The AI evolution of life sciences and health care

Read through the transcript

Heidi: Artificial intelligence is a key next step in the evolution of life sciences and healthcare companies towards the future of health. AI enabled applications, machine learning, and advanced analytics capabilities are already helping companies thrive and widening the gap between those who have embraced these technologies and those who have not. But as adoption of AI increases, so can the risks. AI’s potential risks make it even more important for healthcare organizations to establish appropriate governance and oversight of algorithms and data. 

Welcome to Tales of Transformation. 

Today we will explore the promise of AI, how leaders can establish the right enterprise-wide strategies to maximize their AI enabled technologies and capabilities across the value chain.

I’d like to welcome my guests today, Andrew Bolt, principal in Deloitte’s life sciences strategy practice., Nondas Sourlas, CVS Health, VP of Data, Strategy, Governance, and Products, and Ajit Menon, Janssen Pharmaceuticals, North America, VP of Customer Engagement and Digital Transformation. 

Welcome, gentlemen. We’ve been hearing real promise of AI for quite some time. Nondas, what have we done so far and what would you consider success to date? 

Nondas: Coming from a healthcare perspective, there are pockets of success across the board, probably up to three or four years ago. A lot of it’s coming in automation and operational processes. More recently, we’re seeing the intelligent component of AI that is more real time use of AI.

So, for example, how do you set up next best actions when you have patients or customers, depending on your setting? Or how do you use analytics in machine learning in an AI form to predict risk or assess risk in a near real time type of setting? I think that we have demonstrated success in a lot of those areas, and the next step in that transformation is scaling, which comes along with the right structure and organizational setup with roles and responsibilities, and the right platforms from a technology perspective and governance. So, I think those parameters are key for scaling the AI capability and making them more mainstream. 

Heidi: Setting up the infrastructure for AI to thrive is really critical. Ajit, let me start with you. What have you done so far, and how would you define success to date? 

Ajit: So, similar to Nondas, I think that the technology platforms and the analytics on top of the technology platforms are going to be ever evolving. And especially with regard to machine learning, we’re looking at how historical patterns of data can be mined so that we can predict what future behaviors might look like and what  future patterns might look like. As it pertains to the pharmaceutical and biotech industry, I would say there have been three main use cases for the application of artificial intelligence, or more specifically machine learning.

One is in the prediction of targets from a drug discovery standpoint that can be explored and that can be mined, so that we can develop drugs in a much faster timeframe than years past. The second is in terms of engagement with customers, especially on the healthcare professional side, make sure that we have the right next best action and a tactic to deploy with our customers.

And last but not the least, the use case that is prevalent right now in pharmaceuticals and biotech is patient engagement and interaction patient experience so that we can provide the right support, whether it be copay support, or whether it be other kinds of access support to make sure that patients not only start, but also stay on better.

And it’s not just limited to Janssen. I would say that it’s really prevalent across the industry. To Nondas’ point, we’ve made strides in pockets and I think the key question now is, how do you scale from the foundational successes that we’ve had and make sure that we continue to put the pedal to the metal and sustain the success that we’ve had so far, both from a technology standpoint, from a business case standpoint, as well as from a talent standpoint? 

Heidi: Doing things at scale, what were specific challenges for your respective organizations? 

Nondas: I think how do we structure ourselves inside the organization, the structure we have, how the teams work together, who does what, who lives on which activity was I think, quite important, and more often than not, we didn’t get it right the first time around. We still iterate between the different models and frameworks, and they all come with some caveat and will try to refine and get better for the next iteration. 

Ajit: From my perspective, there are three things that really contribute to success of scaling. One is user experience and wanting to make sure that we have optimized user experience, not just across the board, but also within individual teams, individual organizations, and wanting to make sure that we have adoption and optimized user experience, because once you have the user experience—and especially if you have advocates and evangelists—they can be your force multipliers to ensure that any effort that you put in place from an AI, machine learning, and data science standpoint will scale. The second essential ingredient is to be able to show quick wins and the return on investment from the initial pockets of success and be able to use that to drive continual and iterative investment.

And the third would be, using methodologies like Agile to ensure that you have a continuous test and learn environment, both from an internal technology standpoint, but also with willing participants out in the market to ensure you have a minimal viable profile. You build on it, you show the success, and you continually test and learn, and do continuous AB testing as well to make sure that you have that evidence and the data to prove the hypothesis and take those insights into action. That’s how you’re going to be able to scale. 

Nondas: I would fully agree, and I would probably add the talent and technology, thinking of the resources that go into this list. Thinking back five years plus, most of the talent was a bit niche, difficult to find; technologies were still not fully scalable, or the sheer volume of data, or the lack of data might have created obstacles. And I think in the last few years, we’re overcoming those blockers and it gives us the ability to scale faster. We are, within CVS Health, in the middle of a journey, moving from a lot of legacy on-prem or fractured platforms to cloud enabled environments.

We have actually seen that through some of the use cases that we had and have demonstrated the power. And through resources on demand that that cloud gives us, especially with processing the huge amounts of information—process it fast—being able to scale up and down on demand without long pre-planning processes.

So, I think that is one element, getting the technology, and the longer the technology comes, tools that you use, whether it is cloud native or in various tools that work on one platform, not in the other. So, I think that whole ecosystem of platforms, tooling, has been an area of focus. And the comment around talent goes hand in hand. If talent comes ahead of the technology, you might frustrate people as you hire them, and you don’t have the ability to give them the tools that that they know how to use. And then if the opposite—you move with the technology too fast and you don’t have the people to actually utilize it—you’re again, not optimizing. So, I think it’s two things that go hand in hand. 

Heidi: As a follow up, Nondas what was one of the use cases that helped accelerate your transition to the cloud? 

Nondas: So, an example on the technology as we have the data and the computing power, you start going a lot deeper, and rather than working on averages, you start working on actual observations, or I’d say the price of a particular drug in a particular pharmacy based on buying contracts. Pretty much the data grows exponentially, and the ability to actually operate at that level, the right models that you need to optimize prices with, as well as the technology to actually make this work. The amount of data was one of both the key use cases a couple of years ago that pretty much helped us accelerate that transition to the cloud. There are a lot of other examples. I think next best action is still something that we’re generating quite a lot of value at the moment with the prospects of getting a lot more out of that, both from a financial perspective, but also from a patient experience, and supporting the members with their health at the right time, with the right decisions.

Heidi: For listeners who are just starting out on their journey to build their AI organizations, what’s the right approach in terms of what to build internally? And the right way to think about using partners to help you achieve your AI goals? 

Ajit: So, I would say that there’s three ways in which we’re able to really go along with taking AI and machine learning to the next level, building on initial quick wins. The first strategy that we’ve used is to ensure that we have an up-skilling of our internal count. We have a big philosophy for establishing business data scientists, and that means that we continually upscale our internal talent on an ongoing basis. The second thing is to ensure that you have an ecosystem of partnerships. So you need to ensure that you’re developing partnerships with the right vendors and the right agencies, and you continue to build those relationships and scout for new partnerships over a period of time.

And I would say that the third thing is, you need to have competitive advantage. Whereas you can probably outsource to external companies, the technology— some other places where you need automation, things of that nature—but certainly algorithms, data sciences, making sure that you have intellectual property, and proprietary algorithms that are built and custom made for your specific use case.

And the last thing is you have to have a continual process where you’re recruiting from a whole host of different sources, whether it be inside the industry, outside of the industry—academia as well—and setting up those partnerships with academia so that you have early and career talent that you’re bringing in through internships and co-ops, making sure that they’re learning how to do things the way that you need them to do things and learning along the journey, so you can get most out of your ecosystem of partnerships across technology, data, analytics, and business to ensure that you have this entire ecosystem at your disposal.

Heidi: Andrew, in your work with clients, what are some of the limiting factors in the broader AI ecosystem? 

Andrew: Nondas mentioned before that data was a rate-limiting factor a few years ago in terms of AI, but now what we’re really seeing with clients is data science talent is really that rate-limiting factor. And it makes it challenging to scale and broaden ambitions around AI, not only partnering with the broader AI ecosystem, but influencing the ecosystem through partnerships with building talent to create a larger talent supply of those next-generation data tests is critically important. And then we’re also seeing some companies, as Ajit mentioned, upscale or retrain talent internally in order to get access to the folks that they need to actually execute on their AI strategies. 

Heidi: To Andrew’s point, companies that are just dipping their toes into AI, knowing that it requires senior leadership to understand what’s required, what advice, or maybe examples could you share? 

Ajit: I would definitely demystify AI and machine learning for senior leaders. One of the ways that we’ve done that here is to ensure that we actually take senior leaders, not only through coursework and curriculum that gives them the theory behind AI and machine learning, but also provide practical applications of AI and machine learning to the use cases in their particular area of the business, and have them go through a simulation and a role play where they actually have their hands on a keyboard to ensure that they can work on the use cases that are applicable, that are relevant to their business, and they see what the outputs look like. So, it’s an immersive, intensive experience. Sometimes it takes several hours during the course of a day, or we split it up over a period of days to ensure that they can actually design something; they can role play how it’s used by the end users, look at the feedback that’s coming through the closed loop CRM system, and then act on it. They see the full 360 view of how something goes from development into testing, into production, and back as an enhancement into the algorithms and into the inner-working, so to speak, of the technology. Once they have that immersion through the system, through the technology, and through the application of the systems and technology into actual, real life—practical applications and use cases—then they realize what it means to go through this and how much time and effort it takes, and that makes them bigger believers and certainly leads to much more credibility that then leads to potentially increased investment as you move forward. 

Nondas: The most foundational piece to tackle as you embark in that ML, AI journey, is getting understanding and sponsorship from your leaders, and then, obviously, resources to actually help execute. I think on a practical basis, getting the language so everybody understands the same things when we talk about AI; I think also frequent communication with the sponsors of how things are progressing, what we’re working on, and I think also a bit of a tactical approach on the journey. You probably start with use cases that can demonstrate success, sooner with big impact, probably less resources, if possible. So, something that’s not as complex at the beginning can demonstrate some impacts, so you actually demonstrate the value of AI. 

Heidi: Andrew, as a principal of Deloitte, you’re consistently advising life science and healthcare executives on these strategies. In terms of gaining senior leadership support, what are some of the considerations?

Andrew: I think number one, to Ajit’s point, it’s really important to educate the business on the art of the possible to AI, and demystify what it is and what it isn’t. And so, to the extent you can do that through training or through experiential work is fantastic. I think once you help them wrap their mind around the art of the possible, what is critically important is to make sure that the projects that you’re going to undertake and the use cases that you’re going to tackle are important to the business—that are tied to actual business outcomes—that you’ll be able to measure impact later on. I think one thing that some of our clients are struggling with right now is they’ve done great work, but they’re unclear about to how to measure the value on all the investment that’s been made into the capability, and so being able to ground it in business outcomes is absolutely critically important. Once you do that, then I’ve seen many organizations continue to scale and grow after being able to show that impact. 

Heidi: Nondas, I’ll start with you. With artificial intelligence comes tremendous amount of responsibility. What can companies do and how do they establish the appropriate governance and oversight of algorithms and data?

Nondas: Quite a hot topic at the moment, at least in our business. We have set up our data governance team; there is also a separate team for ethical AI. The initial steps were inventory of models, or use of AI, or algorithms across the business, which you never know whether you got them all. You don’t know what you don’t know, but especially in a business as diverse and as big as CVS Health.

I think getting focused on this by having a dedicated team. Look at an inventory of what kinds of models you have, and which parts of the business are deploying those— and I should mention there are models that we deploy ourselves, and there’s models that come through vendors or externally, and you want to have your oversight on those as well. And method: standardize the methodology to assess whether models have any bias in there. And how do we do that at CVS Health? Rather than building, let’s say, a huge team centrally that tests all the models—especially in areas that we might know a little about of the business—we have tried to educate different parts of the business on how to test what ethical AI looks like, how to test for some of those things in rely on the teams themselves that actually do some of those checks and controls themselves, and we are there to provide support, education, and how to troubleshoot  when some virus is suspected or detected. There’s quite a lot of attention and requests we get from the regulators, states, clients, and employers on the controls we have around are one of those processes need to be auditable, documented. We also look externally for support and work. Besides the knowledge of understanding what others are doing, and what best in class looks like it, it also brings some legitimacy that we’re actually following some industry-wide guidance and protocols. 

Heidi: What’s Janssen’s approach to governance and oversight?

Ajit: I think the first and foremost piece of governance has to be at the data level, because the data in and of itself could have some biases, so ensuring that we have the appropriate data—both internal as well as external—and making sure that there are governance bodies in place to inspect what you expect, making sure that the use cases that are coming forward are tapping into the data with the right intent, making sure that you have a view on privacy, and potential re-identification issues are going to be really important to ensure that the data themselves are as free of bias as possible. Then you have on top of that, analytic governance to ensure that we have the right kind of algorithms, and the algorithms in and of themselves are not biased in any way.

So, making sure that there’s checks and balances in place, test and controls in place, in fact, features that we build into our AI and ML models to ensure that we’re checking on an ongoing basis for any kind of bias. And I think the third point is ensuring that as we go along this journey, that we’re continually checking to not only at this point, but around external authorities as well.

This is not a one and done thing. This is going to be a perpetual journey where new rules, regulations, and governing guidelines are evolving, not only in GDPR, but California privacy laws, other laws in different states, and ensuring that you have a continual up-skilling of your team to ensure that they’re understanding what exact laws those are, what that means in terms of requirements from both our data as well as analytics is critical. And you’re testing on an ongoing basis to ensure that—to the extent possible—you’ve taken the bias out of both the underlying data, as well as the analytical algorithms that you’re applying on top of the data, as well as the use cases in and of themselves; because even if you have taken the bias out of the data as much as possible, ensure that the right analytics are in place. The use cases themselves that are coming in front of the governance team need to be blessed, need to be scrutinized, and we need to ensure that we’re not using the analytics and the data in a way that plots anything.

So, ensuring that those use cases are identified and governed is also critical. 

Heidi: Ajit, how do you do this while not slowing innovation? 

Ajit: Compliance is foundational to all the things that we do. I don’t actually see this as a dichotomy. There’s no trade-off associated with doing things in a compliant fashion to ensure that we predict the privacy of our customers and our patients while at the same time trying to innovate.

In fact, it also means that you’re continually scouting for the right technologies, the right data sources, the right analytics talent that’s out there, bringing in the external perspective as Nondas pointed out, working with external partners—academia to validate some of those use cases, to validate some of the algorithms to validate the data from a privacy standpoint.

So, if you foundationally make the assumption that everything that you do needs to be compliant, needs to ensure protection of privacy, and protect from re-identification of your end customers, it’s actually a good thing that we do and I’m hopeful that we can continue to build on that as we move forward.
Nondas: From my experience, there’s good buy-in in the business and people understand and see the value of compliance; and I don’t see it somehow competing with speed or priorities. So, finding exactly how you set us close to the compliance law and limits without bridging them is a bit of an art, and where you can actually gain close to the competition. 

Andrew: These are all great points. One thing I want to underscore is, it’s an incredibly dynamic space, right? And speakers mentioned that the regulations are evolving, and different from country to country—even state to state—and so, staying on top of that is critically important. But then the AI just in of itself is dynamic, right?

The data changes, the models are developed at a point in time, and the models change. Sometimes they can drift, and so I think it’s critically important that someone have the responsibility—or some organization have the responsibility—for keeping tabs on that, and making sure that as the environment involves, the ecosystem evolves; as the data and the models evolve, that someone’s paying attention in order to do responsible AI.

Heidi: As we come to the close of our show today, Nondas, let’s start with you final words of wisdom. 

Nondas: We have been hearing about AI and machine learning now for about a decade. What’s the next big thing that’s going to come? And then reflecting that—what we have achieved over those 10 years—I’m also thinking we’re probably still scratching the surface of the capabilities and the opportunities with AI. Staying the course, you do need the talent, you need the governance, you need the sponsorship and support, and you need to apply AI—correctly, safely—where it makes sense. Keeping it clear as you go through that and looking at the choices, evaluate them, Doing the due diligence I think pays off. I’ve been with CVS Health for a couple of years, and I think there’s quite a lot of successes and good value generated, but at the leadership level, we all agree there’s quite a lot more to go. The environment is still changing, regulation is changing, technology is changing—so it’s quite fluid—and I think keeping greater open work with other people and companies and vendors outside our own company is important.

That brings in some of the newest thinking that helps you understand where the industry—or all of AI—is heading, or where your competition is heading. It’s important to give an open and broad horizon as we move forward. 

Ajit: Just to build on that, I completely agree with all that Nondas said. I would say it’s a marathon, not a sprint, and that has to be aligned all the way up to senior leadership.

We need to have the right milestones along the way to show successes, but it’s a continued evolution. It’s not an evolution towards 2023, or 24, 25 it’s a continual evolution, and you need to show successes along the way—but recognize that everything is transforming around us. And so, as we transform AI and machine learning, we need to ensure that we’re taking into account that the environment around us is changing as well, and we need to continue to in an agile fashion, adapt and evolve.

So that’s number one. Number two, Nondas brought up the point around talent, and it’s extremely important that it’s a combination of both up-skilling talent as well as ensuring that you have the right partnerships across the ecosystem, between vendors, academia, and industry. So, talent is going to be extremely important and the skills that are needed, both from an analytic standpoint, a data standpoint, as well as a business translation standpoint, are going to be critical. And I would say the third thing is that you need to ensure that you have the right governance, not only in terms of the compliance—the privacy—ensuring that there’s the right ethical AI that’s delivered, but also because you want to ensure that you’re investing in the right use cases, and prioritizing those. So that needs to be governed as well. 

Heidi: There’s no way I could have said that better. Artificial intelligence enabled solutions can provide many benefits for organizations, as we’ve heard today, such as immediate returns through cost reduction and better consumer engagement. But there’s still a lot of work to be done. That means putting strategies into action on a functional level by communicating a clear AI vision, helping that workforce operationalize AI, and finding the right ecosystem partners to supplement technical needs. I’d like to thank Andrew Bolt, Nondas Sourlas, and Ajit Menon for joining me today on Tales of Transformation.

This podcast contains general information only and Deloitte is not, by means of this podcast, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This podcast is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional adviser. Deloitte shall not be responsible for any loss sustained by any person who relies on this podcast.

As used in this podcast “Deloitte” means Deloitte & Touche LLP and Deloitte Consulting LLP, which are separate subsidiaries of Deloitte LLP. Please see www.deloitte.com/us/about for a detailed description of our legal structure. Certain services may not be available to attest clients under the rules and regulations of public accounting.

Get in touch

Andrew Bolt Principal | Deloitte Consulting LLP anbolt@deloitte.com 7032511288

Aditya Kudumala Principal | Deloitte Consulting LLP akudumala@deloitte.com +1 212 313 1562

Kumar Suman Senior Manager | Deloitte Consulting LLP sumankumar8@deloitte.com 3176562497