fbpx

Artificial Intelligence Community Insights

Four Reasons Most Service Robots Fail

The Tally robot, from Simbe Robotics, helps out in stores so people can be freed from the monotonous task of checking shelves for inventory. (Image: The company)

Wouldn’t it be nice if robots could clean our office, greet clients or put away the groceries at home?  But even though we want and often embrace advances that make our lives simpler, the vision of robot as dutiful helper hasn’t come to fruition.  Robotics has made a big difference in manufacturing, but has failed to make it into our homes and businesses in a meaningful way. 

It turns out the challenges are plentiful.  It’s hard to deliver on a consumer or business need without creating operational issues or setting unrealistic expectations of what the robot can do.  Few have been able to do that at a price businesses or consumers are willing to pay. 

Here are the four main challenges that robotics companies typically face:

1. No Product Market Fit

We’ve all seen those YouTube videos where robots do amazing things like climb walls or do back flips.  But such accomplishments are not particularly useful in everyday life.  Sometimes product engineers aren’t guided by market need so much as the desire to solve a mechanical problem.  The result can be an amazing machine no one wants to pay for.  

Theodore Levitt, economist and former editor of the Harvard Business Review, famously told companies to focus more on customer needs and not their own capabilities: “Understanding the customer should come first, and the product should follow—not the other way around.”

What would you pay to not have to vacuum your house?  This is the fundamental question iRobot answered with its smash-hit Roomba.  It isn’t perfect, but its value to you the consumer has always been clear.  For as little as $200, a tedious chore is effectively perpetually done for you.  

(Source: Shutterstock)

Roomba understood a need in the marketplace that it could solve with robotics.  But currently most service and consumer robot companies struggle to either identify or deliver on a clear value proposition.  

Robotic welding arms that can meet a car frame at exactly the same point every time are uniquely suited to adding value in auto manufacturing.   Service robots, however, still have a long way to go in showing economic value for the tasks they can actually do.  

2. Lack of Product Features or Attributes

We have robots that can go anywhere and do just about anything.  Where they still  need to improve is in the software.  

A robot that can’t distinguish between an object and its background won’t be very good at picking that object up.  Similarly, a delivery robot can’t get lost, and a hotel lobby robot has to be able to understand you.  But this is where we see companies launching solutions and coming up woefully short. 

A robot designed to greet you in a hotel needs to be able to see, hear and understand you and then deliver on your needs. It must, for example, check you in or out or offer you restaurant options and directions. The benefit to the business would be in freeing up staff to handle more complex interactions with guests. 

But a computerized robot that can replace a front desk worker, a concierge, or a bellman will require massive computing power and data sets.  It would need speech and/or facial recognition software, noise-cancelling software, cloud computing capability, device security software and API access to reservations and other company systems, all in a form factor with which a guest wants to interact. And this is just part of the software stack needed.  That is an incredible amount of work for an engineering team.

Further, the robot company will likely try to minimize development time by focusing on just one of these use cases to narrow its product scope, get to market faster and maximize performance.  But this will reduce the economic value the robot can provide to a business customer, because it now only performs one task.  That means the cost of the robots vs. the cost of a front desk worker comes perilously close to zero or may even be negative.  Further, if the machine fails more than two or three percent of the time, it won’t make sense to operate. Why buy a robot to do what the staff can do better and with less friction?

3. Real World Operational Hurdles

Burger-flipping robots work best when the kitchen is built around them, not when they come into an existing kitchen that is retrofitted. It’s an example of the operational hurdles these machines face.   

A retail store or a warehouse, for example, has already been optimized for efficiency and effectiveness.   If you are a business operating on single-digit profit margins, you likely won’t risk adding operational complexity.  If a robot can’t work in the business without someone having to follow it, then it creates operational hurdles.  Someone also has to be responsible for the robot, maintain it, keep it in good working order and make sure it is charged.  If that requires extra headcount, the value equation is likely thrown off and a business will choose not to adopt the technology.   

Finally, though robotics software is advancing rapidly, it still can’t deliver on the expectations formed by pop culture.  The benchmarks in peoples’ minds are Rosie the robot from the Jetsons, C-3PO of Star Wars, and the machines in ExMachina.  But the reality is that it’s an incredible challenge just to have a robot understand you in a public setting like a noisy hotel lobby.  

To get a robot to do what we’ve seen in the movies is not technologically possible, at least not at an affordable price. This is one of the main reasons home robots like Anki’s Vector, Kuri and Jibo ultimately failed. You can’t just have a conversation with them. But that’s what we wanted.  Cute robots like these either need to do chores for you or to be a companion, which means they need to be able to interact and converse with you.  If they don’t, Amazon’s Echo may be all you need. 

4. Bottom Line for Success

iRobot is arguably a good example of a company getting it all right.  A more recent one is Simbe Robotics’ “Tally”.  It’s a shy robot that blends into its surroundings, moving slowly and methodically.  It is designed to tackle the massive problem of out-of-stock items in retail stores. Tally checks the store three times a day and relays information to the staff about whether items are out and priced correctly.  This frees up the staff to focus on stocking and helping shoppers.  Simbe identified an industry problem and delivered a robotic solution that works flawlessly.

Robotics companies need to do a better job finding customer problems they can solve well with the technology available now, as industrial robots do.  They need to narrow the scope of their offering to what both the hardware and the software can deliver on, 100 percent of the time, with little or no human intervention.  Finally, these companies need to do a better job setting expectations.  A robot that can move around autonomously generates much bigger expectations than does a robotic coffee maker. 

It is inevitable that the business value of robots will catch up to what we need. The question is: how long will it take?  Robotics companies that address these four market barriers will be the ones that are most likely to deploy fleets soon.

Steve Carlin is CEO of Kebek Ventures and was previously president and chief strategy officer for Softbank Robotics America. Softbank Robotics created the Pepper service robot. A longer version of this piece as well as other recent articles can be found here.

Leave a Reply

Your email address will not be published. Required fields are marked *