Brainstorm Tech 2024: How To Build A Better ‘Bot
Peggy Johnson, Chief Executive Officer, Agility Robotics Interviewer: Jason Del Rey, Fortune
Category
🤖
TechTranscript
00:00A lot of people want to work with robots.
00:02That's probably maybe a good start for you?
00:04Yes.
00:05I like that start.
00:06So, you've been on this job an entire four months.
00:10Right.
00:11Is that right?
00:12Yes.
00:13Okay.
00:14You've had a long, successful career, Qualcomm, Microsoft, Magic Leap most recently at CEO.
00:20I want to talk about that, but maybe that's over drinks or something.
00:25Why agility right now as your next CEO role?
00:29You know, it's a great question.
00:30Right now, humanoid robots are having a moment, for sure.
00:35And I've always been fascinated with robotics my whole career.
00:39My first boss had done his Ph.D. in robotics years ago.
00:44And they're finally having this moment.
00:46And when I looked across the field of players, agility to me stood far and ahead of the others.
00:52It had a working, autonomous robot that was getting ready to take actual work in factories.
00:59And since that point, we actually have done that.
01:01So I think we'll get into some of that, but did you have a treat for us?
01:06I do.
01:07If you'd allow me, I'd love you to let me welcome Digit to the stage.
01:14Digit is the name of our robot.
01:18So I have one obvious question at first.
01:23Actually two.
01:26Why legs at all?
01:27Because a lot of robots maybe roll.
01:30And then why those legs?
01:32Okay.
01:33So you notice they're going the other way.
01:36So we have to go where humans go.
01:40That's what we were designed for, the small spaces as oh, Digit's waving hello to everyone.
01:46The small spaces that humans operate in.
01:49So humans, if you go into some of these big warehouses, humans can only reach so high.
01:55It had to reach that high.
01:57And they can reach so low.
01:58It had to reach that low.
01:59Turns out, knees get in the way of picking up heavy things from the ground.
02:04So we put the legs going the other way.
02:07They don't get in the way.
02:08Do you call those reverse knees or just no knees?
02:11Some people call them bird legs.
02:13I just like to say they're Digit's legs.
02:15What works best in that situation.
02:19I think we need audience over there to wave hello as well.
02:22Yeah.
02:23Digit, don't be shy.
02:31So this was oh.
02:32Yeah.
02:33Okay.
02:34Now he's a ham.
02:35Now we can't get him off the stage.
02:39I'm waiting for him to take one of my tasks.
02:41Yes.
02:42Yes.
02:44So obviously this is a brief little fun demo, but Digit is actually in workplaces right
02:51now.
02:52Yes.
02:53So as of a few weeks ago, Digit is at GXO Logistics.
02:57They're a third party logistics, a big one, outside of Atlanta.
03:01Digit is taking tasks online and doing them versus humans doing some of these dull, repetitive
03:09tasks.
03:10Actually, Mary Daly was talking about it.
03:12We want to augment humans, allow them to be freed up to do other things, and it'll
03:17take the very repetitive things that are part of the everyday job off of the human's plate
03:23so they could be more productive in other areas.
03:26Are we going to see ... I think maybe we have a clip we can show.
03:29We have a clip we can show where we're operating right now, if we can run that video.
03:35So GXO actually moves Spanx, women's wear, and they have these huge automated warehouses,
03:43and then there's still a human moving things from a conveyor belt to a cart, or from a
03:49cart to a foot wall.
03:52It's those points.
03:53This actually is in an Amazon facility.
03:56A real Amazon facility.
03:57A real Amazon facility.
03:58Okay.
04:00So amidst these islands of automation, this is sometimes the Achilles heel, because there
04:08needs to be a human in there, and they're hard to find.
04:11There's about 1.1 million open jobs in the material moving space in the U.S. today alone,
04:19and they can't be filled.
04:21Nobody wants these jobs.
04:22They're doubling.
04:23It's doubled in the last five years, because we all want next day delivery, and you want
04:29your Spanx the next day, you want Digit in there delivering them.
04:33So we have a lot to talk about, but I'm sure the audience will have a bunch of questions,
04:36so I'm hoping to get to a few of you.
04:38So start thinking about what you want to know from Peggy, or maybe from, well, goodbye Digit,
04:43but maybe from Digit as well.
04:47So a lot of questions.
04:49There's a lot of money going into the space right now, a lot of noise.
04:53Like you said, you have a real live commercial relationship with GXO.
04:59You're in Spanx facility.
05:01You're also in an Amazon facility.
05:04I've covered Amazon a long time.
05:07Some of the automation reduces repetitive tasks, but they've also had situations where
05:11it then speeds up the work, and there's some safety concerns as well.
05:15So I'm curious, like when you're going into these facilities, what the goals are and how
05:20much you prioritize safety along the way.
05:25Safety 100%, 100%.
05:28This company was built on a safety foundation.
05:32So from the software to the hardware to how it moves and how it's programmed, you are
05:37operating in an environment with other people.
05:41Right now, we still operate inside of a work cell, but we're headed toward-
05:45That means no humans within are able to interact?
05:49Right, unless we're powered down, because they're 140 pounds, there's a lot of torque,
05:54and so you have to be careful about that.
05:56However, we're on a path to what's called functional safety.
06:00So in the next 18 months or so, we'll be able to incorporate all of the things that are
06:05needed in order to be able to interoperate near humans.
06:10Can you talk about sort of some of the stats about what the robot can do in terms of weight
06:15that Digit can lift, battery cycle?
06:20Obviously this is a cost issue for a lot of these employers, and so I'm curious all the
06:28metrics that go into them figuring out whether they can save money by using Digit or many
06:35Digits.
06:36Yes.
06:37So right now, we're looking at under a two-year ROI versus a human at a fully loaded $30
06:44per hour.
06:45So that's the benchmark that we're focused at.
06:48You're right about power.
06:51You need to be operating.
06:52We operate for hours, but we recharge ourselves.
06:56Right now, we're at about a two-to-one ratio.
07:00So-
07:01What does that mean exactly?
07:02So two hours or units of operating time versus one unit of charging.
07:06Moving to a four-to-one and then moving higher up in our next gen to eight-to-ten-to-one.
07:12So that's the trajectory we're on there.
07:15We can lift, you know, right now between our current gen and next gen, we're moving from
07:20about 30 pounds to about 50 pounds.
07:23But you have this OSHA limit already in place for humans.
07:27So that's the bar we have to meet there, and we'll do that.
07:30Which is the 50?
07:3150 pounds.
07:32Yeah.
07:33But, you know, if you think about 50 pounds, lifting something over and over and over again,
07:38that's the problem.
07:39That's where the injuries come in.
07:40That's where the quitting comes in, the turnover.
07:43You know, in these warehouses, turnover's super high.
07:47Yep.
07:48So I will call on- if anyone has a question, please raise your hand and I'll get to you.
07:54Okay, we'll go to one now and then we'll come back.
07:58Please just tell us your name.
07:59Yep, you're right.
08:00Oh, you need a mic, don't you?
08:01Mike.
08:03Coming up, coming up right behind you.
08:06Just tell us who you are and what company you're with, please.
08:09Hi, Peggy.
08:10Hi.
08:11Ken Washington.
08:12I'm the CTO at Medtronic, and I've known your company for a long time, from my days at Ford
08:19Motor Company and also at Amazon.
08:22And in both of those examples in my past, the digit experience has changed a lot.
08:30But what I haven't seen yet is an integration of digit with other automation platforms.
08:38Is that something that you're thinking about, how to create an API so that other parts of
08:43your automated ecosystem can communicate with digit and vice versa so that they create a
08:48system and work together in an optimized way to improve workflow?
08:52Absolutely.
08:53So we have developed a software platform called Arc.
08:57It has all the APIs to plug into the warehouse management systems that are existing in facilities
09:04today.
09:05So we've integrated already with Manhattan and several others.
09:08So we can step into the existing workflow.
09:11We don't want to have to have the operator change the infrastructure, change anything
09:16about the workflow.
09:17We just step right in, plug into their warehouse management systems.
09:22And we've also integrated with quite a few other robots, whether they're autonomous mobile
09:26robot carts or put walls or conveyor belts.
09:31And so we've got it on both ends.
09:32Where that human goes, digit walks in and we're connected on either side.
09:38How far along are you in that journey?
09:41How confident are you that you can walk into a new customer's facility and whatever the
09:48automation is, digit can currently work within that?
09:52Or is that a work in progress?
09:54Very confident.
09:55The other thing is we have a team of people who started and built robotics companies on
09:59my leadership team.
10:02And they brought that sensibility, the safety, the regulatory, the compliance, the fact that
10:07you have to step into an existing workflow.
10:09All of that is beyond just getting the robot to walk, right?
10:14You have to enter the corporate IT infrastructure and make it work for them.
10:18You don't want to make them change things.
10:21So we have a good sense of that from the leadership team that the folks before I arrived
10:26had built.
10:28Let's talk about the technology a little bit.
10:29And I'll get to another question if we have right after this.
10:33How is digit doing what it just did on stage?
10:37And then how is it learning and or deciding how to work inside these active facilities
10:44that it's in right now?
10:45Right.
10:46So how it's doing it is the company is 10 years old.
10:48It came out of Oregon State University with their robotics institute there.
10:52We have still a big tech force there and also a factory nearby that we have built to take
10:59advantage of the location.
11:02But digit has been, you know, 10 years in the making.
11:05This is highly complex, a highly complex physics project.
11:09The dynamics of locomotion are terribly complex.
11:13The team solved that part of it.
11:15Now when the world changed, November 30th, 2022, with chat GPT, we can start to introduce
11:23that to help with the semantic intelligence.
11:26Now we have been using reinforcement learning for several years on the locomotion to refine
11:31the locomotion.
11:32So that digit can walk over uneasy places.
11:35It can go up and down stairs.
11:38And that's what makes it very different from, say, a wheeled robot.
11:41It can traverse those things.
11:43So that part's been great.
11:45Reinforcement learning helps that, but the semantic intelligence with chat GPT is just
11:51amazing.
11:52It's the embodiment of AI.
11:54Can you give the audience an example of how that might work?
11:58Yeah.
11:59So now we have the ability to direct digit.
12:01So the other day we threw a bunch of trash on the floor, different kinds of trash, and
12:05we put all the bins behind digit.
12:08So, you know, recycle, landfill.
12:11And we said, just gave it one command, digit, clean up this trash.
12:15And digit scans the room, picks up the paper, puts it in the recycle, picks up some cardboard,
12:23does the same thing, goes and picks up some bubble wrap and puts it in the trash.
12:29And we all went, ugh, you know, mistake.
12:31But it turns out bubble wrap is not recyclable.
12:34So digit, with the power of chat GPT underneath, was able to make that decision.
12:39So now we don't have to code all of that.
12:41We can just, with chat GPT and other LLMs, we have the ability to just give digit a direction.
12:50So digit can do one thing in the morning for a company and a different thing in the afternoon.
12:55Do we have any other audience questions?
12:56Otherwise I'll keep going because I love talking.
12:59All right.
13:00I'm going to start, actually, sorry, we're going to start there.
13:04I faked you out.
13:05Yeah.
13:06Right there.
13:07Please tell us who you are.
13:09Yes.
13:10Sterling Audie with Barclays.
13:11So the example that you showed is a single unit in a single unit use case, but I guess
13:17we can all envision armies of these being throughout the warehouse.
13:21So have you already kind of developed the management platform and ability to kind of
13:25have a, you know, a control plane across a larger number of units?
13:30That's exactly what Arc is.
13:32So and what we're finding is the employees who used to do maybe all of that physical
13:38work are becoming the managers of the robots.
13:41So they're along the way getting to be upskilled with the use of our Arc platform.
13:46They're the ones who are managing the fleet and it doesn't have to be just digits, right?
13:50We're connecting to AMRs and other things so they can look at the fleet of automation
13:55and manage that from there.
13:56They have a digital.
13:57AMR just for folks.
13:58Oh, sorry.
13:59Autonomous Mobile Robots.
14:00So they're the carts that you'll see going back and forth in the warehouse facilities.
14:06Got it.
14:07Yeah.
14:08All right.
14:10Cool.
14:11Okay.
14:12I'll come to you right now.
14:13Yeah.
14:14No, you can go ahead.
14:15Cody Cerny with Insignium.
14:16Peggy, great to see you again.
14:17Hi, Cody.
14:18Is there anything that's keeping you up at night or any worries you have about the development
14:22of these robots and the implementation?
14:24Well, the biggest thing keeping me up at night right now is I just got off the phone with
14:28my commercial officer and we are just getting ready to produce our next gen in the fall.
14:34You know, it's already developed and we have this factory up and running.
14:37We've got a lineup.
14:38Why don't you tell me, because I missed that one.
14:41You have your own factory.
14:42We built a factory in, you know, outside of Oregon State and we have the ability to ramp
14:48up to 10,000 units a year because we saw that demand and we didn't want to miss it as it
14:54was coming in.
14:55But right now we have handfuls of robots and we have a lot of interest from automotive
15:02companies, retail, grocers, across the board, anybody who's running these kind of facilities.
15:10And anyway, we're struggling to just to figure out where do they go next until the fall and
15:15we'll be able to open it up.
15:18Because the fall is when you'll start making your next gen.
15:22How many do you think you can produce next year versus that 10,000 capacity?
15:28Next year we're hundreds and the year after thousands.
15:30So that's sort of the ramp that we're looking at.
15:33The good thing is the factory is in the U.S.
15:36My engineers are nearby.
15:37I don't have to put them on planes anywhere.
15:40And it is, you know, we're capable of ramping up.
15:44It's not capex heavy.
15:47It's largely assembly, test, repair, that kind of thing.
15:51So we can ramp up pretty quickly.
15:52If I need a second one, I could get one up and running fairly quickly.
15:58Building wise, raised hundreds of millions of dollars, right?
16:02About $180 million, not too much of that so far.
16:08Okay.
16:10Like how much is that a constraint on building out this vision that you see for many different
16:17types of workplaces?
16:18Yeah.
16:19So the next we are raising now, we're in our series C right now, and that will take us
16:24out the next couple of years until we hit that 10,000 unit.
16:28But the good thing is because we're deployed already, we are getting the very first robot
16:34as a service contract, multiyear contract for humanoids, so we're getting a monthly
16:39revenue already on those units and units that we put on after that, we'll have that incoming
16:45monthly revenue.
16:46Do you know what the discussion is that employers that you're testing with or putting these
16:52robots into their facilities, what they're having with their employees, because I know
16:57what we've talked about, taking tasks, not jobs.
17:00But Amazon, for example, second biggest, largest, second largest private sector employer in
17:06the U.S., 1.5 million jobs.
17:10I've talked to hundreds of these employees over the years.
17:12Some of them, a lot of them see robots and they're concerned.
17:16So what is the message there that is believable?
17:21Well, that's believable.
17:22I think it's reality.
17:23There is 1.1 million unfilled jobs.
17:26They can't find people to do that.
17:28So what does that do?
17:29It taxes the people who are there even more.
17:32And they're ending up doing this repetitive work for longer parts of their day when maybe
17:38it used to be a two-hour job, they're doing it for four or six hours.
17:42And if you see in the video, it's literally moving a plastic bin from here to there.
17:47It's mind numbing.
17:49And so it's not that people are seeking those jobs, they just can't fill them.
17:54And it's a retiring workforce, they can't refill.
17:58So they have a real problem.
18:01So far, everything that we, all of our interactions have been very positive because it does give
18:08you the opportunity to upskill that worker and put them into a higher productive job.
18:12We'll come back in five years and talk some more.
18:14I'm getting evil eyes.
18:15We're out of time.
18:16Thank you so much, Peggy.
18:18Thanks, everybody.