2022 MIT CDOIQ Symposium Session 10-5 - Francesco Tinto and Sanjeev Vohra

Okay, so first of all, good morning everybody and thanks for attending. And thanks, Sanjay, for having me, to exchange a little bit the experience and the commentary. I need to tell you the good but also the bad, and all the scars that we are in this journey. So my career has been always in CPG, as you said. I was in Procter & Gamble, then in Kraft Foods, then Kraft Heinz. So I think that I've seen a lot about CPG, and that was an interest also in terms of saying, in CPG you have a connection with what we call the consumer, but the reality is that you always work through the customer, the retailer, wholesaler, and so on. And there is always this connection that never goes all in. I was very excited when I got the opportunity for WBA for a couple of reasons. The first one was definitely WBA was at the beginning of a massive transformation where we were moving from just a traditional brick and mortar into becoming truly omnichannel. And so this vision of having the customer in the middle and making sure that we have a seamless experience across all the different channels, with the 13,000 locations across the world, you can imagine how we make sure that we have a strong physical presence but also a very, very strong digital experience. The second was that for people that are in CPG, you realize that access to data, first-party data, is really gold. How well you know the consumer, you have everything possible, and you work with the customer to get access to point-of-sale data. I mean, the possibility to really go into a company that has over 100 million loyalty consumers, and we have history across all of them, and we have such an incredible amount of first-party data. I think it's known, first-party data is an incredible opportunity. And so these two things together said, okay, let's go to another area where probably I can build on the experience that I have but really have an acceleration on what I can create there.

Yeah, so the first point that you touch is extremely important. You cannot have an AI strategy, technical strategy, that is disconnected from the business strategy. Everything started from the business strategy. I think one of the key shifts that we did in our strategy was we were really working before as I said, being brick and mortar and a digital experience totally separate. And one of the key focuses was we really wanted to put the customer at the center. So the center is the customer. We want to start thinking and change terminology in terms of saying, I need to know the customer, I need to talk in terms of customer journey, customer experience. Define the customer experience, what I want the customer to be when interacting with Walgreens and Boots across all channels. If you call a call center, if you are in the store, if you are in the website, if you are in the mobile app, I want to make sure that you as a customer are identified and we give you a seamless experience across all the different channels. That is the vision and the starting point. So I started building solutions that are specific for you. To do that, the foundation is really the data. So a company like us has an incredible amount of data by channel. So you as a customer, you are Sanjay because you are Sanjay for a call center, probably a different name because you are in the mobile app, in the pharmacy I have Mr. Sanjivbura. So the key focus was, and the work that we have done together, is how we start building the data and we create all the data in a single data lake that will allow us to say, I have now all the information cleaned. And I would say that this has been one of the reasons why we had the interesting meeting 15 months ago, because it was definitely an incredible journey. You have a company that built this massive data warehouse over 20 years plus, frankly also without any governance from my team many times. So you have the scope of what we believed was such a small one, and then we discovered that there were business tables and even applications that were built on top of it. So the program and the consolidation was a massive change and a massive endeavor. And I would say now we are in a much, much better situation. We have done an incredible cleaning and we have consolidated all the data in the cloud in Azure. So that is the first part, because you can have an incredible asset which is the data, but if they are not curated, there is nothing we can do about it. The second one is now how we can make sure that AI and machine learning are really part of every way we operate. It's not something that we do after the fact, but this is the beginning of the strategy. And I think it has been quite difficult to get to that point, because the reality is that you clean the data, check, you have the technology, all the different tools that we have in the market, and so on, then you build the organization without the data scientist. And even if you have all these pieces together, there is a big one which is the culture and the shift in mindset. It's not easy in the organization to go and say, now you need to trust the machine. And it's not easy when you have a person that is doing forecasting and replenishment, which is his job, and he believes that he knows everything, and you say, now I have a machine that is doing it for you, or an algorithm. But it is even more difficult when you go to a pharmacist and you say, well, now all the checks that you do in the pharmacies... You believe that the pharmacy is just giving you the pills, but behind there is much more science. And I discovered that as well. So the pharmacist has to make checks, looking at all profiles, checking all potential pathologies that you have, all the different allergies, all the different prescriptions that you have, all the different combinations, and then he has to really check if there are potential issues for your safety and health, and also if there are potential even in terms of lower-cost care if there are other brands. So all this part is what is called DUR. And to replace that kind of brain of a pharmacist with artificial intelligence, I think has been a big challenge, because you take years of experience, but also then you say, do you trust? And there is also regulation that you need to have to make sure that you can have the possibility of still having supervision of a pharmacist, because this is a very regulated environment. So the cultural aspect has been one of the most complex as well.

Yeah, so if many of you that are working in companies, you get a presentation from Accenture, but not necessarily Accenture, any other, they will bring you a chart that says the journey of analytics. The general analytics start with reporting, KPI scorecards, predictive, prescriptive, action-oriented, blah blah. It's a famous slide. And the key conversation on this slide seems to say that I need to have all the pieces in place before I go to the last one. And in a company like us, that means for me that basically I will never do that, because it's going to take five to six years even to clean all the stuff that I have from 20 years of history. So that cleaning, just to give a sense, when we made the move, we had 120,000 business tables that we have to move. 120,000. We cleaned them and we moved 3,000, but it took us six months to clean. So if you start thinking that you need to go perfect, with all the foundation, all the KPIs, all the data governance before touching AI, then it's probably you are pushing to never, because you will never be perfect. So I think one of the lessons learned that we had was how do we make sure that we work in parallel across all the dimensions, which is creating an incredible complexity to my organization frankly, and also to our partner, because you need to make sure that you clean the table but also you start building prototypes and so on. So I think that is the first lesson there. The second one that I think is extremely important is that it's cool, AI. Every CEO is going to read about AI, every CFO is going to read about AI, and they would say, yes, let's do something. Then when you go and say, let's invest, it's very difficult for them to grasp what it is and what you are doing, because many times there is always a thinking about you make a program, you make a project, the projects start tomorrow and finish after three months. You talk about use case, use case, and then you say, hey, but this is start, you never finish, because you have a product. So it's really the mindset shift in moving from project to product. And I think it's very important. Then you need to identify few areas where you say, let's try to make a difference and to bring tangible benefits, and you start creating advocates in the business function that are going to come to you and say, this is the benefits, and you create this halo effect. And the third one, frankly, and I'm not saying this because Accenture is here, but it's important that you have a trusted partner to work with. And the trusted partner for us has been not necessarily just a system integrator like Accenture, but many times also a technology partner. So in our case, Microsoft, because we are an Azure shop. But if you are in the cloud, you need to have a very strong technology partner there. You need to make sure that you have the right partnership for another company, whatever it is, Databricks or whatever, for the AI/ML, because especially if you want to really push the envelope, you need to have an ecosystem that is really supporting you. And I think it's been fantastic. And one thing which we really liked in the relationship was because of the very candid discussion on a weekly basis in terms of what's moving, what's not moving, being very straightforward in terms of things which are not moving well. So discussion was always focused on, okay, things are moving well, but don't talk about that, talk about things which are not going well so that we can fix it faster. The parallel processing is very key. I think it's been more complex because human mind is kind of tuned to sequential processing, and we all love to say, okay, let's build the lake, or let's build the platform, then let's do this, then let's do this. But the point is that you don't know what you're getting into. If you keep building in a sequential manner, you may land up after one year doing something which may not be as valuable as you thought initially. And that's the reason what we decided was let's do a parallel, we call it dual velocity. That's a word we normally use, saying you know one side you crank up the engine for foundation work, create the foundation, other one is to start talking to the business and start building the use cases in parallel. And which means also many times that you create a staging environment in the cloud where you move the data just for a specific use case, which is again really truly parallel to the normal migration to the program and so on. You need to, it's not just easy to work in parallel, you need to create as well an infrastructure, and many times also the team, the people that are going to make the migration are totally different from the data scientists and the data engineering that you need to have to make this kind of use case. So the thinking of working in parallel is really an investment, and it's an investment in people, technology, and process. Totally different process when you start thinking about the backlog of the use cases and so on.

So first of all, you all know that the scarcity of, and the war frankly to attract talent in this moment, is incredible. Each of us in the company, we have so many vacancies, and we are really working hard to attract the talent and to engage the talent. What has changed is, I remember that two years ago, it took me after three or four months after the lockdown, I hired my CTO, entire team, without any in-person meeting, fully online. That was my first experience three months after the lockdown. And if I go back to that moment, I would say wow, that has been quite a big change. First of all, because many times we have been hiring for the last two years people without ever seeing them, and I think it's quite a big change. The second one is there is definitely scarcity of resources, but especially in the data space. My experience is telling us that unfortunately it is not enough to have just data scientist, generic data scientist, or it is not enough just to have product manager, because everybody says I go in the market, there is a massive difference if I need you to be a product manager for a pharmacy with pharmacy expertise, with your specific domain expertise, or a data scientist without care, because you have a completely different environment where there are rules, the element of clinical trial, the element of encryption, you need to be PCI compliant, HIPAA compliant, and so on. So it's a massive difference. And to the point that we are also having a little bit of a different strategy, which is a mix. How do we also go within the function where there is a subject matter expertise, trying to understand are the people that have a kind of skill set that we can train to be a data scientist? On the other hand, I have data scientists but I need to train them. So the reality is that I need to make sure that I have a very, very strong onboarding and training program, because it takes several months to have a person that is at the level of productivity that we need with a domain expertise. So my watch out here and the biggest learning is that it's very easy to say I go and find a product manager or data scientist, but don't assume that the productivity is right away if they don't have the domain expertise. You need to build that. I think it takes time, it takes investment. And I think we are seeing the same thing. So we have a lot of data scientists at Accenture, we started building the practice long back in our company, like eight, nine years back, there was a dedicated effort to build data scientists. We actually have multiple locations. One of our European locations, a few of them, are very famous for the colleges which are in statistics and mathematics, which maybe source people from. But we also felt that the generic data scientist, I mean, you need them, but they can't translate the value. It's going to become very hard unless you know the domain and the industry which you're operating, because just to identify a use case and be closer to the use case, you need that understanding of the business. So if you get the people who actually understand the domain and they convert themselves into data scientists, or even if you're a data scientist but you start spending more time in a domain for like five, six years, you really become the person that is able to appreciate business and have a dialogue and generate that confidence in the business that you can trust the system and we will give you something which could be useful for you. So I also think, and probably this is important, at least has been one of the mistakes that personally I did, is that a lot of focus on data scientists and a lot of focus on product manager, and probably not enough focus on the engineering side. It has been always, oh, the time when the engineers were always a commodity, developers and so on. The reality is that engineering in this kind of world are extremely important and crucial, both engineering for product and data engineering. And I think it's a muscle that we need to definitely develop in-house as well, also the process of very good engineering quality and controls. I think it's very important, and I believe that it is as difficult as for the data scientists to find very, very good engineers in that engineering.

I think this is one of the most difficult parts. So right now we have a hybrid policy, which means flexibility for the employee to decide if they want, how many days they want to be in the office, how many days they want to be remote. It's clear though that the employees want to have flexibility, and I think it's good to give them flexibility. On the other hand, I think it's important as well to start recreating those moments where we are all together in a room, especially when we are team meetings, town halls, but also the design thinking sessions, that kind of creative moment where we need really to solve a problem. So the answer is that it is not going well. Probably I don't even remember the statistic, about 25%, but more important what we are trying to do is we are trying to create events where we say for that couple of days or one day a week, let's try to be co-locating in the office, because we want to continue creating this element of team spirit and strict collaboration. But it's clear that we need to give flexibility to the people as well. Now it's a long journey, and the most important part is that we are being clear with the employee that we don't know what is going to be the future. We don't know. And I think we're being clear to say this is for the time being, but let's keep having the dialogue and let's try to learn and then we adjust. Because what we see now, for example, unfortunately with a couple of new variants, the number of cases is increasing, so are we thinking about we need to go back to masks and so on? So it's not a rule that is written in stone and we are not going to change because we have to learn. This is a new way to operate and to work.

I think that first of all, I think it starts with a clear vision that we AI is transformation of all the future. So everything that we're going to do in the future is going to be based on AI. So forget about the process, yes, also digital customer journey and so on, but they are a really big element of differentiation is that how you leverage AI to make sure that you have whatever optimized experience and solution. So starting from that, the key message has been: we have an incredible asset which is the data. We have been one of the companies that really have data for everything. Now, therefore, the expectation has been: if the vision is that AI is going to be across, let's make sure that we really embed AI in everything that we are doing, in every project, in every initiative. Let's really see it can identify use case that we really see a major difference always through a benefits lens. Therefore, we have AI in forecasting, which is most the traditional one, inventory management and so on. We have AI for example in price optimization. Dynamic pricing is done through AI. Dynamic markdown is done through AI. We are moving now in the promotion, the personalization. When you are in the app, your experience is personalized based on you as a customer, based on your profile. We also show the promotion based on what you have done before, the one that we believe based on your profile are more relevant. So for example, in the pharmacy space, in the medical adherence, incredible work on the AI there for the optimization of the workflow for the pharmacists. We have AI in the maker fulfillment center. We do central dispensing through robotic and AI. So instead of having the dispensing of the pill in the pharmacy, we have through a company that we acquired, we are doing artificial intelligence and optimization of the flow for the dispensing really as in a central environment. This is just an example across all the areas. The focus for us has been really to make sure that we have clear use case with measurable KPI because the most difficult thing is when you say what's the benefits, I'm not able to measure. But really to go back to: do I see a difference? And then you can have a different technique. Like I have a control store to see when I use AI, I don't use AI. I have in the for example in some categories you use AI and some not in product to make forecasting. But unfortunately there is a lot of work to prove that the AI works and what are the benefits. So the measurement system is as important as everything else. And that's why you have to bring along as well also Finance with you and thinking totally different in the way we measure benefits and go to identify the primary KPI, secondary KPI that allow us to measure the benefits.

I think that so we keep on thinking about this in a way: what is the value I can produce? So think about pretty simply. The first thing is just the visibility of information to the business operation leaders could be a big value because you know they know more than they ever knew in the past, right, so they can take informed decisions. Right. The second thing is automation of the process itself and then also speed. So automation, cost reduction, and speed. And frankly, at a certain point, also you can do more versus the other and you are in control. So that is one of the key ones. And there is a very thin line between what you develop in-house and what you get through a package because there is a lot of AI that you can get externally when you get a package. But also there are specific areas that are so differentiating that you want to develop in-house. So for example, for us, we are a pharmacy retailer. We have a business which is very clear, which is the what we call front of the store, the grocery, where you can find the grocery, vitamins, cough, colds, but also shampoo and so on. And then you have the pharmacy business. So for the front of the store, are we differentiating for others? No. Can I really say that I can get a product that does forecasting and replenishment based on AI? I can work with the partners. The answer is yes. When I go into pharmacy, well, market leader is Walgreens, the second one is CVS, all the others are very small independent and so on. So I don't find a product. What is differentiating for me is that how I manage that. Therefore, I need, that's where I need to invest my resources to build that. And so this is as well also quite important how you decide where to buy and where to build because at the end we go back to the scarcity of resources. You don't want to spread all your talent across all the areas, but you really want to make sure that you focus them on the areas that are more differentiating for you.

A couple of definitely. So first of all, data is an asset, is an incredible asset, and how you can monetize value out of data, I think it's definitely an alternate revenue stream. So for example, definitely if you have a presence on Walgreens.com, if you have a presence in the mobile app and Boots.com, there is a possibility of optimization of media when you offer to other companies. And that's through AI you start defining as well what these tools that start explaining what is the potential revenue upstream and so on. And we are really building a totally separate division which is Walgreens Media or Boots Media, which is similar to what other retailers have done: Walmart, Kroger, 84.51, and so on. So the other one is actually in the pharmacy business, which has two dimensions. The first one is also honestly to make sure that we keep working through the improvement of the health of the population, and the second one is also you create a revenue stream. We are such an it is very important for the pharmaceutical company, for example, but also for the insurance company, to keep going faster in the market with new product, with new drugs, and lower the cost of care. And to do those two things, there is a fundamental element that is super important, which is the knowledge of the customer and the patient, and the reaction to drugs, and all the clinical history of these patients. And as I said, you have to consider that we dispense about a billion prescriptions per year. So we know basically we have an incredible amount of data about the customer, so the patient in this case also because unfortunately in terms of the population, an incredible amount of population has either one chronic condition or multi-chronic condition. So several people are so lucky in their life that they go to the pharmacies once just because they need an antibiotic, but the reality is that in the pharmacy we have so many patients that have to refill, they have really a relation with the pharmacist, and we have years of history about those patients. And this is an incredible value for the pharmaceutical company that will accelerate their possibility, and then as well also for example for the insurance company in terms of how we can help and guide the patients through the best selection on the insurance plan based on the history so that they lower the cost of the care. And so what we are really creating is capability of insight for either pharmaceutical or insurance company so that they can get the value of our knowledge. Of course, then it's done all in respect because we don't share the data, the information is sold anonymized and so on, so it's really impacted and inside respecting privacy, security, and so on. But that is what we launched clinical trials as a solution. We launched, I think, one month ago or so, and there's been as well another incredible experience on how we have been building on the foundation of having the data lake in the cloud and then building on top the solution for the insights that can be accessed in this case as a solution for the pharmaceutical business.

I think it's more in terms of again how you connect with the pharmaceutical company to provide them specific data that can help you can help them on a specific pathology. So you want to understand, you are working on developing a specific drug on, I don't know, diabetes or whatever, and so you want to have more data regarding what is the pattern. And so that you define some certain analytics that you can define for you can provide to the company based on your pool of data. Clearly, if you have a pool of data of 100 million, you definitely will have statistically relevant information. It can help in candidate selection. You know the consumer behavior exactly, you define a pattern and so on.

So first of all, we started from all possible technology. You name one, we are data integration, data and data in Hadoop, data entities, whatever. I think probably there was totally disparate. We are mainly an Azure shop. So our definition has been: okay, we move into the Azure data lake. And since we are very strong in both and we are in the two worlds of traditional data warehouse and machine learning, we are putting on top Snowflake for the data warehouse and Databricks. So this is very high level the two biggest technologies that we have. Of course, then we have all the other elements around. Because we need still to have the acceleration of the consumption through data warehouse, whereas Snowflake I think it's very strong, but also we need to create an environment for the data scientist. That's where we are leveraging Databricks. And we felt that other technologies like Synapse were not at the level of managing our data. So we use for very simple use case but not for bigger volume. We have to go to Snowflake for speed, scalability, readiness of the products as well. I think software as well. But largely the concept and principle has been to simplify the structure so that total cost of ownership can be managed more appropriately. But then in this world, I don't think you can have one technology doing everything. It's very hard to do it in this world, but you can reduce the number of technologies. Yes, we would aim to really go either way, but in this moment I don't think that there is one single technology that does both the data warehouse and the machine learning and AI. And still when you migrate from a legacy, you have a lot of data warehousing.

Okay, well I think it is putting more on innovation because we pay back.

Finance question: how do you justify your finance to support you on what you're done? How I pay for them? That's easy. I pay for them. How do we pay for them? So I pay for them. No, this is the easy answer. I agree because you need to staff up finance differently. And so I as part of the business case actually I just defined the business case both finance and HR. Because if you create an organization, because in the work that we have done, we are also clearly and probably we need to talk about this, we have done also changing the operating model because many times you have people doing analytics report all over. So we build an operating model where we say we want to have a concept of federated hub and spoke. So there is a hub that is strong in it where we really wanted to do machine operation, we want to do creation of standard report analysis, and the other core of data scientists that are really the ones that are the most advanced, the most difficult things. But the innovation should be and data scientists should be decentralized across the different functions so that we have this kind of model. But we have created this operating model and we've done this together with the Sanjay's organization where we have introduced new job families, 21 job families, so really to make sure that we identify because the first assessment that we did is how many data scientists we went with. I don't know, we had 200, 300 data scientists because everybody is cool and they say what do you do? I'm data scientist. Nobody's going to say I'm an analyst. He's going to say I'm a data scientist. I mean it's, and then when we went there and say and you make a click down, you realize you don't have them. So we really made an incredible job of understanding all the jobs, really the skill set, building the jobs, the job family of the future, and then start mapping the people to do this. And it will create a business case for it and so on, simplifying as well also the external spend through a much more disciplined strategic sourcing. We went to the point of says okay to do this I need to have definitely not only an organization of scientist, engineer, and so on, but I need to have a stronger HR partners because I need to change everything. And I need to, I'm not traditionally HR, but the HR that is really more much more specialized in this kind of skill set, and therefore also Finance. So that's the way we approach it.

Yeah, well, yes, no. In general, so first of all, in general I think they are extremely supportive. It's really supportive as I said because there is so much buzz in the news around that, everybody is hearing about machine learning, importance of the data, data as an asset, and there are enough use points, data points that tell that. Having said that, I think there is a big difference versus the sponsorship level and say hey it's cool versus many times when you start shifting the thinking from okay let's build product, let's start building really what we call the single view of customer or single view finance, and then we start enhancing it and we bring along things, versus okay now I need to have a reporting for finance and then I need another report. So it's a, I don't think the issue is the sponsorship. It is to make sure that we make it tangible for them what you deliver. Because the concept of a product and you keep enhancing it, many times you got the question says: so when is it gonna finish? Or when am I gonna have it completely done? And the answer should be never. And if you say never, they say oh my God, so what? What we are doing and what we are creating is in the spirit of education. We are creating what we call the customer experience meeting. So what we are is really showcasing the functionality. So as much as we can, we showcase the functionality about the experience, we showcase our functionality about analytics. Because although we need to make sure that from the high level sponsorship we make it tangible what you are really developing, and that is the bridge that we are trying to close.

It is the possibility on transformation and how we can change the way we work and operate. I think that retailers started a transformation several years ago in terms of how we put the customer at the center. And I think there is an opportunity to do something similar in healthcare. Healthcare in the US is extremely complex between payer, insurer, pharmacy, and primary care. And I think this is definitely a key opportunity for WBA in terms of how we wanted to pivot into healthcare and making sure that we start defining consumer-centric solutions that allow the customer, the patient, to lower his care costs. I think it's huge, based on data. And I think that's what we are looking at. So thank you very much, and thank you to all of you. Thank you.