In commercial cleaning, artificial intelligence mostly means autonomous floor-scrubbing robots that follow a route mapped for one building and use cameras, lasers, and machine learning to avoid obstacles. Avidbots of Kitchener, Ontario, founded in 2014, is a Canadian maker of such machines, and its Neo 2W model is built for warehouses and factories with hard floors. Performance figures such as hours per charge and area per hour come from vendors, and independent measurements were not found.
Key takeaways
- Autonomous scrubbers clean a pre-mapped route; a technician maps the site and writes the cleaning plan before the first run.
- Machine learning in these robots is described mainly for obstacle recognition, such as pallets and forklift tines.
- Avidbots reported more than 1,000 robots in more than a dozen countries on September 27, 2022; its current About page gives no exact count.
- The U.S. Bureau of Labor Statistics projects 2 percent growth for janitors and building cleaners from 2025 to 2035, and mentions high-tech cleaning methods as a limit on growth.
- Productivity claims, including "doubled" team output and "up to 80%" fewer labor hours, are customer or executive statements, not measured results.
What an autonomous scrubber actually does
A robotic scrubber is a floor-cleaning machine with a computer attached. It sprays solution, scrubs, and recovers dirty water like a ride-on scrubber, but it follows a stored plan instead of a driver. Avidbots describes its Neo 2W as designed for hard floors only, so carpet and outdoor surfaces are outside its scope. The machine is large: the vendor lists about 152 cm in length, 137 cm in height, and a weight of roughly 581 to 688 kg depending on configuration (about 60 inches, 54 inches, and 1,280 to 1,520 pounds).
The company states that its own team maps a facility and generates the cleaning plans before the robot cleans it. The robot has a learning mode, and layout changes can be synced remotely. That arrangement matters for expectations. The machine is not walking into an unfamiliar building and working out what to do; it is repeating a plan that people prepared and updated.
Where machine learning enters
Avidbots' 2023 product announcement lists several features. Localization algorithms handle changing environments, and 3D cameras, lasers, and other sensors support collision avoidance. For warehouses, the company says machine learning detects pallets and the tines of forklifts and pallet movers, which are common floor hazards in that setting. Safety hardware is physical as well as digital: emergency stop buttons, bumpers, warning lights, and a projected blue light ahead of the machine.
The vendor markets the robot as fully autonomous, with no human intervention during a clean. Its own FAQ says otherwise on the edges: a local operator is recommended to supervise, and after a safety stop an operator must investigate and permit the robot to resume. Remote monitoring is offered around the clock. The practical picture is a supervised machine, with people handling exceptions.
Avidbots timeline from company sources
| Date | Event | Source type |
|---|---|---|
| 2014 | Company founded in Kitchener, Ontario by Pablo Molina and Faizan Sheikh, University of Waterloo robotics engineers | Company About page and press materials |
| September 27, 2022 | Series C of USD 70 million led by Jeneration Capital; total raised stated as more than USD 107 million; more than 1,000 robots in more than a dozen countries | Company press release |
| April 25, 2023 | Neo 2W announced for warehouses and manufacturing, with swappable batteries and machine-learning obstacle detection | Company press release |
| October 2026 | About page lists Neo, Neo 2W, and Kas; says "hundreds of Neos" are deployed; lists ISO 9001 and ISO 27001 certifications | Company About page |
The deployment figures do not line up. A 2022 release says more than 1,000 robots, while the current About page says hundreds of Neos, and the two statements may count different things (all models versus one model, or different dates). This article therefore gives no current fleet size.
Claims worth separating from measurements
Product pages in this category tend to blend three kinds of statements: specifications, vendor claims, and testimonials. Specifications include tank sizes of 109 liters (solution) and 135 liters (recovery). Vendor claims include up to six hours of cleaning per charge, with a typical range of four to six hours on the FAQ, and a stated cleaning rate of up to 42,000 square feet (about 3,900 square meters) per hour attributed to the chief executive in 2023. Testimonials include a lighting company saying team productivity doubled and a logistics executive quoted as saying such robots can reduce up to 80 percent of cleaning labor hours. None of these is accompanied by a method, a baseline, or an independent audit on the pages reviewed.
Labor context
The U.S. Bureau of Labor Statistics counted 2,432,600 janitor and building cleaner jobs in 2025, with median pay of USD 36,840 per year, and projects 2 percent growth from 2025 to 2035, slower than average for all occupations. Its occupational page does not mention robots by name; it says continued use of high-tech cleaning methods may limit employment growth. Canadian figures were not reviewed for this article, so no Canadian employment claim is made. What the data supports is narrow: slow projected growth, and a plausible link to equipment. It does not show that robots have replaced a measured number of cleaners.
Questions to ask before a site trial
- Which floors, aisles, and times of day are in scope, and which areas stay manual?
- Who maps the site, who updates the plan when racking or stock moves, and how quickly?
- What happens after a safety stop, and who on each shift is trained to respond?
- Which numbers come from a pilot at the buyer's own site, and which come from brochures?
- What data does the cloud dashboard collect, where is it stored, and who can see it?
- What is the full cost, including service, batteries, and operator time, compared with the current method?
This article is general information, not a procurement recommendation. It does not rank or endorse any product, and the named company is an example of the category.
FAQ
Does a cleaning robot need a person to run it?
It needs less driving but not zero oversight. The vendor's own FAQ recommends a local operator, and an operator must clear certain safety stops before the robot resumes.
Can these robots clean carpet or outdoor areas?
The Neo 2W is specified for hard floors only. Other machines from other makers cover other surfaces, and their specifications should be checked directly.
Is the machine learning in a cleaning robot the same as a chatbot model?
No. In the vendor's description, machine learning is used for recognizing floor obstacles such as pallets and forklift tines. It is a narrow vision task, unrelated to language models.
Do the productivity numbers hold up?
They cannot be confirmed from public material. They are statements from a company, a customer, or an executive, and a trial at the buyer's own site is the only way to produce a comparable figure.
Related reading
- How AI can help businesses cut costs
- AI in goods and services delivery
- AI in e-commerce operations
- AI and economic growth
More in AI Tools and Industry.
Last verified: October 2026
Sources
- About Avidbots, Avidbots Corp., accessed October 2026
- Avidbots raises $70M in Series C funding, Avidbots Corp., September 27, 2022
- Avidbots releases first floor-scrubbing robot built for industrial spaces, Avidbots Corp., April 25, 2023
- Neo 2W product page and FAQ, Avidbots Corp., accessed October 2026
- Janitors and Building Cleaners, Occupational Outlook Handbook, U.S. Bureau of Labor Statistics, last modified August 27, 2026