The pandemic has radically transformed the U.S. working environment. Knowledge workers are stuck on endless Zoom calls and Slack chats; restaurants are having to rely on deliveries; and warehouses are being automated with socially distant workers keeping things ticking along. But will this be the way forever?

Many companies have said that they plan to invest more heavily in automation as the world begins to open back up, with companies relying on inflexible legacy technology likely to be left in the dust. "COVID has given us a new business case for innovation," said Accenture CTO Paul Daugherty. The future will see businesses investing in new technology, but according to Daugherty, that will create new roles for people, rather than just replace humans with machines.

Protocol recently spoke with Daugherty about the future of work, where automation can work alongside people, and how AI systems can perpetuate bias with poor data.

This interview has been edited for length and clarity.

So are we all going to be working from home in the future?

I really don't think that will be true for most companies. For a lot of people, it's really not "work from home" — it's more like living in your office, because you're working all day long. A lot of people like it, it suits their lifestyle. It allows them to balance family needs. It's great for a lot of people, but not for everybody.

We believe the future is really about redesigning work, that everything's virtual, so you never need to be physically together. We're focused on virtualizing the work itself so you don't have to be together, and then redesigning the workspaces so they suit how people will work together in the future. There'll be more convening spaces, more flexibility around how people come and go in the office. That's what we're prepared for. We're doing a lot of work reimagining the way we work. We think it will be substantially different.

The other part I think is important is dealing with the human needs. Because everyone's working from home, our productivity is up. But a lot of that's because people are more focused — in some cases they're working harder. In that environment, we're really concerned about our people, making sure they get back to taking vacation, they take breaks, they don't get too consumed, they get extra support dealing with their families that they need. We have a program going on called "It's OK to say, you're not OK." We didn't want to do social distancing with our people, we didn't ever use that term. We call it physical distancing. So we'll be physically separate, but with greater social intimacy by looking out for each other more through all these types of programs.

What is likely going to be automated first as a result of the pandemic?

I'd highlight four things. One is the acceleration to the cloud that was already a big trend. Companies were moving to cloud services fast before COVID, and it's accelerated in a dramatic way.

The second is a lot of collaboration and virtual work, a lot of focus on companies deploying new tools and work practices.

The third is security. There's a very elevated level of attacks right now, because of people working at home, working in a different way. There's new vulnerabilities that are exposed. So there's a really big need for automation, things like zero-trust approaches to security, better endpoint security, things like that.

The fourth is around data and AI. We're seeing a tremendous focus from companies on data: I need better data on my supply chain and what's happening because the patterns are different than they were; better data on customer purchasing and demand patterns; better data on what's happening in my manufacturing facilities. All that, wrapped around artificial intelligence to better anticipate and predict what's happening.

We're seeing three years of digital transformation happening in three months, and that's really causing this dramatic change.

It must be a tough thing for companies realizing they need to invest in automation to square that with the fact that we're entering probably one of the worst recessions we've ever seen.

It's tough. That's the big question of the day, but this crisis is different than others, in that there was this exponential change in technology happening, and this recognition by companies that they needed to digitally transform. Everybody was at different stages of transformation entering the crisis. Every other crisis was about just clamping things down, cost efficiency and survival — in this one, companies we're working with are investing, but in very different and smart ways.

Companies are looking for ways to continue to invest because they know they have to, they can't just stop because they were in the midst of transformation. Consumers are going to behave differently coming out of this. There's no choice but to figure out some way to transform. We see companies making tough decisions on what they cut versus what they continue to focus on. If you're a retailer, you have no choice but to assume you could have to drive more of your business online. You could have to do more curbside pickup. Companies are investing in those types of new capabilities. Investment levels overall are down, but there's a very careful process happening where companies are looking at how they still keep focused on the key elements of their transformation that they need to continue.

As large organizations look at technology investment, they have a lot of what we call "technology debt" — old legacy technology. That's very expensive and costs a lot to maintain. Moving to the right cloud sometimes allows companies to pick up a cost advantage of moving some old technology to the cloud, and then use some of that savings to transform and invest in new capabilities. It's almost like a zero-cost-transfer transformation. A lot of our clients are looking at it.

Have your thoughts on AI changed at all since you wrote your book? What's the ideal scenario for humans and AI to work together in the office?

My co-author and I were talking about this just the other day. I think our view has been reinforced since we wrote the book. There's just a lot more examples, companies really taking that "Human + Machine" view in what they're doing. I think the other thing we've seen since we wrote the book is the pace of AI adoption has been increasing. A lot of it's driven by the hyperscalers, by the Amazons, Microsofts and Googles.

Companies are really interested in getting their data to a single place where they can have the powerful tools to do analysis and drive new AI capabilities. So a lot of the cloud migration we're seeing isn't necessarily just getting the applications to run in the cloud and compute-level efficiency, it's the innovation capability of getting the data there and then using the amazing innovation tools that the hyperscaler clouds have to drive new AI capabilities.

For example, one of the life sciences clients we're working with, they're moving a huge part of their IT estate to the cloud. And the reason was driven by the ability to do better clinical trial analysis, better R&D, drug discovery — things that they felt they could do once they got things to the cloud.

AI's role is to replace repetitive tasks, which presumably leads to efficiencies that could reduce head counts. Accenture employs more than 500,000 people right now — if AI continues to proliferate, will that always be the case?

It'll be for different things, but I just have to legitimately say I'm not sure. I don't know, but what we see happening is AI is replacing the dirty, dull, dirty and dangerous stuff. But then we see that creating a need for other jobs. The question today is, are we preparing people doing the repetitive tasks with the skills they need to do tasks that require something more? I still think that there's a gap.

I do believe there'll be a lot more jobs in new categories because we've experienced it. As we've been heavily automating our business, we're growing at high levels. For example, the way we do testing, we used to have people entering keystrokes and testing things five-plus years ago. You have very little of that now because the testing can be automated. What you have are testing engineers, test-design and quality engineers, but we have far less people doing testing and test execution. And that plays out across every profession: As you automate to do things more efficiently, how do you retrain the people?

Do you have concerns about implicit biases existing in AI systems, given that algorithms are only as good as the data that's fed into them?

Absolutely. I think it'll be a necessary question to ask for a long time going forward. I think the problem and the solutions are both human. AI isn't the culprit here, it's the way that humans are applying AI. AI doesn't have bias in it. I track new examples of biased AI that I come across almost every week, there's still tons of examples of it. It tends to be what I would call poor or lazy applications of AI, poor training data that were used, poor testing. We're participating in things like the Partnership on AI and other organizations out there to solve these types of problems.

Internally, we've invested in a lot of R&D, and did two things: One, we call the "AI fairness toolkit," which is a toolkit for deep-learning algorithms that allows you to gauge the bias you have along a number of dimensions, so you can test for bias before you deploy the application. And we have another thing called "teach and test," which is a methodology for how you should teach AI and then test to make sure that it's behaving the way you want to. We believe those two capabilities, when used properly, can lead to applications of AI that you can be comfortable don't have bias. If you just take the data you have, develop an algorithm, and put it out there, it's likely to have adverse implications and bias in it, because most datasets reflect a bias.

I think bias in AI, in most cases, is very avoidable, but it takes careful effort and people who know what they're doing. I think you should keep asking the question forever — it's not going to be like we've solved this next year. Like any other technology, it can be deployed poorly or deployed well — it's something we need to be continually vigilant on, especially with some of the issues going on in the world this week.

What do you see as Accenture's role in the Black Lives Matter protests that are happening right now?

Our CEO put on LinkedIn a memo sent to all employees that myself and all of our leadership signed. We have a very clear, unambiguous position that we have zero tolerance for racism in any form. We announced a three-point action plan that we're going to pursue internally and in the communities where we operate to stand up for what we believe in and actively make sure we're doing the best we can internally. We think we're doing pretty well, but we need to do a lot better.

We'll work in the communities we operate in to make sure we're doing what we can and influencing a better environment where everyone can feel safe, regardless of color. And that's tremendously important to us. It's been incredibly disturbing to us, to our employees, and so we took a clear stand that's zero tolerance for racism. We reject racism in any form.