In our new series AI Dimensions, Bijal Shah, CEO of Guild, discusses a dimension of AI that many in tech overlook: its effect on frontline and other workers outside of tech. As AI reshapes the economy, access to the resources needed to truly benefit from it is becoming a new axis of economic divide, layered on top of an already K-shaped economy. Bijal shares what she's seeing on the ground through Guild's work with employers like Amazon, where frontline warehouse workers are being reskilled into technical roles, and discusses why "give everyone a license" isn't an effective strategy for closing this gap. She also touches on what this all means for the next generation entering the workforce, and why the durability of trade and frontline jobs may surprise knowledge workers. Watch the recap here.

‍

10 Key Takeaways

1. Frontline reality: Employers already automating roles are asking which jobs come next - for example, turning warehouse packers into much-needed robotics technicians.
2. What employers need: Employers want two things now: fast pathways into business-critical roles, and a way to get people to rethink existing workflows to take advantage of AI.
3. Driving behavior change: Getting people to use AI well is a change management problem, and empowering peers to be internal champions is a powerful tactic to help an organization adapt to change.
4. Workflow redesign strategy: Nobody can redesign work they don’t do, so pair the person who knows the job best with someone who can step back and see the opportunities.
5. Early-career workers: Skilled trades are necessary but not sufficient on their own, and a degree still teaches the dot-connecting that no single-track path does.
6. Durable jobs: Roles that require a license and a human signature are a safe bet for the next decade.
7. Designing for frontline workers: Assume mistrust at the start - go talk to the people whose work your product is about to change in order to build trust and understanding.
8. Key takeaway for tech executives: Early adopters will try anything, but trust is low, so even a great product has to earn its place if it wants to stay.
9. Message to policymakers: Retrain workers before the layoff, not after.
10. Overall outlook: AI can create better jobs in a shrinking workforce, but it requires collective action to realize its positive potential.

‍

What does AI implementation look like inside a warehouse or a hospital right now, and how are employers supporting their workforce through the transition?

Guild works with large employers in this country, and we're focused on upskilling, reskilling, and redeploying their workers. As we think about where we're headed macroeconomically and what jobs will exist in the future, this is a ripe time to talk about the right ways to think about retraining, reskilling, upskilling, and redeploying. We work with 70% of the industries represented in the US economy, from health care to advanced manufacturing to retail. We help employers drive their future-of-work strategy by building the talent they need inside their own organizations.

We're seeing some interesting things. Warehouses, distribution centers, and health care already use a lot of robotic and advanced technology. When you talk to those employers about the disruption coming, the conversation takes a different lens than it does at businesses with a high proportion of knowledge workers. They already see how this technology can help evolve and adapt their workforce, so we spend our time on what the jobs of the future are.

A good example is Amazon's robotics division. They need more reliability maintenance technicians and engineers, and they can't just go hire them. So they're taking warehouse workers who might do packaging or traditional maintenance and skilling them into reliability maintenance technician roles. Those require robotics, electrical engineering, and mechanical engineering competencies. There aren’t enough people on the open market with those skills today, so they need to build the capacity in-house.

What are employers asking you to do today, at a tactical level, that's different from one or two years ago?

For the sectors that are further along, the workforce strategy is clear, so we work with them on two dimensions. The first is building career pathways and pipelines into new jobs we know are persistent and durable, and finding innovative learning providers to help. You can't go become a reliability maintenance engineer at any university today - it isn't a degree you can get, and a degree would take too long if you're trying to get people into these roles fast. So we innovate alongside employers to create those pathways and that competency as quickly as possible inside their organizations.

The second is that many employers want their people to use general-purpose models to work more efficiently, and to rethink their workflows so they can use skills-based or agentic technology. That is a very different problem: how do you drive behavior change inside an organization? We started off with "give people the tools and everything will be fine," and it's become clear that doesn't drive utilization. It works for people who are savvy, naturally curious, or early adopters, but most employees don't fall into that bucket. It's not about providing tools. People already have access to tools. It’s that they don't want to use them, or they're not sure what to use them for. We're getting a ton of requests that say, "We've deployed the tools, which is great, but adoption is really low, or we don't feel people are getting the maximum benefit. How can you help us drive behavior change?"

Everyone, it seems, is grappling with that same problem. So what's the answer - how do you drive behavior change?

I see a lot of leaders who don't use the technology themselves offering examples of why it will be good for someone else. It lacks the authenticity of having seen the benefit yourself, so it actually reduces credibility inside the organization in a way that is hard to come back from. You're now operating at a trust deficit. I work with a lot of heads of HR and talent leaders, and I tell them that before they push this technology out, or even enable it in their orgs, they should understand its bounds themselves. There's already mistrust in the system, and employees will trust you less when they realize you aren't using it yourself.

One of our core theses is that behavior change around AI is no different from general change management. To bring along late adopters or naysayers, you need their peers to prove the technology is powerful and not to be feared - so we run AI academies, broken into three persona types: leaders, technical people who use the technology to generate code or train their own models, and a broad-based group who need the tools to work more efficiently or rethink how they do their work.

We drive peer-to-peer learning, because the fastest way to get people to want to adopt this technology is to see how their peers use it, and to see that it can help them do their job more effectively, rather than being scared about it disrupting their job or leaving them unemployed. It helps when this happens inside one organization. For example, we rolled out a 65,000-person academy across an entire health system, to the whole employee base in one go, which is pretty unheard of. It's been phenomenal to watch people learn from one another, including across divisions. Someone says, "I applied it in finance this way," and someone in operations realizes the same thing applies to them. The tried-and-true principles of change management still hold even when the technology is novel.

How do you distinguish an adoption or training problem from a workflow that needs redesigning? Who should own that redesign?

I came from management consulting, and I think it's really hard for someone who doesn't know how the work happens to walk in and tell you how it should be redesigned. At Guild we've taken a two-pronged approach, because you also need someone who isn't jaded by the existing process, who is willing to step back and see the picture more clearly. So inside our own business (and I'm starting to see other businesses do this), we pair the people doing the work day to day with someone willing to take a step back, and we co-build what the new version looks like. It's a new muscle that hasn't historically been exercised.

You're asking the person who knows how the work happens to divulge that information on the promise that it will improve their workflows and they'll still have a job on the other side. We're navigating that one by one, and we have some awesome stories inside Guild, but it's tender. You need both sides to come together. That's why adoption is hard to see, productivity gains are hard to see, and workflows are slow to change. Without it, you end up with someone redesigning a process that didn't actually work or couldn't be implemented, or someone who is only willing to change their portion of the system and not the whole system.

I think about the hurdles we have as a 700-person organization compared with the 40,000-plus-person organizations we work with regularly. Our micro hurdles are their macro hurdles. This is so much easier if you're starting something from scratch, and most businesses aren't. So the hurdles are particularly acute in existing organizations, but we've seen this rodeo before.

How is AI affecting early-career workers, and what does it mean for those with a four-year degree compared with those without one?

One thing the generative AI conversation has done is normalize that multiple pathways matter, which has always been true. Not everybody needs a four-year degree, and there are great pathways outside college for people graduating high school. As we've seen this craze around software and SaaS companies, we've put on a pedestal the notion that you must become a knowledge worker, and the way to do that is to get a four-year degree. Our thesis at Guild, and mine for a long time, is that there are multiple pathways to economic prosperity in this country and we should value all of them. When supply and demand are off in a field, there's more money to be had than in a sector with an oversupply of talent, and we're starting to see that rebalance.

I always hear parents ask, "Should I even send my kid to college?" I think that's a little shortsighted. For young people who do pursue a degree, multidisciplinary learning is really important. To have context, connect dots, and solve problems across a domain, you need it, and college is one of the best avenues for it because you're forced to learn things you aren't interested in. A requirement says you need an English class even though you're a software engineer. I think we've lost the plot on why degrees exist. They were meant to drive critical thinking, contextualization, dot-connecting, and the ability to build a social network. Instead we're in a debate about whether four-year degrees matter or don't.

At the same time, we're all seeing a rise in skilled and vocational trades. If you look at where jobs are net growing in this country, it's manufacturing, advanced manufacturing, construction, and health care. The AI boom and data centers drive some of that, but not all of it. We're also trying to re-domesticate work that lived in other countries for a long time.

One of my worries with the push toward skilled trades, which is a great push, is that those jobs are physically hard on your body and don't last forty or fifty years the way a knowledge worker's job might. Every one of our healthcare employer partners is asking for radiology technicians and radiologists. Ten years ago people said we wouldn't need radiologists because AI would read the reports. Now we have a shortage. So it's important that we don't stop at "just get a skilled trades job." I think it's necessary, but not sufficient.

What is a durable job today, and what will be durable in the future?

I can't totally predict what happens ten-plus years out, but in the window of now to ten years there are pretty clear signs of what will be disrupted versus what will be durable. Many jobs in this country require licensure and certification, and the regulatory hurdles to change that are high enough that I think those jobs will stay durable. Take radiologists: there are many reasons for the shortage, and one is the licensing hurdle to become one. Another is that you still need a human in the loop. Even if technology reads the scan and makes the diagnosis easier, our regulatory environment still requires a doctor to sign off, and that will take time to change. Much of that regulation sits at the state level, not the national level, so you'd have to change it state by state. Combined with people worrying about job security and economic uncertainty in other sectors, policymakers won't be quick to make changes that displace workers. Where displacing the human feels existential, even if it might not be, I think those licensed and certified roles will remain.

I also see some large employers who went down the path of fully automating checkout reversing course a bit, because they've realized customer experience really matters. So I think you'll see some rebalancing in frontline roles. There will be fewer of those jobs, but they will still exist.

I do worry about what happens in the future to software development, product, marketing, and design. But I'm bullish that those roles will become more multidisciplinary. As the CEO of a technology company, I always have an endless backlog I can't seem to get to. Maybe the ability to generate and commit code faster and release features faster means that instead of 80 items on the backlog at any given time, we have 20. Inside Guild, what's a software engineer versus a manager versus a designer? Those things are melding together. People aren't necessarily equipped to be multidisciplinary, because we've siloed their jobs so much, so it will take retraining, reskilling, and even just exposure to that type of work.

And ten years from now? I don't know. I think regulation will start to catch up. In health care, for example, people will push for broader access and for changing the regulatory environment. I think a lot about how we're entering an era of continuous learning and relearning. I went to business school to pivot my career, and I think everybody is going to need to pivot on an ongoing basis. The art of the pivot will become critically important.

Many people in our community build products or run teams that touch frontline workers, directly or indirectly. What should founders and operators keep in mind when designing these products?

First, there is a general distrust of AI right now, given the FOMO rhetoric of "you're out if you're not using this" and some of the fear. That doesn't necessarily exist on the coasts, but I live in Denver. I was at dinner last night with 12 executives from pretty big companies, and 10 of the 12 said, "I might wait and see. I'm not sure. What are the ethics behind this technology? Who's making decisions about what I'm being shown?" There's a lot of distrust, and not just at the frontline level. You don't necessarily see it where companies are technology-forward or in a very competitive race against their competitors, but you see it in small and medium businesses. I see it in Denver constantly. So you're not starting from a place of trust. You're starting from mistrust, and how you educate, design, inspire, and motivate is different when people start there.

Second, get out and talk to people. Ethnographic research should be in its heyday right now. If you want to understand why people aren't adopting the technology, what the use cases are, or what would be valuable, get as close to the customer as fast as possible. That's more important now than ever, because those are the people who need to buy into the fact that level-one support tickets should no longer be handled by them, and that they still have a place in the world on the other side of that.

What's the most important thing for tech executives to take away from this conversation?

Remember you're operating from a trust deficit, not trust. The way you build and design things when your audience trusts you is very different from when it doesn't. Early adopters trust things, so they'll play with them, try them, and use them. But as a country we're at the lowest level of trust in any institution, including employers, that we've been historically. That changes the psychology of how you interact with customers and how you get them to want to use the technology, no matter how good it is.

How can non-corporate players like a foundation support Guild's connectivity to the broader corporate ecosystem, so more people have access to these services?

McKinsey put out a study two weeks ago called Workforce in Motion; the premise is that we used to be in a war for talent, with a scarcity mindset, and now we're in a war for mobility. The individual worker needs it, the employer needs it, and we as an economy need it. Guild has been trying to prove that this is possible. We've helped 100,000 individuals change job families and move from one type of work to another, and there are 10 million+ people who need that. I think it's a win-win-win for employers, their employees, and the broader economy. I would love help in amplifying that message.

We're starting to see it work. About twelve months ago we stood up a public sector team, not because we wanted to go into states, but because we wanted to redeploy public sector dollars and braid them with employer dollars so the dollar goes further. Everyone told us working with states, cities, and the federal government is slow, and in twelve months we closed four states, which is pretty unheard of. State deals always start small, but there's a lot of potential behind them as we keep proving what we can do. The same applies to nonprofits.

What do you want policymakers to be thinking about with regard to AI regulation and frontline workers?

I won't focus on ethics-related regulation, because a lot of discussion is already happening there. I'll take the premise that some incumbent workers inside the workforce now will need to move into new types of jobs, and that we will have displaced workers who get laid off. We're already seeing this. They need to find new occupations or opportunities.

I wish there were more conversation about taking existing dollars and freeing them up or redeploying them for those two populations. The first is incumbent workers at risk of being laid off, who need to be retrained and reskilled so they can move to a new occupation when it happens. The second is displaced workers who have been let go and don't have the economic means to retrain into a new occupation that exists in their local or regional geography. Those dollars are hard for workers to access, hard for states to deploy, and hard to track outcomes on. We have a lot of wasted spend in this country because there isn't transparency, accountability, or efficacy. I wish more people were focused on what works, and on policy changes that could redeploy those dollars into programs that work.

When a worker gets displaced, they're less likely to take on training, because their only goal is to get a new job, and they don't have the economic resources to spend on retraining. By the time they're displaced, it's too late. We should spend more time on incumbent workers, making sure they transition to new occupations in a timely fashion, so that when they are displaced they can move from tech to health care, for example, in a reasonable amount of time.

Overall, are you optimistic or pessimistic about where we are headed?

I worked at the MIT Media Lab about fourteen years ago, and there's a professor there who uses computer vision and computational models to evolve how health care works, as well as autonomous vehicles. I remember listening to one of his talks about what this technology could do to augment or offset what a radiologist does. Then yesterday I was in a meeting about how we get more radiologists and radiology techs in this country, and I chuckled, because when I sat in that room years ago, every investor and business person was raising their hand to say we'd just need fewer radiologists. Everyone in their bubble was talking about needing less of this type of worker. The reality is that fourteen years later, we have a shortage and can't get out of it fast enough.

So I am optimistic. If you get the technology right, there's a lot of upside. The population is aging and leaving the workforce rapidly, our workforce is actually shrinking, and current immigration policies are shrinking it further. When I think about where the work is, I'm optimistic that radiologists can get through more images faster, or that a radiology tech with a robotic assistant can do more scans faster. What does that mean for access to imaging that is preventative and gets people diagnosed earlier? Look at the Amazon example I mentioned. They need 10,000 reliability maintenance technicians and don't have them today. That's 10,000 workers who could have a new job that is probably more intellectually fulfilling than being a forklift operator or filling packages, which is what they're doing today.

What worries me is that not enough collective action is happening. I see all these commitments, a hundred million dollars here or a billion-dollar fund there - but the rubber has to meet the road. We have to use these resources to retrain, reskill, and drive mobility, and it's not any one organization's problem. It's a collective action problem. So I'm optimistic - and I'm worried.

‍

AI is here to stay - how it impacts the world around us is up to us. Hear from the Operator Collective community about dimensions of AI beyond tech.