkout method for years. Usage at 10% of sales is respectable but also a ceiling; most grocery transactions will still move through traditional lanes and self-checkout. Retailers should view smart carts as a data-collection tool and a loyalty accelerator, not a checkout replacement.
The move also signals that edge AI for loss prevention and produce grading is now table stakes. Toshiba's high adoption rate (94% of customers using produce recognition) shows that when the feature works and doesn't slow checkout, shoppers use it. Albertsons' in-house build rather than vendor purchase suggests that grocers with scale and tech talent will increasingly build bespoke supply chain AI rather than license off-the-shelf.