New year, new tech, no problem

In this episode of The McKinsey Podcast, partners Michael Chui and Roger Roberts speak with McKinsey editorial director Roberta Fusaro about McKinsey’s latest report on technology trends, with a particular focus on the benefits and challenges of applied AI, cloud and edge computing, and bioengineering.

After, we’ll hear from McKinsey partner Brian Rolfes, who came to McKinsey firmly “in the closet.” But a chance meeting with a client showed him the value of being authentic—with himself and others.

The McKinsey Podcast is cohosted by Roberta Fusaro and Lucia Rahilly.

The following transcript has been edited for clarity and length.

 

AI is changing how business is done

Roberta Fusaro: We’re here today to talk about the recent McKinsey research report on technology trends. We don’t have time to talk about all 14 noted in the report, but thought we’d focus on three, starting with applied AI. How would you say applied AI has changed?

Michael Chui: In recent years, the one thing that has really changed is that the applications of artificial intelligence are starting to have an impact on business.

The definition of AI is not particularly clear, and that’s OK. I sometimes describe it as analytics turned up to 11. It is this ability for these systems to do the sorts of things that people used to have to do cognitively.

We’ve been serving thousands of executives over the past few years in their use of AI. And we do see adoption continuing to increase in applied AI. An increasing number of technologies are moving from the lab into business. That’s one of the reasons we identify applied AI as one of the 14 disruptive technology trends that is important right now.

Roberta Fusaro: Roger, what industries are using applied AI the most? Are there particular industries where we’re seeing spikes in use?

Roger Roberts: Quite frankly, what we find most exciting about this trend is that it touches every industry. While there may be a few that are further ahead in building their own capabilities, like financial-services institutions or large media and consumer tech players, we absolutely see the opportunity expanding rapidly across all sector boundaries.

Part of that is because it’s so widely applicable to many different kinds of data types. As we’ve seen over the last few years, AI has expanded rapidly from being focused on quantitative data, structured data, and textual data and into spoken audio, imagery, and all forms of video. Not just now for classification of those, and recognition, but also moving toward understanding question answering and language understanding, for example.

Roberta Fusaro: Can you provide an example of a typical use case, in a company or an industry, where AI is now being pushed to the foreground, even more so than several years back?

Roger Roberts: What could seem simple but is challenging in practice is the use of AI to support customer care. The notion of customer service agents that can be truly responsive to queries, whether in the form of chatbots on websites, or through audio and spoken-language input, or even supporting a human agent in the midst of a conversation, to bring them the right information at the right time to support that customer.

Understanding and responding to conversation, understanding even the customer’s mindset or the nature of their concern—whether they’re happy, sad, angry, worried—can now be recognized and understood in ways that allow for a better and more adaptive customer service experience.

Roberta Fusaro: Can you describe a little more how AI is able to sense and interpret emotion?

Roger Roberts: When we think about language understanding, the first step is in a conversation, understanding the words and translating meanings. But learning models can also make inferences about the nonsemantic aspects, the emotions that are built into a conversation.

That might mean they could understand that someone is excited, angry, confused, or upset, and convey that information to the customer care agent who is picking up the call. And when that agent maybe transfers the call to a second agent to escalate a problem, that information could also go along with it.

There are lots of different kinds of uses there, and it’s starting to become quite possible to make strong and accurate inferences about the emotional state of people in conversation. And that [information] used in the right way can allow for better and more effective service.

Michael Chui: Here’s another example of science moving into the workplace. There’s an artificial intelligence technology called reinforcement learning [RL], which has famously been used mostly in its early years to demonstrate the ability to learn video games. What we’re starting to see is that RL is now being used in business. There are some videos online about using RL to improve the design of boats used in the America’s Cup.

We also see RL being used to help train robots, where, previously, you had to instruct a robot on exactly where to move the hand and how to move the arm. Now these systems are starting to learn how to do it in a game-like way.

Roberta Fusaro: Roger and Michael, can you share an example of a company that solved a problem using applied AI?

Roger Roberts: I had a client trying to figure out how to best fulfill orders during the holiday shipping crunch in 2021. They had used typical rules-based algorithms that pointed pickers to places and had a queue of orders.

They realized that that was not the most optimal way of moving people through a warehouse, especially a lot of people who were, in many cases, temporary workers who’d only signed on to help with the holiday peak. So they designed an application that allowed for waves of workers to move through and use AI to guide them to the best pattern so they wouldn’t bump into each other, yet they could fill their orders faster and more accurately than they ever had before. It really showed dramatic improvements in productivity.

Michael Chui: I have an example that’s a bit of a parable. This is a consumer products company. Forecasts are incredibly important for consumer products companies. It will determine how much you’re going to manufacture. It’s how you’re going to prepare your supply chain, how you’re going to prepare your logistics for an upcoming season.

They used to use analytics for forecasting, but then started to use machine learning to try to improve the process. Truthfully, the accuracy of the forecast improved by a few percentage points.

But for folks who are listening and thinking about your own forecasting, would you like to have a few more percentage points? Absolutely, because that is a huge driver of business performance.

Stock-outs are a really bad thing. A part of this parable is that if you think about AI as not some magical thing where it’s data from Star Trek, but it’s improving your performance and analytics, there are lots of places in the business where that is a really material change.

Don’t get deceived by the flash. It may seem like the boring parts of business, but it is really exciting to see business performance increase.

It doesn’t have to be this amazing humanoid who shows up to your board meeting. It is oftentimes the core of your business—the most important places where value is created—where these technologies have the most impact.

It is oftentimes the core of your business—the most important places where value is created—where these technologies have the most impact.

Michael Chui

Busting machine-learning bias

Roberta Fusaro: What are some of the concerns that executives, companies, should be thinking about when it comes to applied AI?

Michael Chui: There are a lot. But let me start with one, which is an important one for people to think about. As part of artificial intelligence, a lot of people talk about machine learning. They’re roughly synonymous. I find the term machine learning [ML] to be slightly misleading because it’s really not machine learning.

It’s really machine training. We use data to train the model. One of the most important things is understanding your data. Because if your data has problems, whether it’s data quality or bias, that will affect the model that gets trained.

Take, for example, the use of AI to read résumés. You might say, “We’re just going to train our ML model based on who’s been successful in our organization previously because that should be a good predictor for who will be successful in the future.”

If it turns out that people who’ve been successful in your organization previously are disproportionately male, then you might end up training a model that disproportionately preferences male candidates.

Cloud versus edge

Roberta Fusaro: I want to shift to another one of the technology trends that we covered in the report: cloud and edge computing. In the most simplified terms, the cloud serves as storage for data, and edge offers a faster way to transmit data without taking up space. How do these two types of computing complement each other in business?

Roger Roberts: We’re recognizing that the incredible power of shared computing resources needs to be connected to sensors and actuators that are sitting close to the action, as in, inside the retail store or inside the warehouse, because data that’s being captured in real time may need to be acted on in real time.

Moving significant amounts of data back only to make small decisions can get expensive, both in terms of how much bandwidth it consumes in your networks but also expensive in time—what a technologist would call latency, the time between request and response.

As we think about that, we’re getting a better understanding of what should happen at the edge and what should happen in the cloud. What’s exciting is that it’s not an either-or trade-off. It’s a both/and. So now we look at this, and we say, “We want to make our edge computing as fast and furious as our cloud computing. We want to be able to bring new ideas, new capabilities, from concept to creation as fast at the edge as we can in the cloud.” That, to me, is why we think about this as cloud plus edge and not cloud or edge.

Roberta Fusaro: Can you provide an example of, say, a retailer or a telecom company that would get some benefit from this concentration on the combination of cloud and edge?

Roger Roberts: Take, for example, a retailer. One kind of datum that’s high in volume and can be captured with high fidelity is video tracking of retail shelves during the day.

If I’m in a store—let’s say it’s a grocery store—I’m tracking what’s being bought on aisle three. And I see that cereal boxes are moving off those shelves—with my video streams I can recognize stock-outs. I don’t need to send the entirety of that video back up to the cloud. What I do need to do is send a signal, a simple data element that says, “Wow. We’ve got a stock-out of Cap’n Crunch on aisle three.”

When I do that, it can trigger a replenishment action, whether that’s from the back room of the store, or it might be more Cap’n Crunch needs to go on the truck, and we get it overnight. As a result, those kinds of closed-loop decision-making applications can keep our shelves fuller faster, but they can also reduce the burden of stocktaking at the end of the day, waiting until the store is closed before I even know where my stock-outs are.

Closed-loop decision-making applications can keep our shelves fuller faster, but they can also reduce the burden of stocktaking at the end of the day.

Roger Roberts

Roberta Fusaro: The benefit is clear. What are the potential risks for companies that are thinking about this combination of cloud and edge computing?

Roger Roberts: It’s not so much a risk, but a barrier is making sure that you have the network in place to support this. It means that I have to have network reliability between my edge locations and my cloud data centers. Oftentimes, those are quite physically separated, and there’s a lot of network hops in between, so making sure that network is solid, reliable, and fast.

A second is making sure that my in-store or my edge architecture is built for agility. A lot of systems that were built in a prior generation were built with the expectation that they needed to run on their own and never be updated unless there was a major software release. What we’re now seeing is we need to deliver lots of small, fast changes into those systems.

Roberta Fusaro: I would imagine that because there are fewer and/or faster hops that security becomes less of an issue. Is that true? Or is that a bad assumption on my part?

Roger Roberts: Not necessarily. I would say security remains always a concern. The more attack surface area through systems that are exposed externally, the more we’ll have new and evolving risks. So I’m not sure we really change our security profile here.

We might allow for better and tighter monitoring, because I can have my in-store or in-warehouse systems managed and monitored on a more granular level. And I can make sure that that information’s being relayed into my security ops [operations] center and other monitoring and event management tools all the time.

Roberta Fusaro: Michael, do you have anything more to add on the cloud/edge topic?

Michael Chui: The only other thing I’d observe is that companies and executives have been concerned about security when you’re taking it off of your premises and putting it somewhere out there in some data center that you don’t own.

But more and more executives are asking themselves, “Is my cybersecurity staff going to be more capable than a cloud provider who is hiring the best people in the world and has a giant effort against this capability?” The complication with a people-powered security staff is ensuring their technical capability and managing them well.

Bioengineering our way to sustainability and better health

Roberta Fusaro: Now I want to shift over to a topic that’s also covered in the research on the future of bioengineering. I personally know less about this, so, Michael, I wonder if you could define what we’re talking about when we talk about bioengineering. What are the technologies that make up this category?

Michael Chui: When we talk about the future of bioengineering—we also had another report that predated it called The Bio Revolution—what we’re actually thinking about is the convergence of a set of innovations and inventions in biology. There are things like CRISPR, which is a way that you can start to edit genetic code or the ability to scan a human genome or any organism’s genome relatively quickly and relatively cheaply, with a set of innovations in information technology.

People have heard the word genomics, but there are a whole bunch of other molecules other than DNA and RNA. There are proteomics, looking at proteins, for instance, and lipidomics, looking at lipids. There are a bunch of things that you can do at the molecular level.

You can level that up to cells and organisms. Many of us are aware of the climate implications of, particularly, livestock because cows are flatulent, among other things. They eat a lot, and you need a lot of water.

Now people are starting to grow meat from cells in a lab. They call it cultured meat. That’s an entire set of alternative proteins, which has implications not only for health and taste but also for the climate. So those are some of technologies that we’re talking about when we talk about the future of bioengineering.

Roberta Fusaro: What’s the uptake been for bioengineering technologies?

Michael Chui: It varies greatly by the different industries or areas of application. We’ve seen things like mRNA vaccines on the cutting edge of healthcare. We’ve seen the vaccine for COVID-19, but now people are talking about how we could potentially vaccinate against other diseases that have been challenging for us before. RSV [respiratory syncytial virus], for instance, a lot of us who have kids are suffering through that.

You can imagine a vaccine for certain types of cancer caused by infectious disease. That is changing the game. So certainly in the life sciences/health field, it’s quite advanced.

We also found applications, for instance, in consumer products. Imagine putting on a skin cream so that you wouldn’t have to shave. We also look at materials of various types, whether it’s energy, or, for example, there was a big biofuels trend a few years ago.

Turns out it’s very difficult to compete with an industry that’s been around for one hundred years and is really optimized very well. Even here in the Bay Area, there are about three different start-ups creating artificial spider silk for apparel.

Imagine what’s possible there. We’re also seeing different types of biomaterials that are stronger, more sustainable. And we’re seeing more and more of that happen as time goes on.

Addressing the unease around bioengineering

Roberta Fusaro: What are some of the potential concerns and risks associated with the deployment of bioengineering technologies? What could prevent companies from moving forward with this group of technologies?

Michael Chui: When you’re talking about the building blocks of life, it is something that concerns people. So there are a lot of different things that companies and policy makers are thinking about. Take, for example, the old saying from Jurassic Park, “Life finds a way.”

This requires a deep understanding of what’s actually possible and not possible. For example, the mRNA vaccine will not change descendants’ DNA.

It doesn’t actually work that way. Even though it is genetic material, that’s not how the science works. So I think a deep understanding of how the science works and then couple that with the appropriate policy makes a lot of sense.

There are other questions as well. What’s patentable, for instance? There’s been disputes about that. How do you think about cross-border questions? There are genetically modified organisms and genetically engineered organisms. The second use genes, which already exist in nature. So there are a lot of subtleties here, particularly on the regulatory side.

Finding the right talent

Roberta Fusaro: Do we need new types of talent in any of these areas in order for there to be more adoption? Are we lacking in cloud/edge computing talent? Are we lacking in any of these areas?

Michael Chui: We’ve seen organizations report that, oftentimes, talent is a real limiter. That said, one of the potential ways to address that is to create tools that allow people to be more productive so that people aren’t spending as much time doing data plumbing because you can actually automate data pipelines. That’s part of a trend that we’ve described as industrializing ML.

But we do need more. All of these trends we’ve discussed today are related, so you will need data engineers to do the future of bioengineering.

Roger Roberts: Sometimes we think of talent as people who graduated with degrees in X. I think going forward, we are all going to have to be lifelong learners and explorers of new domains and bring these cross-connections of fields together even mid-career, even late in our careers, to learn about the possibilities that can be unlocked here.

Michael Chui: This idea of being able to combine different trends to create creative solutions is actually what’s most powerful. It’s very rare that you’ll be able to just use this one trend and solve the entire problem. Usually, you will have to bring a number of things together. That’s really fun to do.

Tech is forging the future

Roberta Fusaro: What excites you most about any of the three technology trends we’ve discussed or any of the others in the report, frankly? Where do you see the most possibility?

Roger Roberts: Well, Roberta, I am a techno-optimist by nature. The reality here is that tech is stepping to the forefront. It’s stepping to the forefront of how we think about strategies.

Fifteen, 20 years ago when I was earlier in my career at McKinsey, I might’ve thought we set our business strategy, we think about technology, and then we determine how we’re going to execute our strategy. But tech is a way to execute our strategy.

Now tech is what is bringing the new building blocks, the new possibilities. You have to look at tech as a first-order topic in strategy. And you have to think about it as a creative process. How do we mix and remix these technology trends? How do we bring their combinations and permutations to life in new forms that then create unique and differentiated strategies that serve customers and serve our societies better and better?

Roberta Fusaro: This has been a fantastic discussion. Roger, Michael, thanks so much for joining the podcast.

Michael Chui: Thanks, Roberta. It’s been great to be here.

Roger Roberts: It was a pleasure, Roberta.


Lucia Rahilly: As critical as data is to serving our societies, it might lack human touch. Authenticity and empathy also play essential roles in making the world a better place. And our next story is full of both qualities, as told by McKinsey partner Brian Rolfes as part of our My Rookie Moment series.

Brian Rolfes: The story I’d like to share comes out of my very first study at McKinsey. It goes all the way back to 1995. What you need to know is I didn’t join McKinsey “out.” I was in the closet when I joined the firm, and I did that because I mistakenly thought McKinsey was this conservative institution where I couldn’t be authentically myself.

The good news is that within a couple of months, I discovered that I could be who I wanted to be here. I could be out and I was, but for that first study, I wasn’t out. I was serving a big Canadian bank. My boyfriend, Brad, and I had recently moved to Toronto, and we found ourselves on a Saturday walking down the street.

And, of course, who is approaching us on the sidewalk but a very, very senior client, the senior executive on this bank study that I’m on. His name is Jim. Jim’s approaching me. So I throw my boyfriend’s hand aside, and Jim says, “You’re on that study, the McKinsey study, aren’t you?”

I said, “Yes, I’m Brian. I’m one of the brand-new associates. Nice to meet you.” Jim says, “Well, this is my boyfriend, Alberto.” And I said, “This is my boyfriend.” And that immediately created a bond and a relationship between Jim and me that then lasted five or six years.

As it turns out, at the time, Jim wasn’t out at the bank. And I, as I said, wasn’t out at McKinsey. I came out within a month or two at McKinsey and helped create the LGBTQ+ affinity group here at McKinsey, along with others at that time. A few years later, Jim did the same thing at his bank.

I came out within a month or two at McKinsey and helped create the LGBTQ+ affinity group here, along with others at that time.

Brian Rolfes

But it created this bond that has lasted through the years. Jim and I continue to be friends today. The learning from that first study was, surprisingly, even as a gay man, don’t assume others are necessarily straight, and you’d think that would be self-evident [to me], but it wasn’t. The second thing I learned was that there can be a bit of a superpower in terms of being gay and meeting other people who are LGBTQ+ identified. It creates that bond in that relationship, and that relationship with Jim has lasted to this very day.

Explore a career with us