Amazon, BurgerFi, Chili's, and other companies are ending or rethinking delivery and restaurant robot trials, unable to balance their costs against usefulness
In retail, fast food and delivery, robots with names like Roxo, Rita and Scout have delivered mixed, sometimes quirky results, at costs some companies find too high. Tweets: @rakeshlobster Tweets: Rakesh Agrawal / @rakeshlobster : Bots are much better at destroying democracy than navigating the real world. Is there a more successful “bot” than an ATM? https://twitter.com/...
Context & Ripple Effects
The robot trials now being shelved were products of the pandemic automation push, when chains rushed in kitchen and delivery bots from startups like Miso Robotics and Chowbotics while delivery demand spiked. The WSJ's reporting closes that loop: Amazon's Scout, BurgerFi's and Chili's experiments with bots like Roxo and Rita couldn't make the cost-versus-usefulness math work outside controlled settings.
The retreat isn't uniform across the corpus — service robots such as Moxi in hospitals and Jesper in restaurants are still seeing high demand, which suggests the dividing line is environment predictability rather than robotics itself. And the hidden-cost question was visible early: Kiwibots' campus fleet ran on remote human pilots in Colombia paid $2/hour, a reminder that 'autonomous' delivery often carried labor costs of its own.
First-order effects
- Amazon, BurgerFi, and Chili's stop spending on trials whose operating costs exceeded the value of the deliveries or table runs they performed, freeing capital but ceding first-mover learning.
- Robot vendors that rode the pandemic wave — Miso Robotics, Chowbotics, and peers — lose marquee restaurant references just as funding environments tighten.
Second-order effects
- Surviving operators like Starship Technologies, which proved out grocery delivery in Milton Keynes during lockdown, and campus-focused fleets face pressure to show unit economics that big brands' abandoned trials could not.
- Restaurant and retail buyers redirect automation budgets toward back-of-house and structured tasks — the kitchen-automation segment — where repetition is high and sidewalks aren't involved.
Third-order effects
- If the pattern holds, delivery robotics bifurcates: robots persist in closed or semi-closed environments (hospitals, campuses, planned communities) while open-sidewalk last-mile delivery stays human-powered until per-task costs fall — echoing why the ATM remains the rare unqualified bot success cited in the coverage.
- The failure mode pushes the industry toward measuring robots by cost per completed task rather than novelty, making procurement discipline — not pilot enthusiasm — the gate for future rollouts.
The trend: Post-pandemic automation is consolidating around structured, high-repetition environments as open-world delivery robots fail the cost-per-useful-task test at major brands.