How robots and AI disrupted grocery shopping

Hello, and welcome to?New?Scientist’s?Business?Insights?newsletter.

Artificial intelligence and robotics are two areas that researchers have been working on for decades. Now these technologies are maturing, they can improve one another – as well as the way businesses operate.

Few people know this better than Sam Lloyd, head of data at Ocado Technology. The online grocery group has been trying to harness robotics and AI to pick and pack its items for more than 20 years.

Lloyd shares a few of his insights into this in today’s newsletter, but you can hear more at New Scientist’s Emerging Technologies Summit on 20 May, where he will be speaking more about Ocado’s innovations, and the lessons he has learned.

What journey has Ocado been on over the past 10 or 15 years when it comes to robotics and AI?

We were founded in the early 2000s with a mission to change the way the world shops for good, but we were really built to disrupt the e-commerce grocery sector. No one thought that was an industry that could be automated and disrupted for online, mainly because we have large basket sizes in grocery. Forty or 45 items is the average basket size – and that’s across a range of 50,000-odd stock-keeping units. You’ve got different temperature regimes, and things that are super fragile and super heavy all in the same bag. And the margins for grocery in general are super, super low.

The only way we thought we were going to tackle this and make it economically viable at scale was through robotics and automation.

How did things evolve?

Our first iteration was a customer fulfilment centre (CFC), a big warehouse, basically, with lots of dock bay doors for incoming and outgoing goods. The first CFC was opened in Hatfield, just outside London, in the early 2000s. We shut it down last year as part of an ongoing evolution, but it had something like 35 kilometres-plus of conveyor belts. It worked from the get-go.

We were thinking, “this works”. So, we built another CFC near Birmingham, which was going to be bigger. It was at that point we began to reimagine how we build a lot of our tech.

Instead of having conveyors that bring totes [boxes] to people, we have bots that pick up the totes from the top of a grid and then drop them down to the grid, then bots that pick up the customer order totes – big red containers. These get dropped down through the grid to pick stations, where manual pickers pick up your tomato ketchup and put it into a box. It’s super-efficient in terms of being able to supply customer orders. The totes move at about 4 metres a second and pass each other with millimetre accuracy. There are all sorts of complicated algorithms and artificial intelligence behind how we run these.

What lessons have you learned from that evolution? I imagine a lot of readers will be thinking, “We’re at the stage Ocado was maybe 5 or 10 years ago.”

One would be you have to continue [to adapt]. Technology at that point in time often didn’t exist, and we had to invent it ourselves. I think this has become more widespread and available to other businesses, such that you shouldn’t just go, “Well, we've built this now. We've got a sunk cost.” We need to continue to innovate and invest.

The other thing is we didn’t get it right from the get-go – at all. We had Hatfield, the first CFC, and completely reinvented that. You can play around with things and change how you do stuff. We’ve got three different series of bots at the moment. The first bots, called the 400 series, are heavier and slower than the 500 series and 600 series, as you can imagine. Our latest bots, the 600 series, are 3D printed with carbon-fibre reinforcement. They are topologically optimised in terms of strength for minimum weight. The message would be: don’t settle. Always be aware of what’s out there. And don’t be afraid to reinvent how you do things.

A lot has changed since Ocado started doing this, so I imagine there are more off-the-shelf options, so the company doesn’t have to spend as much time developing proprietary tech.

Absolutely. We are increasingly deploying robotic pick arms on our grid. We don’t build the arms ourselves – they are from a company called Universal Robotics. These arms wouldn’t have existed in the early 2000s at a [price] point we could have afforded to deploy these things. Technology has become more ubiquitous and affordable, and it has allowed us to think about how we can buy these things, then integrate them into our systems at a point that makes sense for us economically.