Joe Fuller: Amazon is one of the largest employers in the United States, with roughly two hundred and seventy-five thousand employees spread across the country. In July 2019, Amazon announced an ambitious goal to upskill a third of its workforce in five years, citing the centrality of training in the upward mobility of their employees. Welcome to the Managing the Future of Work podcast. I’m your host, Harvard Business School professor and visiting fellow at the American Enterprise institute, Joe Fuller. Today I’m speaking with Ardine Williams, Vice President of Workforce Development at Amazon’s HQ2 in Virginia. Ardine will tell us about the variety of training programs at all levels of Amazon and how the company engages with localized workforce development ecosystems around the country. We’ll also discuss Amazon’s innovative approach to building new pipelines of talent and the calculus they use to measure the success of those programs. Ardine, thanks for joining us.

Ardine Williams: I’m delighted to be here.

Fuller: Ardine, when people talk about the use of automation and how it’s affecting the workforce and the prospects for workers, they very often cite Amazon as an example. How does the company view the deployment of technology and how it’s affecting your workforce needs?

Williams: Rarely does technology replace a job wholesale immediately. What we experience—and we see it, whether it’s in our warehouses or in our office workers—is that, as technology comes in, it changes specific aspects of the job and creates either new opportunities for people to work or changes the nature of the work that they do. In the warehouses with our middle-skills workers where we have deeply deployed technology, what we find is that it creates opportunities for workers to learn new skills. In our robotics warehouses, workers have the opportunity to become trained to actually go out onto the floor where the robots are—to be able to clear the lanes where the robots operate, to reset roadways, to perform basic repair and maintenance on the robots—and then, in the case of becoming more skilled, to move into the mechatronics field. Mechatronics is a field that combines mechanical engineering, electrical engineering, optics, software.

Fuller: So it’s not a Hollywood character from one of these movies with the Transformers or something.

Williams: Not yet.

Fuller: In July, the company announced a program to upskill or train 100,000 workers. Can you tell us about that program and the motivations for it?

Williams: So we announced that we would upskill 100,000 employees—that’s about a third of our US workforce—between now and 2025, backed by an investment of $700 million. And we believe that, as a large employer in the US, we have a role to play in creating great jobs. And, as we think about great jobs, there are really kind of three things that comprise them. The first is great pay. We announced and implemented last year $15-an-hour minimum wage—and just for context, that’s $30,000 a year without overtime. The second is robust benefits. So from day one, our warehouse workers have access to full medical and dental, as well as innovative programs like 20 weeks of parental leave. We believe that great jobs offer the ability to learn skills, to build on those, and to create a career path. The upskilling announcement was a recognition of a lot of the work that had been going on internally—so pilots that had been very successful that we doubled down and increased in size and scope.

Fuller: Amazon, like a lot of its peers in the high end of technology, is renowned for doing lots of experimentation. But we think about that as experimenting with offers or feature sets or things like that, not workforce development.

Williams: As a company of builders, we are exhorted to experiment, which for me has been a tremendously enriching experience. It’s also scary, because you have to decide what you’re going to try and then be willing to …

Fuller: … live with the consequences.

Williams: … willing to fail. And so the experiment that I’m particularly proud of was on the apprenticeship side on workforce development. So, when I first started with the company, I was working in Amazon Web Services, AWS, and talent acquisition. And the business was growing rapidly. And we were struggling to fill jobs for cloud-support associates—the folks that worked with our customers. And what we found was that, when we interviewed veterans, they were a great cultural fit, but they lacked the specific technical skills. Veterans were doing great on the cultural assessment, but we couldn’t get them past the technical bar. And so the initial option was, well, we can kind of down-skill them into a lower role, but that really didn’t really make a lot of sense. And so we knew that we had good onboarding programs, and some of our managers had very robust skills plans. And so we reached out to the Department of Labor and said, “Hey, we’re interested in building a registered apprenticeship program.” And they were super. This was in 2016. The professionals there helped us build out the program. And we began with 15 veterans in the first cohort and grew that program to the end of the year to 63. And then the question comes in, “Okay, how do you measure it?” So we were looking at, saying, “Are people making it through the program? Are they performing comparably to external hires? And are we retaining them?” And when we looked at those metrics, the program was successful. And then the next challenge was, how do you actually scale this in a business that’s growing at a far more rapid pace?

Fuller: And where do you stand today with that?

Williams: So at the end of the year we had about 500 apprentices who completed the program.

Fuller: Exclusively veterans?

Williams: The vast majority are veterans. We opened up to military spouses as well.

Fuller: So that’s an example of an innovation to source some new talent into the company. What about, you mentioned earlier, investing in people’s prospects for advancement?

Williams: There’s a wide variety of programs. For example, the Amazon Technical Academy trains employees who don’t have a technical degree and provides them with the training that would be equivalent to what they would receive in a computer science degree and qualifies them to take on a role as a software development engineer [level] one. The Machine Learning Institute trains people who already have a computer science degree in what they would normally receive at the post-baccalaureate level in machine learning or AI, artificial intelligence.

Fuller: Do you develop your own curriculum for those things? Are you partnering with academic institutions? How are you ensuring that the material goes beyond what’s relevant to Amazon to being relevant for career choices outside the company?

Williams: It’s a combination of internal training, as well as some external providers. For our Career Choice program, which is focused exclusively on our hourly workforce, the vast majority of that training is actually developed and delivered with external partners, predominantly community colleges. Career Choice is a program that provides training for in-demand jobs in the local community that pay more than Amazon does that offer a career path. And that program trains for right now about 37 distinct job types in five general families.

Fuller: How are you judging that jobs are in demand?

Williams: We began using the Bureau of Labor Statistics’ job census data. And as you know that’s rearward looking. It’s about a two-year look. And you can see some unusual blips in it. So a new business comes into town or a warehouse starts, and so there’s demand. And so that demand that you see historically may not currently exist. So we added to that Burning Glass’s jobs analytic data. But there’s still a judgment piece to it that says, “Are there enough roles that someone with the training has a good chance of transitioning?” And that’s where what I like to call “conveners” come into play, like an industry association. In Dallas-Fort Worth, the Dallas-Fort Worth Aerospace Coalition was very important in helping us recognize that there was a significant demand for advanced manufacturing technicians. And there were programs in the local community college, but those programs weren’t fully subscribed. And part of the challenge was they were full-time programs. And so you had an adult learner in a situation where they had a job with good benefits, and they were faced with this decision of, “Well do I quit this job to go take the training to get a better job?” And in many cases, we don’t have that luxury—to stop working to go to school. So we partnered with the community colleges to make the program part-time, brought it onsite into our fulfillment centers, and offered it at periods of time during the day so it was shift friendly. And then, as our employees completed the program, helped work with the coalition to schedule job fairs. And that program has been very successful. Through a completely unrelated meeting, we learned that there was the same demand in Southern California. And so we’re in the process of building the Advanced Manufacturing Technician program in Southern California to support demand there.

Fuller: But across locations, you’re looking for jobs that are available locally.

Williams: That’s correct. The point that we train for local jobs is incredibly important. We know, the Janesville example is a painful lesson …

Fuller: … so, the Janesville, Wisconsin, GM plant …

Williams: … where workers were trained, but the jobs for which they trained weren’t available locally. So you were asking people to train for a new role and then to uproot from their support network—family, friends, community—and move to a new location. And that creates friction. And in order to help people move up the wage scale, in order to help people on that career ladder, removing friction is incredibly important.

Fuller: Ardine, this description really turns the logic that we so often hear from employers on its head, because when you talk to employers about investing in improving the skills of their incumbent workers, you very often hear a reluctance to do that because the worker might go work for a competitor or leave, and then I’ve made all these investments that have walked out the door. You’re viewing it in the opposite way—that you’re happy to do that. Could you explain why? Because I don’t think that our listeners view Amazon as an institution that’s dedicated to educating people for the sake of educating them.

Williams: This isn’t philanthropy. It makes absolute good business sense. We talk about creating great jobs. So it’s an attraction mechanism. It’s a retention mechanism—folks are eligible for the training after one year. And we’re a retailer. In that example I gave you in Dallas-Fort Worth, those employees who train with us and become advanced manufacturing technicians and graduate to a job that pays 10%, 15%, 20% more than we do add to the local business community. They fill jobs that would otherwise go unfilled. Production gets met. New employers come into the community. Their discretionary income goes up. And we’re a retailer, which means that they’re going to continue to shop with us and, we hope, think of us as a good provider in the community, a good partner in the community. When you have employers with similar requirements, it benefits us to work together, because it makes the pie bigger. And employees have a choice. Candidates choose based on the type of work they’re doing and the culture of the employer. And while the person who interns with me may not ultimately come back to work with Amazon, they’re going to work in the local community. And someone who interned with them may come to work with me. And so that collaborative piece allows particularly smaller employers to work together to help increase that labor pool.

Fuller: How does that work? Because certainly when we look at models of skills development in Europe, what we see is that in a lot of countries—particularly in central Europe, the German-speaking area—are very strong organizations led by employers to define skills, update them, interact with the education community. Outside a few industries in the United States, that doesn’t happen.

Williams: It varies by site. In many locations, there is a convener, like a workforce board or an industry association. The Washington Technology Industry Association is an important convener in the state. There are state initiatives, like Career Connect Washington, that help provide the incentive.

Fuller: In Washington State.

Williams: Washington State, yes. They’re also, though, where you have rallying groups like veterans, for example. But there are other pieces to it, and you touched on them. It’s important that, as an employer, we’re signaling and communicating regularly with the training providers. So for example, in the case of a community college, if they’re training data-center technicians, it’s important if we require Unix, for example, vs. another operating system, that Unix is included in the curriculum. If their graduates are struggling with particular concepts upon transition into the workforce, we need to provide that feedback. If they don’t have someone who is working with state-of-the-art technology, what can we do to help bring that to the community college— whether it’s through curriculum consultation, helping design labs, or providing someone from our staff who can come in and do a guest lecture? It’s really that collaborative piece that makes it possible.

Fuller: One of the things we often hear also from employers is that they find the idea of working with local post-secondary educational resources appealing in principle but very hard to implement—that the education providers often don’t have appreciation for what constitutes state of the art, maybe have trouble getting new programs approved by state-level departments of education. What have been your experiences in working with those educational resources, and what are a few steps or practices that you found help advance those relationships and get that partnership to create the opportunity for people to upskill?

Williams: Community colleges are incredibly flexible and an important partner. Every state is organized differently, and what works in one state may not work in another. But I think, if you were to ask community colleges, I suspect that they might come back and say business is pretty hard to work with, too, because they don’t understand …

Fuller: They do. I’ve asked.

Williams: … they don’t understand what it takes to change a curriculum. And, oh, by the way, I can’t change a curriculum three times during the quarter. You guys need to figure out what you want and tell me what you want. So I think that there’s opportunities on both sides, and when we approach it collaboratively …

Fuller: … it sounds almost like you’re applying classic supply-chain management principles to this dialogue—that you’re giving feedback to the skills provider about performance, that you’re helping them upgrade their technology, and all the other steps that any company takes in dealing with any supplier of an important part of their value proposition.

Williams: Talent acquisition as at its most fundamental level is really a supply-chain challenge. It’s about what are the capabilities you need, what are the specs for this job? What are the things you need to do to help create that competency and capability? And then, what’s the quality as people move through the system? But because it involves humans, it isn’t quite as straightforward as parts.

Fuller: So how do you think about the future of the workforce and try to get out in front of what your skills profile is going to have to look like in 2025, as opposed to next quarter?

Williams: While we can’t say with any certainty what jobs will look like in 2025, we can look at the technology today and the impact it’s having and make some pretty good predictions about the kinds of skills that we think people will need. So, for example, if you think about workflow automation—so these are, they’re frequently called “bots” that will do simple automated processing for resolving issues on a time card, for example, or approving vacation—we know that the people in HR who are currently doing that work manually or who are working with that will likely need to understand and interface with an automated workflow program—“How are we thinking about the logic in the program? What does the quality control look like?”— because they’re going to have to come back and understand the corner cases that get kicked out of the system so that they can address those. And so I know that the line between tech and non-tech, or between STEM and non-STEM, is blurring. And so the more that I can do to help educate that workforce in technology, the more prepared they will be to deal with the changes in their job that are coming four and five years down the road. I think the one piece is this idea of experimentation. And that was a big change for me when I came to the company—this idea that it was okay to start with something scrappy, to experiment and iterate. And I didn’t have to go get approval, for example, to run the apprenticeship program, at least the first couple. Having everything completely baked and completely planned wasn’t a path to success.

Fuller: The enemy of innovation.

Williams: Exactly, right? It’s “Perfect is the enemy of good or done.” Whatever we attribute to Voltaire, right?

Fuller: Right.

Williams: So I don’t need to remap all my jobs and understand what all the skills competencies are, and bring in a new training provider, and build a top-to-bottom training program, and then re-architect my enterprise resource application so that it can understand those skills and competencies. If I wait to architect all of that, I’m never going to build that program for the apprentices and figure out what works and what doesn’t. And we made changes through the program as we experimented in that first year. And we continue to experiment. And so I think the most important lesson for me is that starting small and innovating and experimenting is incredibly powerful. And that also means that it’s within the reach of companies that aren’t big. If you’re smaller, it’s, in many cases, easier to be innovative and scrappy.

Fuller: Our experience in talking to large companies and in studying them at the Harvard Business School is large companies generally do pilots; they don’t actually do experiments. And here you’re describing a very experimental mindset. So talk a little bit about what are the design criteria when you’re doing an experiment that affects prospective employees or employees, as opposed to doing A/B testing on a website.

Williams: An experiment, truly I don’t know the outcome. I may have a hypothesis about the outcome, but I don’t know the outcome. If I’m at the pilot stage, I need to have a pretty good idea that it’s going to be successful, and I’m in the process of refining details. When we think about the experiment and what constitutes success, we think about two pieces: both the input metrics and the output metrics. And so, on the apprenticeship program, what I was looking at was, “Hey, what are my conversion rates? When I bring people into the program, how many can I successfully move from the classroom-based training to the supervisor-on-the-job training and then through the phase gates that are in the journey-worker path?” And then I looked at how do those graduates compare in terms of on-the-job performance to people that we hire without coming through an apprenticeship pipeline. And so I had the opportunity to say, “Is this an effective pipeline? Am I equipping people with the kinds of skills that they need in order to be successful? Is it the right duration? Is there a cost benefit that makes sense for the company?”

Fuller: And when you’re judging effectiveness, can you share some of the variables you’re looking at, the specific variables that would indicate that this approach is superior to the historical approach?

Williams: So I think it’s important on the apprenticeship pieces. It wasn’t an “or” gate. Right? So I was adding …

Fuller: … a new …

Williams: … if you think about the back to our supply-chain analogy, I was qualifying another supplier.

Fuller: A new vendor.

Williams: And I was taking folks who were a good cultural fit that we assessed were capable to successfully complete the apprenticeship program, and then looking at whether I was effectively moving them through that training program.

Fuller: For something like Career Choice, would there be specific metrics you’re looking at for those incumbent workers that you’re creating new opportunities for to judge whether or not this is having the right impact?

Williams: For Career Choice, the metrics are similar. We look at our employees successfully completing the training program. And, again, the challenge there is these are adult learners who have full-time jobs. And so I’m asking someone, in addition to their 40-hour-plus work week and their personal family obligations, to take on …

Fuller: … fit in some more hours.

Williams: Fit in some more hours. And do some homework. And maybe dust off, maybe they need to go back and take algebra, for example. So it’s a big ask. But if I can, if I get them into the program, are they able to successfully complete? And both with the apprenticeship program and with Career Choice, we need to back up and say, “What are the capabilities and skills that they need coming into the program in order to be successful?” So math is a great example. Most of the programs that we’re training for in that job family requires some level of math. And sometimes folks, it’s been a while since they’ve been in math.

Fuller: Their math skills have atrophied.

Williams: That’s right. So let’s go back and provide that piece so that they can in fact be successful in the program.

Fuller: What about measures like “reduce voluntary turnover” or “higher employee engagement.” Do you track things like that?

Williams: We do. I think that that’s a ... because it’s a multivariate analysis, it’s tough to come back and attribute success to one specific program.

Fuller: Yeah, sure. No one who’s seriously considering leaving signs up for a program like this or ... so you have to be careful how you weigh it, but …

Williams: Well, what I would say is we do have people who are seriously considering leaving who sign up for the program, because this is a path. So last year, our most popular program was commercial driver’s license. And that’s because it’s a three- to five-month program. It offers a very significant increase in wage. You move to ... the average starting wage is about $44,000 a year. And it offers a lot of flexibility. I can work as a hostler in a yard moving trailers around. I can do short haul in the local area. Long haul, I can become my own ... I can own my own tractor. And so those are people who are thinking about leaving. And we’re very proud to be part of helping them on that career journey.

Fuller: And one of the few jobs that invariably appears in the top 10 unfilled positions in every state in the union.

Williams: Yeah. Commercial driver’s license, and IT, healthcare, are really our big three. So when we take a look at what job paths there are, we think about a career. And that can either be a vertical career ladder, which you would think about, “Hey, I progress up.” Or it could be something where I take a job and it provides a stepping stone into a parallel qualification.

Fuller: Ardine, one of the concepts that you hear regularly discussed in this areas is stackable credentials—the idea that, if we can build a system where workers can have a job experience or have an educational experience that will provide a credential that is recognized not just by their employer but by other employers, allowing them to begin to build momentum toward advancement into higher paying jobs. How does that reflect in what you’re doing?

Williams: I think it’s incredibly important, because those stackable credentials make it easier for an adult learner to decide the path they want to take and to learn a little bit about it and decide whether or not it’s something they really want to pursue. And those stackable credentials give them the flexibility to take the training and immediately go out and earn more money, or to continue the training and then transition out into a higher-level degree. In the IT space, when we think about it, one of our most popular certifications is the Computing Technology Industry Association [CompTIA] A+ certification, which is really for desk-side support. So it’s the people, when I get a patch and I can’t figure out what happened with my computer, I call them up and they save me.

Fuller: Classic help desk person.

Williams: And so someone can get that credential and transition out of the company from, say, $15 an hour to $17 an hour. Or they can take that credential and invest another, say, three to six months and become a computer-support engineer. And that takes them into the, say, $20-an-hour range. And then they can transition out of the company, take a job, and then continue that career path elsewhere, or they continue with us and get a two-year degree. And so those are really off-ramps and on-ramps.

Fuller: When you’re looking at local jobs, how do you distinguish between those jobs that actually constitute a springboard for further opportunities to grow? It’s not a dead-end job, because there are a number of jobs—an example that we’ve cited in our research is pharmacy tech—that have a lot of appeal prima facie but what we describe as a step-ladder job. It may be a step up from where you are, but it doesn’t lead much higher, as opposed to other jobs, which are really suspension-ladder jobs, where it’s no guarantee you’re going to go all the way to the top, but the prospect is there.

Williams: The reason it’s important that there is that path—and we really went back when we raised our minimum wage and evaluated jobs—we found that we had had some roles that people were interested in, where there were local programs available, that were really cul-de-sac jobs. And a great example of that is, we talked about the pharmacy technician. This is a job that the requirements for it don’t path into any other healthcare job. And, as a result, people invest time and effort and money into training for that and then don’t really have that next rung on the ladder available. Where, in other healthcare areas—whether it’s medical records, coders, medical assistants, licensed practical nurse—the training that you take for each of those applies to the next.

Williams: And as we went back and analyzed those jobs that we were no longer going to train for because they didn’t pay more than we do, we found that there were some of those cul-de-sac type jobs in there.

Fuller: Some cynics say programs like this are rooted in either the very tight labor market—so companies are making moves now because very low unemployment rate, hard-to-fill jobs, and when the economy softens and there is more available labor, companies are going to back off these investments—and they also say that companies, particularly large companies like Amazon, are doing this to burnish their brand a bit. How would you respond?

Williams: We’re a large employer, and regardless of the availability of talent, creating good jobs, which we talked about, great pay, robust benefits, and a career path are important in attracting and retaining talent. Even when, at the nadir of the Great Recession, people did change jobs—maybe not as frequently as they do now—but having programs where you can, in fact, upskill is incredibly important as an employer. As far as burnishing brand, I think that every employer has to think about brand, because who you are and what you represent from an opportunity perspective and what it’s like to work at the company is what draws talent to you, ultimately. And people have choices.

Fuller: Well, Ardine, thanks for joining us on the Managing the Future of Work podcast.

Williams: Thank you for having me. It’s been great to talk with you.

Fuller: Thank you for listening to this episode of the Managing the Future of Work podcast. To find out more about our project on the future of work, visit our website at hbs.edu/managing-the-future-of-work.