MOLLY WOOD: Today I’m talking to Delphine Zurkiya, senior partner at the strategy and management consulting firm McKinsey and Company, maybe you’ve heard of it. Zurkiya works at the intersection of technology and healthcare, helping medical and life sciences companies accelerate their digital and AI transformations and their growth. She’s also a member of the McKinsey Technology Council and currently leads generative AI initiatives for her division of the company. In this episode we discuss AI adoption journeys, reskilling teams for the new era of work, what AI means for consulting, and the value of being completely transparent about when and how you utilize AI.
MOLLY WOOD: Delphine, thanks so much for joining me.
DELPHINE ZURKIYA: Thank you. Thanks for having me.
MOLLY WOOD: I want to ask about your kind of long history with AI. You have an engineering degree from MIT and went on to work with AI but have said that the state of AI 15 or so years ago was part of why you pivoted to consulting. What can you tell us about where it’s been compared to where it is now and how you’ve seen that evolve?
DELPHINE ZURKIYA: It wasn’t really the technology itself that excited me, it’s what you could do with it and how you could support, in my case, physicians in better diagnosing cancers. The reason I was frustrated is not about the technology, it’s about all the change management that is needed for AI to be adopted. And I felt at the time we were too focused on the technology and not focused enough on the people and the process aspect around the technology. And what does it mean for a physician to adopt AI and change their workflow? Do they have time? Who trains them? How do you measure?
MOLLY WOOD: Do you find that the challenges of change management are particularly hard in these industries, in healthcare and life sciences?
DELPHINE ZURKIYA: I think they’re hard everywhere. These are industries that are regulated, and so we have to be very responsible in how we deploy AI, but they’re also industries that are very hungry for assistance. In healthcare there’s a lot of burnout, and also we want to innovate much faster, and we know AI can assist. But change is hard, and I use AI in my own workflow, and it’s taken me a while to say, okay, I have to stop and reskill myself. We’re running so fast, it’s just sometimes hard to just take even 10 minutes of our time and experiment. So I think change is hard everywhere, and there are tools, there are a lot of playbooks for how to do that, and that’s the dialogue we should be having in the enterprise.
MOLLY WOOD: I wonder within your sort of specific area, what are you seeing in terms of improvements, operational efficiency, or automated tasks, or even, hopefully, some life-saving technologies?
DELPHINE ZURKIYA: In general, I think healthcare is middle of the pack when you look at the adoption. It’s been accelerated lately because of the burnout I mentioned, especially on the provider side, and because of the availability now more of the shelf tools. It is still starting for the most part with the operational workflows. So, how do we get the back office to be more effective or more efficient. It really runs the gamut of whenever you have a human being that has to go fetch information, that has to read through and synthesize, that has to reconcile, this is where we’re seeing a lot of creative use cases. The area where I’m particularly excited, and this is now, I think we’re seeing that more at the forefront is how do you really assist the middle office and the front office? You know, these are clinicians that, of course, they do their job and they’re the ones who are going to give the recommendation, but why not have an assistant that, you know, like you would have a fellow or an intern? Why not have an AI that assists you in the number of simulations that you could be doing to look at whether or not your device is working in the human body? So more and more now we’re seeing AI starting to make its way into products.
MOLLY WOOD: That’s fascinating about the product development and testing aspect. Can you say more about that?
DELPHINE ZURKIYA: Product development is obviously very different in different parts of healthcare. So if I maybe start in pharma, which is where there’s been a little bit more in the news about AI helping with drug discovery, I think that is certainly an area that’s really ripe for AI, and especially these large language models that can really model ultimately anything that’s a series of tokens. So if you think about chemistry, about biology, these can be modeled using these techniques. It’s also being used quite a bit in the clinical trial space, because this is where we still spend most of our time, making sure that drugs are effective and safe. And that’s where there’s just a lot of back and forth—all the various trials and you need to see how the patients react. All of that, AI is really assisting. If you move now to MedTech, which is, you know, more about developing devices, and a lot of them are software-enabled, you’re going to start having techniques that help software engineers write code much faster. And then on the payer and provider side, the innovation is really in the care. We are starting to see AI be there to connect the dots in a way that a human brain cannot because you haven’t seen every potential case, you haven’t read every paper. So that’s where the assistance is coming.
MOLLY WOOD: But also, on top of all of this, you are leading McKinsey’s own AI adoption journey for your division of the company. How has that unfolded?
DELPHINE ZURKIYA: It’s been fascinating to take our own medicine. [Laughs] I’ll start with why we did it. When I think the world was waking up to ChatGPT, we realized it is really important to understand how to deploy at scale. We have more than 35,000 consultants, and that was the goal always from the start, to make sure this is available to everyone. And it is available to everyone on the use case for us that is the most important, which is all the knowledge that we have that is within the walls of McKinsey around every time we help one of our enterprises—what is it we learned and how can we make sure everybody at McKinsey can benefit from it? You know, we went from, I’d say, a pretty traditional search bar to be able to look for information to now an interface that not only helps us find information, but we’ve been surprised at the creativity of our consultants and how they’ve actually come up with use cases we never thought of. So we’re now at 40 different ways that our consultants use AI. And some of it is still in the bar where you can type, and some of it is now starting to look much more like its own application and pretty specific to a workflow for a very specific group of consultants. So that’s sort of what we’ve had to learn, because, initially, we thought, there’s going to be one product, it’s going to look like this, and everybody’s going to use it the same way. And then we realized, no, actually it’s a platform. You know, I’ll say we take our own medicine because I think we were probably initially also thinking, you know, we’ll build a technology and they come. And then we very much realized, no, it is important to really reskill, understand how people use it, iterate the product as quickly as possible. We now have a full infrastructure that we, you know, using a lot of folks that are already at McKinsey, but, you know, have sort of evolved their job description to be able to support now this assistance.
MOLLY WOOD: Were you surprised at how much change management you needed internally, even though you know this is the hill to climb?
DELPHINE ZURKIYA: I wouldn’t say we were surprised, we were expecting it. But we were surprised at where change happens and where it doesn’t. So, for example, I think we had assumed our business analysts would embrace the technology very fast and senior partners wouldn’t. And in reality, if you look at the data, there are super users at every level of the enterprise. There are folks that have become really creative and actually had spent the time to learn how to change their workflow so that they can become more effective. The other surprise is how people have been using the tool and how much people are asking for these various use cases. We initially thought we had access to all the data. It turns out people have plenty of data on various repositories, SharePoints, et cetera that we didn’t know existed. They didn’t really necessarily want to give it to us at the beginning. And then once everybody realized, Oh, that’s how now information is being consumed, I think now there’s a lot more of that data flowing into the system. And then finally, I think we’ve realized that training can only do so much. We’ve gotten really creative at ensuring, for example, in every office, we have someone in IT that can be there and support you. And now those folks actually train people how to use Lilli. And that’s been interesting because that’s really changing the job description as we speak. But a lot of that came from them. They said, Hey, can we start, because that’s where the action is. That’s where people are excited.
MOLLY WOOD: And Lilli, just to clarify, is your internal platform.
DELPHINE ZURKIYA: Exactly. That’s the name of our internal platform. Yes.
MOLLY WOOD: I feel like, more than someone in a different type of company, your adoption journey feels like it also makes you better at your job. Like, it’s sort of extra accretive. Like, not only do you become more efficient but you take those learnings and pass them on to other companies going through this journey.
DELPHINE ZURKIYA: At multiple levels I think how we build the technology is a lot of lessons learned, and we’re working very closely with all the LLM providers, all the hyperscalers constantly to exchange ideas and push our collective thinking. So that’s also been very helpful and something our clients can benefit from. The other part is in our client service itself. I do believe we have access to information faster. I was meeting with a head of HR for a company, and she was asking me, How do you set up a group such that we can do the product platform? And I knew there was a slide somewhere. And so I just typed in Lilli, you know, product platform, blah, blah, blah. And it pulled up a page from our rewired book, which is where we explain how this works. And I just switched my screen and I said, let me just show you the diagram because that’s the best way to think about it. And I said, Oh, and by the way, I just pulled that up using Lilli, and then she asked me a ton of questions as to how that works.
MOLLY WOOD: She was like, You are a sorceress. Tell me everything. [Laughter]
DELPHINE ZURKIYA: But it’s showing, yes, as a head of HR, you’re going to set up a group to do product platform, but by the way, that group’s going to enable your company to do something like what I just did, and it’s important not just to role model but to explain to people, you know, this is really the way you change your workflow. In the end, that’s really what we all need to be doing. You know, my workflow would have been, I would have texted an associate, they would have looked for something, and at the end of the meeting, if I was lucky, I would have pulled up the slide.
MOLLY WOOD: Right. Absolutely. How do you think about how the existence of these tools, the continued development of these tools, the ability to do things like that so quickly, analysis, right? Projection. How does that change the consulting industry writ large?
DELPHINE ZURKIYA: One question we get asked a lot is, you know, does that mean you’re going to get rid of your entry-level folks? And I do think the answer is no. What’s going to happen, though, is our entry-level folks are going to now do tasks that are very different. So if you think about the work of a business analyst, and even it’s evolved in the last 10 years, a lot of the calculations were initially on paper and then they moved to Excel and then they moved to Tableau, and now they write Python code. But they’re still at the end doing what’s essential, which is, we have a very important question from our client. We have a lot of data and we need to get to that answer as fast as possible, ideally with more and more data. The way we work evolves, but ultimately what we have to do as consultants, which is really answer a business question, that’s going to be the same. I want to be eyes wide open. Of course, there will be some shift of tasks from people to the machine, but then I think we’re also going to realize there’s going to be a need for people that check the quality, and we’re going to realize there’s more tasks that we hadn’t envisioned. And I think it will be an equilibrium, but at the end, I do hope we get to answers faster.
MOLLY WOOD: You co-authored a piece early last year, which is roughly one hundred years it feels like in AI time, about what every CEO should know about generative AI. Even then you were saying, you know, it’s more than just a chatbot. I wonder, with a year now under your belt, what updated advice might you have for CEOs?
DELPHINE ZURKIYA: So I would say I’ve noticed, you know, the question from a lot of CEOs has shifted from, should I care about this technology to, I now know I should care, but what do I do with it? The biggest advice I have for CEOs is to say, you know, what is it you’re trying to solve? What’s your strategy? What’s your business problem? And then we can think about how AI can support. You don’t want to lose sight of that. But certainly there is now, I think, a need to be able to articulate to a board, to employees, to investors, why is it you might have not thought about how AI can support? So you do need to be ready and have an answer to that question. You also need to be ready to explain how you’re going to scale AI. I think we have seen—unfortunately, gen AI is no different than AI, meaning there’s a lot of pilots that just never lead to anything changing in the organization. Roughly 90 percent of pilots really don’t scale, because in the enterprise it is about changing people and processes. And if that’s not put in place, again, the technology won’t work. So that’s the other area that CEOs really have to think about is—if we take Lilli, our own example, if you look at the people we’ve deployed, it’s a one to one. So one is a technical person and one is a change management person. It might seem controversial, but when folks want to deploy AI, my first question to them is, do you really have the budget to do the change management in addition to the technology? And if the answer is no, then I would say just focus on one area first, get it working as opposed to spreading yourself too thin. And so we’re big fans of picking a domain where there’s an executive sponsor. There’s enough, ultimately, resource reallocation to make sure you don’t just fund the technology, but you also fund the change management. And just that alone is really what a lot of CEOs are learning needs to be in place.
MOLLY WOOD: That is really powerful. I kind of want to reiterate that. I mean, it has been a theme of this season for people to say the companies who focus on outcomes first will have the best success. If you just try to implement a technology solution and you don’t manage the people in the process, you will fail. You will have a pilot that goes nowhere.
DELPHINE ZURKIYA: That’s right. And will frustrate a lot of people along the way. So it is important to think about that.
MOLLY WOOD: Right. Another thing you do is work with what we’re calling AI-native companies, companies that are being born in this era and using AI in ways we might not have thought of. But you have this kind of people-first, it sounds like, approach to that. What can we learn from AI-native companies who are doing this well?
DELPHINE ZURKIYA: So what I really enjoy in AI-native companies is everybody’s a technologist. It doesn’t really matter who you are, what your title is. Everyone really understands what technology can do and what it cannot do, and really gets impatient when the answer is, we’re just going to throw a process at it without having thought about, well, can technology also help? The legal teams in digital-native companies understand how LLMs work, right? They will ask questions around, what is the filtering technique for that particular large language model? The L&D team, right, learning and development team, really understands that it’s hard. You need to explain and have a curriculum such that you stay on top of how the technology evolves. So that’s what’s been fun. To some extent, you don’t need to teach that. Now, of course, you also need to stay grounded in the business problem you need to solve. And I think great companies strike that right balance of not over-investing in what technology can do for you.
MOLLY WOOD: I feel a little targeted. I definitely struggle with that impulse to just throw a process at everything, but I know plenty of leaders do too, so can you talk about how to avoid that mindset?
DELPHINE ZURKIYA: Yeah, I mean, in the end, it’s very much design thinking. And I think all of us have been trained to think, okay, let’s figure out everything that went wrong. And let’s try to fix it with, again, the way people work and processes work. The comment would be to say, do we even need the process? Does it make sense? You know, I’ll use one example, which is more on the commercial side, which is often you have pricing councils that approve if you’re going to get more discount for a particular deal. And that’s a very heavy machinery that is usually set in place. And there’s a lot of different stage gates and conversations. You could certainly get technology to help you and support you in going faster through that process, or you could just ask, can there be more self-serve, more assistance directly, such that you don’t need a lot of people advising, you know, sales reps, for example. And if you could do that, then I think you make everyone happy because often the part where that technology can help, the automation part, is not the most fun part of your job. It is very specific to each area. We often get asked, Hey, can AI do this? And we say, yes, it can. But if it’s going to save an hour of somebody’s time, is it really worth it? One of my colleagues, Rodney Zemmel, calls it the dog paradigm. You know, dogs will be really happy, they get walked an extra hour, but, you know, it’s not going to help the enterprise. So there’s a little bit of a threshold you need to say, is this meaningful enough that you actually get rid of a piece of the process, such that we’re talking much more than one hour a week.
MOLLY WOOD: And so if you’re an AI-native company and you’re tackling, let’s say, an existing industry, you have an opportunity and you will have a workforce that, from the get go, may say, everybody else has a process for this, we’re not doing that. Or if you’re attempting to set up a process, you could in fact say, actually, let’s think about this in a completely different way and apply technology instead.
DELPHINE ZURKIYA: That’s right. Yeah, you really can and should challenge the way everything’s done such that in the end you get to the customer as fast as possible in a, you know, the less friction way as possible. And there are many examples. I mean, we read them in the news, right, of companies that are starting to rethink, you know, do we need these very large systems and all these approvals, and do we just get rid of that and go much faster? And that’s what digital-native companies can push us on.
MOLLY WOOD: Well, and that’s what I wonder, how to turn that then to larger, more entrenched companies who are maybe addicted to their systems. They maybe have a little of, you know, they’re pot committed to processes. How do they handle that rethinking and maybe try to adopt some of that AI-native thinking and behaving?
DELPHINE ZURKIYA: I certainly think there’s some industries where there’s going to be a lot more pressure. We see it in the education space, for example. It may be the less regulated areas where, ultimately, it is easier to disrupt the way ultimately how you create products and way to go to customers. But even in the regulated areas, the way I see companies do it successfully is what we’ve discussed, which is they pick one area, which is as much about the business value. The business value needs to be there, as there is a set of sponsors that are willing to fail fast and commit that this is not about, does the technology work or not, because we know it typically works. It’s about how do we make sure it gets adopted and you break through every possible silo that exists in a company and you truly are cross-functioning, you truly are one team. And there’s no blaming different parts. That’s agile. I think we’ve been talking about agile a lot, but this is very different.
MOLLY WOOD: Right. I mean, I do feel like I want to ask you, have you ever actually seen that happen? Does that happen?
DELPHINE ZURKIYA: It does, but it happens when people trust each other. It doesn’t happen because somebody says it has to, or there’s meetings being scheduled. It happens when there’s usually a moment where everybody realizes, I got to trust you. And that’s really when I think the magic happens. You have to trust your technology company. If they tell you that it’s important to go through this responsibly or, no, you can’t just patch together this, and a data platform’s really important. Well, you have to trust them, but also the technology company has to trust the business when they say, You know what, I’m going to be okay if the AI is 80 percent correct because I can do something with that and they don’t feel like they’re going to get thrown under the bus six months later because the AI is 80 percent correct and the business loses patience. So all of that has to be discussed and there has to be this constant iteration that, you know, we are one team and we’re going to figure it out together, just like you would in a startup. I mean, startups rarely make money with the first idea they had. Same thing with deployment of AI.
MOLLY WOOD: It feels like to the third leg of that stool is that good leaders have to make sure that their people trust that they will be reskilled, that they will be brought along on that journey.
DELPHINE ZURKIYA: Yes. And there is again, the budget to do that. It’s not somebody doing that on the weekend or as an additional part of their job. It’s really rethinking the way, again, your learning and development team works and the peer-to-peer teaching that can happen.
MOLLY WOOD: So you mentioned, actually, with your clients that you try to be really up front when you are engaging with AI, let’s say in a meeting, but talk a little more about the importance of, you know, I think people tend to sort of use AI in the background, and maybe it’s the dirty little secret of how you produced a report or got some information. Can you just talk a little more about the importance of being really open and transparent about how you’re using these tools?
DELPHINE ZURKIYA: Yeah, absolutely. I think it’s important to be transparent around any data that you’re using when you are making decisions, whether or not you vetted the data. So often people are afraid if you get access to data through AI and there hasn’t been a human in the loop. And so I think it’s 1) important to explain that AI was used and 2) to also say that there was a human in the loop, and that’s very important for responsible AI. I think the more we talk about how AI is used and not used and exchange information, exchange ideas, the more we are going to get creative as business folks to be able to say, okay, I see you’re using it, it seems to be working, you’re getting some business outcome out of it. Maybe then let me try on my side. And that’s happening within enterprises. Within enterprises you have one business unit leader that is really creative, and then other business unit leaders say, Oh, well, it’s working for them, so let’s try it on our end. So talking about it is important. It’s important to talk about when it doesn’t work because you learn a lot and, you know, you don’t want to be the one that’s restarting a pilot that didn’t work. And then it’s very important to talk about how people feel about AI. You do have to listen and you do have to use that as a way to say, okay, well, can we put some processes in place or specific policies in place to be able to address some of these fears. Otherwise, it’s anyways there, so might as well put it in the open.
MOLLY WOOD: Okay, I want to ask you a quick lightning round, if you don’t mind before I let you go. So, you talked a little bit about AI in your workflow. I’m hoping you can tell us how you use AI in your work and in your personal life.
DELPHINE ZURKIYA: Yeah, I mean, so one example is my, you know, the gym I go to, there’s a lot of ability to collect data around your workout or even, you know, yourself. And so I had taken a picture of that and I was exhausted and I didn’t want to put it in an Excel spreadsheet, so I loaded the image into the large language model and I said, you know, please analyze this and give me recommendations on what the workouts should be to be able to, in my case, increase my muscle mass. And I gave a little bit of information about who I was, and lo and behold, they actually gave me the recommendation. I was like, okay, well that’s pretty game changing.
MOLLY WOOD: Okay, so last question. Fast-forward, if you wouldn’t mind, three to five years. What do you think will be the most profound change in the way we work?
DELPHINE ZURKIYA: Five years, probably 10 years for sure, we will step into a company and it will look incredibly different in terms of everyone will be a technologist at this point, but it will be really knowing how to use AI to their advantage. I really do think it’s going to become, everybody will have their assistant, their intern, their expert at their disposal. So it’s going to feel very different. I can’t tell you if it’s going to be better or worse, but it’s going to feel very different. And I think they will have a lot less patience with the way we work today, which often is very inefficient.
MOLLY WOOD: Delphine, thank you so much for the time. I cannot wait for that future, and I have enjoyed talking to you.
DELPHINE ZURKIYA: Thank you. It’s been a pleasure.
MOLLY WOOD: Please subscribe if you have not already, and check back for the rest season 7, where we will continue to explore how AI is transforming every aspect of how we work. If you’ve got a question or a comment, please drop us an email at worklab@microsoft.com, and check out Microsoft’s Work Trend Indexes and the WorkLab digital publication, where you’ll find all our episodes along with thoughtful stories that explore how business leaders are thriving in today’s new world of work. You can find all of it at microsoft.com/worklab. As for this podcast, please, if you don’t mind, rate us, review us, and follow us wherever you listen. It helps us out a ton. The WorkLab podcast is a place for experts to share their insights and opinions. As students of the future of work, Microsoft values inputs from a diverse set of voices. That said, the opinions and findings of our guests are their own, and they may not necessarily reflect Microsoft’s own research or positions. WorkLab is produced by Microsoft with Godfrey Dadich Partners and Reasonable Volume. I’m your host, Molly Wood. Sharon Kallander and Matthew Duncan produced this podcast. Jessica Voelker is the WorkLab editor.
Microsoft is a technology company, a small local company, with few employees, no offices, and almost making no profit… >>
Please visit the firm link to site
