Tutor Intelligence, which is building warehouse robots using artificial intelligence, today brought in $34 million in Series A funding. This brings the company’s total funding to date to $42 million.
Josh Gruenstein, the co-founder and CEO of Tutor Intelligence, said he that when he launched the company in 2021, he felt jaded about the state of robotics research.
“At the time, people were thinking about how we build these learning systems where robots, through trial and error, could develop skills,” Gruenstein told Automated Warehouse. “I thought that every single one of the 8 billion people on Earth all knew how to do this stuff. It’s really intuitive to us
Gruenstein wanted to find a more efficient way to teach robots these tasks, so he founded Tutor Intelligence with a fellow researcher at MIT.
“Since then, we’ve built a team of over 50 people, including some really gifted engineers and researchers on the R&D side, and some awesome people who have recently joined us on go-to-market and manufacturing to spin up that side of the business,” said Gruenstein. “We’ve deployed robots from coast to coast. If you walk into a grocery store, there’s a good chance something in there has been touched by a Tutor robot.”
Tutor uses teleop as a stepping stone to full autonomy
Tutor Intelligence uses data gathered from its deployments to constantly train and level up its AI models. This AI and data flywheel method has become increasingly common in robotics, but getting that flywheel started can be a challenge.
“The first time we ever deployed a robot as a pilot, two or three years back, was to a perfume company in New Jersey,” Gruenstein recalled. “That robot was 0% autonomous. It was 100% remote-controlled. Over time, that data compounds and it scales, and you’re able to use that to support a larger and larger fleet.”
With teleoperation, Tutor was able to get robots up and running quickly, without any programming.
“We operate quite a large fleet, and it provides a massive benefit to our customers,” Gruenstein said. “Normally, to install automation, it’s a million-dollar project, with many months of a lot of customer engineering to program for exactly the thing you want that robot to do.” “
“But, because we’re able to support the sort of instant live learning, we can deploy a robot to a customer in a day or two days,” he added. “We can have that robot up and running without them having to invest any time or money in anything even close to resembling programming.”
Today, Tutor still utilizes teleoperation to keep operations running smoothly. From a customer perspective, it doesn’t really matter if the robot is acting autonomously 100% of the time; what matters is that it keeps running without someone having to intervene, noted Gruenstein.
“We still have a fleet of what we call ‘Tutors,'” he said. “This is a staff that we call robot operational support. If a robot is in a situation where it doesn’t know how to do something, it hasn’t done it before, then it enters an error situation. Those Tutors are on call and available to both get the robot out of that jam, and that’s also experience we can learn from.”
Save the date for the 2026 Robotics SummitWhat tasks can Tutor’s robots perform?
Tutor Intelligence said its robots use visual intelligence to identify, adapt to, and handle a variety of SKUs in live production. The company assembles and programs all of its robots in-house at its headquarters in Watertown, Mass.
“Today, we’re selling a robot named Cassie, which is designed very explicitly for stationary case-handling tasks. So, that’s palletization at the end of a line; it’s depalletization at the beginning of a line. Anything that’s a fixed task,” Gruenstein said.
“One of the ways we’re using our Series A funding is building out our product line so we can do a wider range of work, and build on our technology platform for customers in both manufacturing and logistics,” he added.
Currently, the team is focused on palletizing tasks, and Gruenstein didn’t specify whether Tutor is doing each picking yet for consumer packaged goods.
Tutor’s robots adapt to new sites quickly
Tutor Intelligence said it delivers its systems to customer sites in just 30 days after signing, and these robots are typically operational just one day after delivery. The company offers these robots through a robotics-as-a-service (RaaS) model.
At each new site, Tutor is able to use the data it has gathered from other sites to speed up deployments.
“There are specifics on a per-customer basis. If you have a specific way that you’re building your pallets in your factory in Columbus, Ohio, we want to ingest that information and learn how to do your task,” Gruenstein said. “So that’s certainly specific, but there’s obviously also learnings that we’re able to do benefitting from our fleet scale, and we’re able to pass that learning on to all of our customers every two weeks through over-the-air software updates.”
Gruenstein also emphasized Tutor Intelligence’s reliance on real-world data rather than synthetic data.
“We’re big believers in physical data,” he asserted. “I did my master’s thesis at MIT on the sim-to-real gap. That convinced me that collecting real-world data and building the real-world infrastructure for physical data is tremendously hard and operationally complex, but the data you’re able to collect from it is massively valuable in terms of your ability to immediately deliver robot capabilities, either in a lab or to a customer.”
“Our perspective is, at least at the moment, by doing the hard thing, we can get a pretty major advantage in terms of what our technology can achieve for our customers,” Gruenstein said.
Tutor works with large and small customers
Tutor Intelligence works with companies of all sizes looking to automate.
“We have some customers that are Fortune 50 enterprises, and they just have tasks that are too challenging or diverse [for them to handle], or they have so many different facilities,” Gruenstein said. “Having a coherent automation strategy requires these generally capable robot workers that can just show up at the facility and get to work.”
“At the same time, there’s a benefit in our ability to adapt that intelligence,” he continued. “We’re making these robots accessible for the first time to the mid-market and to small businesses that are just getting started.”
In addition to accelerating robot training, Tutor’s platform can provide higher-level management visibility.
“A lot of these customers, they don’t have other robots,” Gruenstein said. “Often, they don’t have that MES [manufacturing execution system] or ERP [enterprise resource planning] layer, so not only are we the first automation in the building; we’re also the first real-time data source for production that enables a level of observability in production that’s way higher fidelity than a guy walking by with a clipboard once a day.”
Startup to invest in commercialization
Union Square Ventures led Tutor Intelligence’s Series A round, which included participation from Fundomo and Neo. With the funding, Tutor said it plans to:
- Accelerate the commercialization of its robots
- Scale its consumer packaged goods fleet
- Advance its central robot intelligence platform and research infrastructure to power a new suite of form factors and capabilities
“We’re always trying to build robots that are smarter, do a wider range of work in the facility, and are more accessible for our customers,” Gruenstein said. “We think of AI as an enabling tool.”
While he didn’t specify what robot designs the company is exploring, Gruenstein did say he sees potential in humanoids.
“The product vision of what humanoids can do in the world is really right and crisp,” said Gruenstein. “That’s definitely where this AI technology is going to enable us to go in terms of building a new paradigm of industrial automation, but I think we can be a lot more pragmatic in terms of delivering these solutions to customers tomorrow, and then using that fleet scale to build the capabilities a lot earlier.”
“We look forward to either implementing humanoid hardware, or partnering with humanoid hardware when the time comes,” he concluded. “But for now, our goal is to build large fleets of robots that are able to solve large categories of problems in factories and warehouses.”
