What problem does it solve?
Frying is a repetitive, physically demanding job in a restaurant kitchen: standing over hot oil during a rush while timing baskets precisely and handling other tickets at the same time. It is also a hard shift to keep staffed, particularly overnight, and inconsistent timing on the fryer shows up directly in food quality and ticket time.
White Castle's director of operations, Steve Foreman, describes Flippy's effect on the team as, "It takes some of that load off the team, allowing members to step away from the fryer and focus on other tasks," while Miso Robotics, Flippy's vendor, frames the wider goal as reducing the stress of peak hours.
How does it work?
- Sense what is loaded. Miso describes Flippy, at White Castle, as an "AI-driven robotic fryer." For a later pilot of the next generation Flippy 2 and the Sippy drink dispenser at Jack in the Box, Miso states the two products "leverage artificial intelligence, machine learning, computer vision and data analytics." Nation's Restaurant News separately reported that, as part of a planned expansion, Flippy's cameras and sensors "will also give" it the capability to show employees real time inventory needs and offer recommendations for complex or bulk orders, a planned capability with no source confirming it was delivered, and a description of inventory visibility, not of vision guided frying itself.
- Run the fixed task autonomously. The robot arm performs the one task it was built for: lowering, monitoring and lifting a fry basket. Miso describes Flippy's frying implementation as tapping into "sensors, intelligent monitoring and anticipated kitchen needs to keep food temperatures consistent."
- Hand off to a person. The finished output, a basket of fries ready to season and bag, moves to a hot holding area or prep counter where a team member takes over the next, less repetitive step.
- Flag what a robot cannot handle. A well built station should hold and alert a team member rather than guess when it meets a jam or a fault it cannot process. None of the public deployments on this page describes its exception handling in detail, so treat this as a design goal to check for your own robot.
- Expand at the pace the organization sets. White Castle began Flippy at one Chicago area restaurant in September 2020 and announced a plan to expand the pilot to up to 10 more locations in 2021; Jack in the Box announced its own pilot in April 2022 with "possible plans for further integration" left undisclosed. No source here confirms either expansion took place, or documents a guest feedback step for either chain.
- Audience
- Back office
- Autonomy
- Autonomous
- Adoption
- Emerging
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
No public deployment has disclosed a measurable outcome yet.
Value drivers: Employee productivity, Lower cost to serve, Customer experience.
Indicative value
A quick service chain piloting one kitchen robot in 20 restaurants
USD 78,000 to USD 343,200
Crew hours redeployed, valued at fully loaded labor cost per year
How this is calculated
Formula: restaurants * hoursFreedPerRestaurantPerWeek * 52 * fullyLoadedHourlyCost. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Restaurants with the robot installed restaurants, restaurants | 20 | 20 | Editorial illustrative scale, not tied to the deployments on this page: White Castle piloted the original Flippy in 1 restaurant, then confirmed a plan in February 2022 to install the newer Flippy 2 in 100 more locations, with no source found confirming how many of those installs are complete; Jack in the Box ran Flippy 2 at 1 San Diego location, with Sippy announced for the same pilot but no source confirming it was installed. Replace with your own rollout plan. |
| Crew hours freed from the automated task per restaurant per week hoursFreedPerRestaurantPerWeek, hours per restaurant per week | 5 | 15 | Editorial assumption, replace with your own time study of the specific task before and after. No deployment on this page discloses this figure directly. |
| Fully loaded cost of a crew hour fullyLoadedHourlyCost, USD per hour | 15 | 22 | Editorial assumption for US quick service labor cost including taxes and benefits. Replace with your own figure. |
What it leaves out: This is redeployed labor value, not a cash saving. None of the organizations on this page has disclosed a labor cost or headcount outcome, so this number should not be read as a saving from reduced headcount. It also leaves out the equipment, installation and maintenance cost of the robot, which is substantial and not disclosed by any named organization on this page.
Who already uses it?
2 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Jack in the Box Inc.
United States · Travel and hospitality · 2022
Jack in the Box began piloting Miso Robotics' Flippy 2 fry station and Sippy drink dispenser at a single, standalone San Diego location, announced in April 2022 by its Chief Operating Officer, Tony Darden, who said the chain was "looking forward to testing Flippy 2 as our new hire at our San Diego location." Miso Robotics describes the two products, "built on the Miso Platform," as leveraging "artificial intelligence, machine learning, computer vision and data analytics to maximize efficiency in the commercial kitchen." Miso's Chief Executive, Mike Bell, confirmed in an interview published 4 October 2022 that Jack in the Box in San Diego had adopted the robotic fry cook, alongside White Castle and CaliBurger locations, so the pilot had moved from announcement to an operating installation by then; no source found confirms whether Sippy was also running by that date, or discloses an outcome for the pilot.
No outcome disclosed.
White Castle Management Co.
United States · Travel and hospitality · 2020
White Castle began piloting Miso Robotics' Flippy, which Miso calls an "AI-driven robotic fryer," at a single Chicago area restaurant from late September 2020; Nation's Restaurant News separately reported that White Castle would also implement the newly revamped Flippy ROAR (Robot on a Rail), a rail mounted version announced in October 2020 that can glide between workstations, though no source cited here states which version was running at the September start. In October 2020, following early positive results, White Castle announced plans to expand the pilot to up to 10 more undisclosed locations in 2021; no source cited here confirms that expansion took place. Miso, Flippy's vendor, credits the robot with reducing the stress of peak hours so team members can focus on other tasks; White Castle's own stated reason for piloting autonomous frying was the need to limit staff while maintaining cooking speed and freshness during a pandemic demand surge. In February 2022, after upgrading the Chicago area restaurant to Flippy 2 in November 2021, White Castle confirmed a plan to install Flippy 2 in 100 more locations, with Miso describing those rollouts as "phased by region" and "planned out and scheduled in the months and years ahead"; no source cited here confirms how many of those 100 installs are complete.
No outcome disclosed.
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Training images covering the real range of product variation the fryer will see (basket types and loads)
- A kitchen layout and workflow map showing exactly where the unit hands off to a person
- A maintenance and cleaning plan that meets the same food safety standard as the task it replaces
Systems to integrate
- The fryer, oven or prep station hardware the robot operates
- Kitchen display or ticket system, so the robot's task starts and finishes in step with the line
- A safety interlock and emergency stop reachable by any team member
Complexity: High
This is a physical integration project, not just a software one: the robot has to fit an existing kitchen footprint, meet food safety and electrical codes, integrate with the fryer it replaces, and keep working when baskets and loads vary in size and quality.
- 1
Pick one high volume, low variability task first
Frying is a proven starting point because the task is repetitive and the failure mode (a bad basket) is easy for a person to catch.
- 2
Pilot in a small number of restaurants with different traffic patterns
Run it somewhere busy and somewhere quiet, and in more than one kitchen layout, before deciding whether and how to expand. The public deployments on this page each started as a single restaurant pilot before any broader plan, so treat multiple traffic patterns as a recommendation for your own rollout, not a description of what White Castle or Jack in the Box did first.
- 3
Build the exception path before the happy path
Decide up front what happens when the unit cannot process an item (a jammed basket) so a team member is never blocked waiting on the robot.
- 4
Measure the task, not the kitchen
Time the specific task before and after, and track crew feedback, rather than assuming labor savings across the whole chain from a single station pilot.
- 5
Plan for maintenance from day one
A robot that is down during a rush is worse than no robot. Agree service response times and a manual fallback procedure with the vendor before go live.
Guardrails
- Emergency stop and manual override reachable by any team member at all times
- Automatic hold and alert, not a guessed output, when the vision system cannot identify or size an item
- Food safety and cleaning procedures documented and audited the same as for the equipment it replaces
- No expansion beyond the pilot scope without a review of the specific task's before and after numbers
KPIs to instrument
- Time per unit of output (basket, portion) with and without the robot, on the same shift pattern
- Crew hours redeployed from the automated task to other work
- Downtime and manual override rate of the unit
- Output quality on a checked sample, compared with the manual process
Human in the loop
A team member loads the station, does a visual quality check on the output, and handles anything the robot flags or stops on. White Castle's public sources describe a staged, location by location rollout plan, and Jack in the Box's April 2022 announcement left further integration undisclosed; neither details a guest feedback step.
Common failure modes
- A single, well maintained pilot unit does not predict a fleet
- One robot in one clean, well staffed restaurant understates the maintenance burden of hundreds of units across ages of kitchens. Track downtime and manual overrides, not just the demo footage.
- Load variability breaks a narrow vision model
- A model trained on a limited range of basket types or loads misidentifies or mishandles the outliers. Keep the training set current with the real range the kitchen actually processes.
- The exception path is slower than doing the task by hand
- If clearing a jam or override takes longer than a person would take to do the task, the unit becomes a net loss during a rush. Time the exception path, not only the happy path.
What are the risks and rules?
EU AI Act
Depends on design
Frying alone is not a use listed in Annex III, so a robot that only plans and runs the frying task is usually minimal risk. That changes under Article 6(1), which is high risk when the robot, or an AI or machine learning component of it, is a safety component of a product covered by the Union harmonisation legislation in Annex I, and that law requires a third party conformity assessment of the product. For a kitchen robot the relevant product law is the Machinery Regulation (EU) 2023/1230, which requires that assessment for machinery whose safety function relies on machine learning that can change its own behaviour, for example vision guided collision avoidance or an interlock near a crew member. Since the Digital Omnibus on AI, Regulation (EU) 2026/1744, moved the Machinery Regulation into Annex I Section B, only Article 6(1), Article 60a and Articles 102 to 112 of the AI Act apply directly to that safety component, and the detailed requirements reach it through delegated acts amending Annex III of the Machinery Regulation; these AI Act rules apply from 2 August 2028. None of the public deployments on this page describes its vision or control system performing a safety function rather than only guiding the task, so check your own robot's design against this before assuming it is out of scope.
Rules that apply
Guidance
- Article 2, scope (European Union, Europe). Article 2(2), as amended by Regulation (EU) 2026/1744, says that for high risk AI systems related to products under the Annex I Section B laws, such as machinery, only Article 6(1), Article 60a and Articles 102 to 112 apply.
- Annex I, list of Union harmonisation legislation (European Union, Europe). The product laws behind Article 6(1). Since Regulation (EU) 2026/1744, Section B lists the Machinery Regulation (EU) 2023/1230 as point 21, and the Machinery Directive 2006/42/EC is deleted from Section A.
- Article 113, entry into force and application (European Union, Europe). Sets 2 August 2028 as the application date for the high risk rules that reach an Annex I Section B product, such as a kitchen robot regulated under the Machinery Regulation.
- Regulation (EU) 2023/1230 on machinery (European Union, Europe). The EU Machinery Regulation applies from 14 January 2027. Since Regulation (EU) 2026/1744 it is listed in Annex I Section B of the AI Act. It is the product law that can make an AI safety component of a kitchen robot high risk under Article 6(1), for example a vision guided collision avoidance or interlock function that can change its own behaviour.
Controls to put in place
- Documented, audited cleaning and food safety procedure equivalent to the manual task
- Emergency stop and manual override always reachable by crew
- Maintenance response time agreed with the vendor before any expansion beyond the pilot
Frequently asked questions
- What kitchen tasks do restaurant robots automate today?
- The public deployments on this page are narrow and specific: Miso Robotics' Flippy fries food at White Castle, and Flippy 2 fries food at Jack in the Box, where Miso also announced its Sippy drink dispenser for the same pilot, though no source confirms Sippy was installed. Neither runs a whole kitchen; each hands off to a person for the next step.
- How big are these deployments really?
- Small and staged. White Castle began Flippy at one Chicago area restaurant in September 2020, announced a plan in October 2020 to expand to up to 10 more locations in 2021, and in February 2022 confirmed a plan to install the newer Flippy 2 in 100 more locations, with rollouts "phased by region" over "the months and years ahead"; no source found confirms how much of either plan has been completed. Jack in the Box separately announced a pilot of Flippy 2 and Sippy at a single San Diego location in April 2022, with "possible plans for further integration" left undisclosed; Miso Robotics' Chief Executive, Mike Bell, said in an interview published 4 October 2022 that Jack in the Box in San Diego had adopted the robotic fry cook, a vendor's claim, so Flippy 2 was operating by then. No source found confirms whether Sippy was also installed by that date, describes a guest feedback step for either chain, or discloses an outcome for either pilot.
- Do these robots replace kitchen staff?
- The organizations frame it differently. White Castle's own reasoning, per trade press covering the rollout, was the need to limit staff while maintaining cooking speed and freshness during a demand surge, and its director of operations describes Flippy as taking load off the team so it can focus on other tasks. Jack in the Box's Chief Operating Officer, Tony Darden, described the pilot as a steppingstone for its back of house restaurant operations, aimed at "promoting safety and comfort to our team members"; the same Miso Robotics release announcing the pilot separately states the goal as team members spending less time at the fryer and drink stations and more time in front of customers. Miso Robotics, the vendor behind Flippy and Sippy, has separately framed the wider goal as redeploying staff to front of house work. Neither organization has publicly disclosed a labor cost or headcount outcome.
- Is a kitchen robot high risk under the EU AI Act?
- It depends on the design, not the category. Frying alone is not a use listed in Annex III, so if the AI only plans and runs the frying task, it is minimal risk. Under Article 6(1) it becomes high risk only when the AI or machine learning component is a safety component whose safety function relies on machine learning that can change its own behaviour, for example vision guided collision avoidance or an interlock near a crew member that keeps adapting on its own, because the Machinery Regulation (EU) 2023/1230 requires third party conformity assessment for that kind of safety function. Since Regulation (EU) 2026/1744 moved the Machinery Regulation into Annex I Section B, only Article 6(1), Article 60a and Articles 102 to 112 of the AI Act apply directly to that safety component, and only from 2 August 2028. None of the deployments on this page discloses whether its system works this way, so check your own robot's design before assuming it is out of scope.
How to cite this page
Blits.ai AI Use Case Library, "AI robotics for kitchen frying", last verified 30 September 2026, https://www.blits.ai/ai-use-cases/kitchen-robotics-and-food-prep-automation. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 30 September 2026: Published after review by an automated review workflow (independent skeptic review).
- 30 September 2026: Editor pass 5, responding to an adversarial skeptic review: reworded the indicativeValue note and the "What kitchen tasks" and "How big are these deployments" FAQ answers so that only Flippy 2 is stated as confirmed operating at Jack in the Box's San Diego location by October 2022 (per Miso's chief executive, a vendor claim); Sippy is now presented as announced for the same April 2022 pilot with no source confirming it was installed. Unlinked the Chipotle Autocado evidence record from this page and deleted it (it supported no other page), and removed the Autocado and produce prep material from the definition, aliases, problem, howItWorks, feasibility, implementation, humanInTheLoop, failureModes and FAQs, keeping the page on the two remaining named, public frying deployments (White Castle and Jack in the Box); also dropped the now stray "produce prep" wording from the EU AI Act basis and FAQ, and softened the Jack in the Box evidence record's "own article" description of the Times of San Diego piece, which carries a Reuters photo credit.
- 30 September 2026: Editor pass 4, responding to an adversarial skeptic review: added an independent October 2022 Times of San Diego source confirming Jack in the Box's Flippy 2 pilot moved from announcement to an operating installation, and reworded the meta description, FAQ and evidence to match; narrowed the title, definition and meta description to AI robotics for frying, dropping the produce prep scope that only Chipotle's non AI Autocado supported, and removed the metaDescription's unsupported "non AI" assertion and its promise of per deployment limitations; removed the unsupported claim that Miso framed staff redeployment as "rather than cutting headcount," and reworded the indicativeValue caveat to say only that no organization disclosed a labor or headcount outcome; removed the unsupported sensor timing and "either way" claims from howItWorks and restricted step 2 to what Miso and Chipotle actually say; corrected the NRN cameras and sensors sentence to the future tense the source uses; added the EU AI Act's self evolving qualifier, the reduced article set and the 2 August 2028 date to the FAQ, to match the basis field; removed "not just vendor demonstrations" from humanInTheLoop, which Chipotle's release does not say; dropped the unsupported retail-and-ecommerce industry; added risk.guidance entries for the Digital Omnibus and Machinery Regulation citations already in the basis field; and softened the White Castle evidence record's ROAR timeline and 360 baskets attribution to match the source exactly.
- 30 September 2026: Unpublished by an automated review workflow (independent skeptic review).
- 29 September 2026: First published