Data, artificial intelligence and machine learning are having a profound influence on healthcare, drug discovery, and personalized medicine. On this episode, CXOTalk host, Michael Krigsman, talks with two data experts innovating in healthcare with data.
Healthcare Innovation with Data and AI
Data, artificial intelligence and machine learning are having a profound influence on healthcare, drug discovery, and personalized medicine. On this episode, CXOTalk host, Michael Krigsman, talks with two data experts innovating in healthcare with data.
Milind Kamkolkar is the CDO of Sanofi, a company focused on empowering lives through human health. Milind’s focus is driving and transforming Sanofi from a data generation organization to an insights generating organization where data is a monetizable asset class on par with product and shareholder value.
Hicham Oudghiri is Co-founder and CEO of Enigma. Previously, he managed the private sustainable finance program at BMCE Bank, in partnership with the World Bank Group, to create energy models for large-scale alternative energy projects across Africa. Prior to that, he supported an energy fund in Dallas, Texas where he was responsible for hedging physical energy assets against modeled counterparts in the electronic markets.
Transcript
Michael Krigsman: We all are so interested in healthcare. We all, at least the people watching this show, are interested in artificial intelligence, innovation, data, [and] machine learning. Today, we bring it all together.
I'm Michael Krigsman. I'm an industry analyst and the host of CxOTalk. You are watching Episode #275 of CxOTalk.
I'm so thrilled because we have two extraordinary guests. Milind Kamkolkar is the chief data officer at Sanofi Pharmaceuticals, and Hicham Oudghiri is the CEO of Enigma Data, which is a very interesting startup. We'll learn more about his company, and Milind's company Sanofi, in a moment.
I want to say a thank you to Livestream. They have been supporting us for the last several years, and they supply our video streaming infrastructure. If you go to Livestream.com/CxOTalk, in fact, they will give you a discount on their plans.
Now, this is the part of the show where I beg you to please tell a friend, and please like us on Facebook right now. Tell a friend. Tell a friend to like us on Facebook, too, and subscribe on YouTube.
Milind Kamkolkar, you have been on this show before. Tell us about Sanofi and tell us about your work.
Milind Kamkolkar: Sure. Sanofi is one of the world's largest French pharmaceutical companies specializing in a number of different areas: consumer health, pharma, general medicine, specialty and rare disease, and oncology and, of course, in vaccines. Most of my work at Sanofi--I joined hitting up the ninth month now--is really focused on helping Sanofi moving from a data-driven or, let's say, data generation organization to an insight generation organization. As part of that data transformation journey, [it's] really bringing together the best parts of our organization and covering the areas of where we need opportunistic growth and, frankly, helping us make decisions better across the firm.
Michael Krigsman: Okay. Fantastic. You're the chief data officer, and I'm sure, as we have this conversation, we'll learn more about what that actually means and what you do.
Our second esteemed guest, and this is his first time here on CxOTalk, is Hicham Oudghiri. Hicham, welcome to CxOTalk. How are you? Thanks for being here.
Hicham Oudghiri: Thanks, Michael. Thanks for having me.
Michael Krigsman: Hicham, tell us about Enigma.
Hicham Oudghiri: At Enigma, what we do is really help companies create a center of gravity around their data model. If you look at Amazon, Google, [and] Facebook, really what separates them from the rest of the Fortune 500 is this notion that all of their products are built upon a central data model, these business objects, and every single thing that you do on their platform enriches that database and that view of their world.
For a lot of companies, this has to be cobbled together, right? Data is coming from legacy applications, applications that weren't designed to speak to each other. We really help companies link their data to make smarter decisions on a unified view.
Michael Krigsman: I thought this was very, very interesting because you are also aggregating many, many public data sources that organizations can then use in their machine learning, AI, and predictive analytics efforts.
Hicham Oudghiri: Absolutely. When we give people back these enriched views of the data that they have, we layer on top of it really the whole world as a context. We aggregate information all the way from H1B Visa records to cargo container shipments, FDA data, [and] Medicare spending data. For us, being able to integrate data and allow companies to use more and more heterogeneous data in every line of what they do is pretty mission critical, and it's making sure that people can actually interface and refer to things in the same way.
Michael Krigsman: Okay. Fantastic. Now, Milind, let me begin with you. This notion of data and healthcare, set the stage for us, if you would. Why is this so crucial today? What's unique about today's environment that we need to pay attention so closely to this issue?
Milind Kamkolkar: Yeah. I want to start off by saying data is nothing new. It's been around for years. We've just been really good at accumulating more and more and more of it. But, I don't think there's ever really been a business design behind how and why we accumulate this data. Particularly in healthcare today, we're often still focused on the same questions we've been trying to answer for the last 25, if not 30, years. Personally, I think the opportunity that we have today is really around being able to ask not only those foundational questions but perhaps asking questions of new data sources, new environments that might shed a whole new light into the way in which we operate.
I call this phenomenon, at least the way in which we're organizing it in our company, all around the notion of doing things better versus doing better things. The doing things better is all around the operational effectiveness of your decision-making. Are you looking at data that's findable, accessible, interoperable, reusable, and really using, if you will, that fair data standard as the FICO score for your information sets?
But, it doesn't stop there. Once you've gotten that piece in play, how do you start using these new data sources to be able to ask questions to uncover new insights that you may never have had before? I think that opportunity with the advent of big data, if you will, even as a data management infrastructure, the advent of machine learning algorithms that can actually operate faster because, again, that's nothing new. It's been around for some time. But, it's really around this high compute infrastructure now and the availability and accessibility of this data that's allowing us to do new things in a far more profound way. And, more importantly, being able to use those new insights in ways that impact healthcare in a more positive way.
Michael Krigsman: Hicham, I love this notion that he was just speaking about of using data to do things better. You're in the data business, so how can we use data to do things better, to actually make better decisions?
Hicham Oudghiri: Think about it this way. The whole industry is moving towards more and more personalized delivery. When you do a clinical trial today, it's possible you're doing it with hundreds of people as opposed to the thousands or even much more so than you were doing before. The barometer for data quality, using data well, and doing things better, it's not just opportunistic. Literally, the science is leading us in that direction as well.
Milind Kamkolkar: Yeah.
Hicham Oudghiri: Everything from the manufacturing process being tighter from a quality perspective to being able to, on the backend, manage and triage adverse events faster, to the regulatory submission process when you're researching drugs, all of that is moving towards having data being corroborating evidence to deliver more and more personalized drugs and to be much, much faster in the process of research and development.
Michael Krigsman: Milind, I know you have thoughts on this. [Laughter]
Milind Kamkolkar: Yeah, absolutely. I couldn't agree more. I think it's one of those areas where, to echo what Hicham has been saying here, ultimately, this comes down to, how well do you operate your business and how effective is your business in being able to address unmet needs in the consumer space, in the patient space, or customer space in this instance.
I think what we've begun to do much more effectively now is use that data to deliver personalized content for the right customer through the right channel at the right time. It's not limited, right? When you think about that, that's almost marketing 101. Frankly, it's not marketing 101; it's basic human information engagement. That notion of basic information engagement, I think we've gone through this dawning of the digital age [where] there are many companies that are incredibly innovative in their digital presence but are absolutely pathetic when it comes to decision-making for their business.
I think that's the big game changer here. When we talk about data as being oil, gold, or whatever other appropriate highly quantifiable entity is, it's that notion of saying, "Can this data actually drive better business outcomes? As a result of that, can our firms become smarter in the way in which we make those decisions?" I think that's when you start tailoring into the world of algorithms as well.
I don't want to take away, though, the importance that with good algorithms and good data comes the need of basic foundational elements of data, which includes governance, change management, the area of data quality. I mean at what point do we agree the threshold of data quality is good enough to at least get the ball rolling?
I think, once you start doing those and that to me is more the operational effectiveness of how well you run your data operations, you can really start managing a portfolio of investments across your data and analytics workspace. I like to call it your systems of record. It should be no more than 40% of resource allocation. Systems of innovation where you're competing -- sorry, differentiation where you're competing should be the last 40%. Finally, the leapfrog capability comes from those newer data elements and the newer algorithms you deploy in the systems of innovation.
Today, almost 98% of our resource allocation is still kept captive in systems of record. Most of the time, we're not even doing those well. I think that's where the opportunity is to start embracing these newer technologies, capabilities, data, and so forth with, of course, a profound impact on change management. What do people do now versus what do machines do now?
Hicham Oudghiri: If you think about it, there is so much scale inherently in that supply chain from research to patient outcome. If you just get one part of it right, you're affecting millions of people, you're affecting matters of life and death, and sickness and not, right? The ability to do something that is robust enough and sustainable enough that you can actually build on it and start innovating on top of something that you can push out to a variety of areas, I think that's very, very important for pharma. I think the accessibility of data is changing ways in which pharma companies even organize themselves.
Milind Kamkolkar: Yeah.
Hicham Oudghiri: I see a lot of pharma companies who organize themselves around basically brand franchises; this drug has this research database, this marketing database, and this quality database. Data is being able to kind of substantiate more of a hub and spoke model around information access for the first time. I think that's a big change too. These are cultural changes happening because of the way information is liberating, basically, resource access.
Milind Kamkolkar: No, absolutely, and I think one of the things that tie into that is this whole notion of, with data being more available on the Web as well through the Internet, the one really nice thing that's happened is where we used to have these, let's call it, cultural boundaries that existed, well, the Internet doesn't have those per se. When we're seeing things like GDPR, et cetera, all starting to introduce itself now into the world around data privacy and security, but really patients who are describing or customers that may be describing disease outcomes or parts of their patient journey in a language that is non-native to, perhaps, say, folks in North America, you can still actually leverage their data to see, particularly in multicultural countries, do the same kinds of people, for example, share the similar kind of impact or not? I think this is where this world of personalization makes such remarkable difference because, whilst the Internet may not have geographical boundaries per se, the reality is you can now start leveraging that and translate that into your domestic markets where things start to become quite real as well.
Hicham Oudghiri: I totally agree. The only problem that kind of comes out of all of this is, we do increase that noise to signal ratio immensely. Kind of what we need to do as practitioners in this space is remember that basically there has to be some domain expertise that comes back into the creation of the data, the curation of the data science behind all of it. It's why data science, frankly, is a multidisciplinary effort.
Milind Kamkolkar: Yeah.
Hicham Oudghiri: There's been, in my opinion, a little bit of a rush in the last couple years towards the infrastructure and the compute and not enough of a marriage in between the underlying science and the domain to basically make sense of all of this flood of information that's coming. We've seen it in a variety of areas where it really, really can hurt, and it's been nice just anecdotally as a young company to be able to work closely with people who actually understand it. That's, I think, the cool thing about pharma is that there is so much domain expertise. The problems are very deeply rooted.
Michael Krigsman: Milind, can you give us some concrete examples of this that the audience, who may not be healthcare or data experts, would be able to understand or relate to?
Milind Kamkolkar: Sure. Let's take the example of outcomes-based evidence. This is a big area for many pharmaceutical companies. We know that contracts with payers, with insurance companies, or even regulators in, let's say, more socialized healthcare systems across the world, are delving into i.e. the notion that you only get reimbursed if your medication proves a positive health outcome. In these instances, this is where this new data makes a profound impact because, for example, geographies that might be very high in pollen count might not always be producing or, in some cases, might be producing the right kind of results depending on patient population in that area. You'd be surprised how often we take for granted the notion of weather, but don't often integrate it into our treatment paradigms, into our outcomes-based contracting.
I think what we're starting to see now are people saying, "Well, hang on. Are these new variables that are coming into, let's call it, clinical settings actually more important, because they start to give a more holistic view of a patient's journey through their treatment paradigm?"
I think the other one that's very interesting here is what I would call the Fitbit paradox where we're seeing a tremendous uptick of digital gadgets. I think the word of today in this year's CES conference is you could literally put Alexa on anything. As we're seeing this sort of profound impact in digital health, one of the struggles we're facing is a lack of data that's being used for preventional studies; i.e. people who buy a Fitbit are generally fit, right? The people who probably really need Fitbit may not be using it as effectively. You start getting into data biases that come into play in the world of evidence and evidence-based outcomes.
The last area that I would see where data is being used into this specifically is the world of effective computing; i.e. how do you really measure pain? Should it still be on the same ten-scale threshold? Should it still be using the same emojis? Now we've got emojis online, and we can start collecting that more frequently through online measures. Do those now become the new clinical biomarkers, if you will, or digital biomarkers of pain?
All of these sorts of things, when you start combining them together, start giving you both a more real-time context under which you can start creating new interventional studies or interventions that prescribers and/or patients can sometimes self-administer, versus how that gets reflected now into the economics of healthcare management, i.e. when is a reimbursement most relevant? If you ask them to get high pollen counts, for example, the likelihood of payouts during that time in an economic burden or burden of disease area would also be probably for those who suffer asthma and other pulmonology related or immunology related diseases.
Michael Krigsman: Hicham, any thoughts on practical implementation examples of data similar to what Milind was just describing?
Hicham Oudghiri: Yeah. I'm really fascinated. I like Milind's example of real-world evidence because it's this notion that data can connect parts of the system that haven't been connected together tightly, so the actual drug manufacturers, the payors, the providers. I love the problem of pharmaceutical safety as well. I'm really passionate about what it takes to pull drugs off the shelves when you hear, historically, root cause analysis takes close to a year to bring together an instance of an adverse event. Someone took a specific medicine and got really sick in a way that they didn't expect, and the time that it takes to understand how that drug was manufactured. What are the quality indicators of that specific batch?
We now have sensors on boat shipments and pallets recording the temperature of each and every movement of the pharmaceutical batch across the chain. That stuff is being analyzed in real time and so, when an event like that happens, we can trace the lineage of that event all the way and do the analysis in real time. Furthermore, we're connecting that supply chain together, but we're also connecting kind of external evidence to the regulators, the FDA, the World Health Organization, who will report these. We'll have individuals call into call centers. We'll have doctors and hospitals kind of be redundant in the way they report these things, and it costs a lot of time and effort to triage all of this information, and all of it kind of comes in an incomplete way.
What we can do now with data is actually stitch it all together really, really fast in real time. Instead of having doctors figure out kind of who is on first, we're actually investigating why things went wrong as opposed to gathering evidence. Kind of reducing that rule to something that's much more palatable and much more efficient for everyone's safety.
Milind Kamkolkar: Yeah. I think, to echo Hicham's point, there was a really good point raised around the connectivity. In the early 2000s, we had a lot of these buzzword phrases like bench to bedside, molecule to market, and these sorts of things. Whilst the hype cycle of that was very promising, it went through a bit of a lull because we realized we just didn't have the infrastructure to do that work.
Nowadays, I think, with the newer capabilities that are coming through, that data has the providence and lineage to really be able to be tracked at a discrete level. What machine did this particular batch of compound get produced at? Could you really quantify if a product label change has to be made as a result of the adverse event that's being reported in a call center? Is that propagation going to happen consistently? Is it going to happen in a way that's also meaningful where maybe the pharma companies don't need to report to the regulators anymore, but we give the regulators direct access to a portal to say, "Hey, you know what? Take a look through our process. You can see it anytime you want to make sure that this kind of quality and compliance is coming through the system." I think that's kind of the power of the transformation that we're seeing today as well.
I think the other part around cost is a big one. All of these processes cost money. Any kind of interventional state, whether it's algorithmic, data stitching, or data transformation that can happen in play, if it does reduce the cost, the anticipation is that these, of course, will be benefits that get transferred back to the patient population.
Michael Krigsman: I want to remind everybody that you're watching Episode #275 of CxOTalk. We're speaking with Milind Kamkolkar, who is the chief data officer at Sanofi Pharmaceuticals, and we are also speaking with Hicham Oudghiri, who is the co-founder and CEO of Enigma Data, a very interesting startup. Right now, there is a tweet chat going on. You can ask your questions of these two very smart guys using the hashtag #CxOTalk.
Let me address to either one of you. You've both been painting a vision of data linking together disparate parts, what today are disparate parts, of the healthcare system: payers, insurance companies, doctors, hospitals, let's not forget patients.
Milind Kamkolkar: Of course.
Michael Krigsman: And then, their devices. You're both painting this holistic picture. What do we have to do? Where are we today, and what do we have to do to achieve that vision you've both been describing?
Hicham Oudghiri: I'm happy to take a quick stab.
Milind Kamkolkar: Sure.
Hicham Oudghiri: Definitely an incomplete answer, but there are a lot of very encouraging, big, bold bets being placed. If you take a look at CVS and Aetna, or even the announcement a couple days ago of Amazon partnering up with JP Morgan and Berkshire Hathaway to deliver more unified care, these are moonshots and quite interesting. They're very calculated. I think that the risk on the data side of the equation coming together, becoming lower, and becoming more a game of, how well can you do it?
There is a lot of low hanging fruit in our industry to make data better, and I just want to harp back on that one point that you made earlier, Milind, which is, if it's not done with an operational mindset, you're creating a little more debt despite the amount of innovation that we're springing out.
Milind Kamkolkar: Yeah.
Hicham Oudghiri: We won't be able to kind of reduce the entropy in this complex system.
Milind Kamkolkar: Yes.
Hicham Oudghiri: I'm very encouraged that people have the confidence that they're doing it the right way. We certainly see that. I certainly see how pharma is investing in that. As to how it gets done technically, I have a bunch to say, too, but obviously, I'd love to hear what you think about people's confidence in that respect.
Milind Kamkolkar: Yeah. I think, on the pharma side, I would say it's a cautious trepidation. I think we are taking much bolder bets, but it's all relative speaking. To that extent, I'm incredibly encouraged just seeing the recent investor calls between Roche, Novartis, [and] certainly our own. I think pharma CEO today, or every large healthcare CEO today, is in their investor calls relating to the importance of digital and data.
This really is a golden age of seeing these things come together. I think, when you have leadership that is coming down and saying, "This is a strategic objective for us, and we're going to apply it in the areas of finding the right patients for the right trials, optimizing clinical trials so that the cost can be transferred further down the line, ensuring that we're engaging more effectively so it is the right content, and it's not this superfluous advertising spend that we often see today." I think those are the areas that we're seeing some really big moves.
I'm particularly encouraged by some of the work that's going on in deep learning and the applications of machine learning technologies in operational effectiveness, simple things of saying, "Can we automate FDA submissions through using national language processing?" We're doing some pretty cool work on that, and my peers are doing some really cool work in that stuff in the other companies.
The world of blockchain, I know there's a really nice healthcare consortium coming together between a couple of us in the industry where we're openly sharing our experiences in working with these different platforms and technologies and trying to address, "Okay. Look. There's a really cool tech that's out there, but is it really going to make the best sense for us?"
I want to give you a really clear example. We talk about blockchain, for example, in the world of counterfeit and being able to prevent counterfeit. But, you often wonder and, sadly, that most of this counterfeit action often happens in emerging growth markets and/or parts of Africa.
Here we go. We take a really big energy consuming infrastructure like blockchain and apply it to areas that have very little to zero bandwidth. It makes you wonder, "Is that really the best application of blockchain that we see today?" Structurally, it may not work, but the idea is correct.
I think what we're starting to see now is that we're not jumping to technology first. We're actually addressing what is the problem we're trying to solve before we go into those spaces. I'm starting to see at least more relevant questions, more interesting questions being asked and the technology, honestly, taking the seat it should, which is, okay, let's see how we can best do this now.
Hicham Oudghiri: You can really proxy into the answer of any question.
Milind Kamkolkar: Yeah.
Hicham Oudghiri: It's this notion that kind of all models are false.
Milind Kamkolkar: Right. [Laughter]
Hicham Oudghiri: Some are more useful--
Milind Kamkolkar: Yeah.
Hicham Oudghiri: --than others, right? And so, we work a lot with public data. We see people all the time doing things like trying to cobble together claims data from a variety of providers. It turns out Medicare puts out this information for free for everyone.
Milind Kamkolkar: Yeah.
Hicham Oudghiri: Sure, it's not the whole population, but you can get a sense of what's going on. I see this on the data side. I see this on the technology side, vis-à-vis your question about blockchain. Absolutely right, the best answer may actually not be the best answer.
Milind Kamkolkar: That's right.
Hicham Oudghiri: It may just be the best theoretical answer. [Laughter]
Milind Kamkolkar: Yeah.
Hicham Oudghiri: This space has, I think, a culture of kind of engineering and setting things up to deliver. It's different than if you are kind of marketing and delivering ads on the Internet, which is what big data has been for a really long time. The challenge that pharma has is it can't be as fast as loose, and I think that that ingenuity is coming through that challenge.
Milind Kamkolkar: Look. I fully agree. I think, to me, the most satisfying thing certainly I've seen in the last nine months here at Sanofi and certainly speaking with my peers in the other companies is, we're finally embracing agile. We're finally embracing the notion that it's okay to be a little scrappy within reason. I would say areas with respect to patient safety, et cetera, absolutely no compromises in terms of quality and safety. But, I think in some of the other areas, this notion of you can work agile, and it's okay. I would not say it's okay to fail. I would say it's more okay to experiment smartly. When you do that, being scrappy, but recognizing that much of this is still very much a marathon, I think is a really positive sign that I'm starting to see come through.
The other area that I see is incredibly encouraging is the world of open source. We're starting to see a bit more around data sharing, around data observations, around algorithms that are being published now in open source as well. What I would love is, and I guess this is a challenge not only to myself but also to the industry, can we get past the open data stuff but really start thinking about open data model sharing?
I think, when you can start deriving data models that are more relevant to either disease or other such areas, the data flows that come through actually feed into a significant amount of data prep work that no longer has to be done. I'm not saying it goes away, but at least when you have an open data model, we're speaking from the same platform. In the absence of that today, this is where you do start getting those nuances in data clarity, data quality, and sometimes observations that, frankly, may not be as intuitive as what they appear.
Michael Krigsman: We have an interesting question from Shelly Lucas on Twitter. Shelly, by the way, is a content marketer, Internet influencer, strategy type of person who is just among the very best out there. I know Shelly well. She asks, "Pharma has marketed directly to consumers, but will it need a new engagement model with this increased data sharing?"
Hicham Oudghiri: Yeah.
Milind Kamkolkar: Yes, absolutely. Shelly, I think you hit the nail on the head. If you think about the kind of work that we've done in the past, content and engagement strategy wasn't necessarily the top of the list. At least, if it was, it wasn't always done effectively because you have these long brand plan sessions over a year where the ability to change interventions through, let's call it, the approval process was quite difficult and often taking time.
I see there definitely needs to be a new engagement model where the rep is, in fact, one of many channels. Historically, it's generally been rep led conversations. I don't think human interaction is going to go away. I still think there is a relevance for reps, but the nature of what reps look like, I think it's going to change more into this scientific liaison and much deeper conversations around patient archetypes, around genomic discussions, around things that patients are actually searching for answers. Maybe it's a more sophisticated Dr. Google [laughter], for example, that comes through, and that becomes a new channel that's relevant.
I think content marketers, in general, and the way in which we've been engaging, in general, is going to go through a whole new world. Even the notion of an agency today is already going through its own challenges. You're seeing this evolvement, if you will, in the content marketing and marketing space, in general.
Hicham Oudghiri: Can I take that question with a fast-forward 10, 15, 20 years from now attitude?
Michael Krigsman: Sure.
Hicham Oudghiri: The thing that I keep thinking about at night is, what if we had the data work for us and what do we need to do in order to get there in a safe way? Right now, the diagnostic capability of pharmaceutical companies to deliver personalized medicine, we could literally be going to the doctor and being told exactly what we should be taking, right?
Milind Kamkolkar: Mm-hmm.
Hicham Oudghiri: Multiple companies could be getting that information and analyzing it in real time to deliver very personalized medicine. On one hand, that is a really cool feature--
Milind Kamkolkar: Yeah.
Hicham Oudghiri: --where we're being taken care of. On the other hand, there is a lot of data privacy concerns going on. I know Europe is at the forefront with GDPR in this respect, and I know that there are regulations coming in May around basically what is the management of PII data look like. I think these are some of the biggest questions over the next 10, 15 years. The almost ethical question is, how do we enable pharma to know, essentially, our body in a way that we feel comfortable with? That's quite an engagement strategy.
Michael Krigsman: On the subject of the ethical question, Milind, let me direct to you that's coming from Zachary Jeans on Twitter, a really interesting question. He's asking, "What are the potential dark sides of big pharma companies leveraging all of this data," as Hicham was just describing?
Milind Kamkolkar: Yeah. Look. I think one of the things we've done specifically is established an ethics board, an external ethics board, that looks at not only content but also algorithms that appear in black boxes. Honestly, my biggest fear, independent of whether I work for a pharmaceutical company or not, is exactly the same fear that you have, which is, we talk about digital disruption often in a highly positive way, but we can't forget that there is the Black Mirror effect in all of this.
Hicham Oudghiri: Yeah.
Milind Kamkolkar: Which is, the outcome of abuse. Sadly, this happens across the board. It's probably happening already today. Do we really know how our data is actually being moved across the board and is being used for different reasons? I would hope most of it is positive, but I think we can all agree there's probably nefarious things going on that we simply don't know about.
I get the sense that, as a company, we need to do everything we can. I think there are new, let's call it, ethical terms that we need to start addressing now as a pharma company, as n industry body. It's not just ethics in terms of what information we can or cannot use. There are a lot of intended use guidelines around how we use data but, more importantly, the ethics around the societal impact of that data, the societal impact of black box algorithms perhaps negating certain functions that physicians do today. We're already seeing this happen in radiology. We're seeing this happen in numerous other professions in the physician world.
My fear is, this is something that a pharma company alone cannot handle. I do think it's an industry, it's a healthcare industry consortium thing that we need to address. I think that things like blockchain may bring about a better-trusted relationship there with our information sets and so forth. To be honest with you, there are many things that could be done negatively, none of which I anticipate our company is doing today, and I don't think, by design, they intend to do that. But, like anything in this world, the minute you go online, your information could be used ten ways to Sunday, and not all of it is going to be great.
Michael Krigsman: Hicham, Arsalan Khan is asking, in this same vein, "What about data corruption that takes place either on purpose or accidentally? Who governs these open data models and the data sharing," that you both have been advocating so strongly and with very good reason, of course?"
Hicham Oudghiri: That is a great question. One of the most important things of any data transformation journey is lineage, understanding everything about where the data was produced. Much like it's very important to us to know how our food was produced, it's very important for us to know how the data landed, who transformed it, why, and what. We're trying really hard to do that, I think, as a community of practitioners, without slowing down results. There are many, many ways. AI and machine learning have been helping us to do that now. I'm seeing a lot of positive stuff, but it's a quality standard that we have to keep.
Michael Krigsman: Hicham raises a great point. How can machine learning, AI, and these new kinds of techniques that are based on data help push innovation forward without suffering the potential kind of issues that may happen when data is aggregated and shared like we were just describing?
Milind Kamkolkar: Yeah, sure. I think the first thing we need to recognize is that AI and machine learning, any of these coding languages is inherently a model, i.e. it lacks morality. It's the people who code them, the discipline, and the ethical boundaries under which either they've been raised or otherwise lends itself to the behavior. Also, once those algorithms go open, i.e. they're part of an engagement plan, it depends how we engage with them as well. What do we train those algorithms on and the data that gets trained into it?
When I think about how we approach this, one area that I've been very adamant on is the world of diversity and knowing that diversity and diverse thinking, if you will, and people coming from diverse backgrounds almost forms the crux and basis under which you should even deploy an algorithm, let alone design it. That's number one.
Number two, I think, as a society, we need to start raising the bar on how we start engaging with these platforms. We've seen Microsoft had, unfortunately, a disaster of an experience with an algorithm or chatbot that they put out that unfortunately was getting trained on by a 16-year-old or a 15-year-old, rather vivacious teens that were using all sorts of language and very demoralizing kind of language, highly sexist. They actually had to pull the chatbot off. I think it's a two-sided equation. Not all of it is just on the pharma side. Most of it has to come from a diversity of talent that you put into the program, but then also the appropriate coaching then of how we engage so that, as the model starts enriching itself, you have the right ethical guidance in terms of how we engage with it.
Michael Krigsman: Any thoughts on this, Hicham?
Hicham Oudghiri: The diversity point is probably the most important one. I could not agree more that challenging this sort of technology from as many viewpoints as possible is the best way forward. I've been very optimistic about what we do as a society and people, as for how we come together. We want to create that opportunity to see things differently.
Milind Kamkolkar: Yes.
Hicham Oudghiri: Remember, the impact of AI and machine learning is that it scales fast, and it's designed to learn and reinforces biases.
Milind Kamkolkar: Yeah.
Hicham Oudghiri: It's a self-optimizing system. The opportunity to challenge it, the design of multiple points of view, that must be inherent in how we go about it here.
Michael Krigsman: How do we do this right? Milind, how do we do it right?
Milind Kamkolkar: I think it comes in talent acquisition, too. When you put teams onboard, make sure they do have opposing views. I think the echo chamber phenomenon that we're seeing, be it in politics or otherwise, is something that's come culturally over a period of time, but it doesn't mean it has to be integrated into our algorithms, let alone our data systems.
The data exists. It's what we do with and how we do with it that's important. I think it starts there. Putting teams together, an appropriate ethics board around it or some kind of principles of design that go around it, and really start thinking about, as a human, how would you feel about this, these responses and this engagement? It comes down to delivering a customer experience that is relevant, an engaging experience that's relevant and meaningful.
Michael Krigsman: We have about two minutes left, and so in 140 characters or so -- [Laughter]
Hicham Oudghiri: [Laughter]
Milind Kamkolkar: [Laughter]
Michael Krigsman: Hicham, I'm going to direct this one to you. [Laughter] Milind just raised the issue of customer experience. What does customer experience in healthcare actually mean? What does it consist of? In ten seconds, can you summarize that entire body of knowledge and way of looking at the world?
Hicham Oudghiri: Honestly, it's abstracting away most of everything we talked about and not making it complex. Just better patient outcome. It's more transparent, and it comes with fewer caveats. Leave that to the people who are working for those patients and make that relationship clear.
Michael Krigsman: I love that. Milind, your final take on this notion of customer experience and algorithms.
Milind Kamkolkar: Yeah. Make it convenient. Make it fun. Make it meaningful.
Michael Krigsman: Boy, you guys are quick. [Laughter] Any final, final thoughts? We're really out of time, but how about final thoughts from each of you. Who wants to go first?
Hicham Oudghiri: All right, final thoughts. I think one of the opportunities specifically in healthcare is that it is a system that has all of this compartmentalization. You do have the providers. You have the manufacturers and the drug makers. You have the people doing research in a more isolated way.
I really think data is going to bring the system together. I think it's a complex system for a reason. It's amazing what we've achieved, but I think we'll be able to iron out a lot of the kinks in the next coming years and make the whole system feel like more of one thing for a patient. I know that's a huge source of confusion and anxiety for people who are just trying to get better.
Milind Kamkolkar: Yeah, I would just leave it with, I think if more people contracted obsessive compulsive data disorder, the better it's going to be.
Hicham Oudghiri: [Laughter]
Milind Kamkolkar: I think when we're aware and conscious of how we use information and for what purpose and be relentless about the problem you're trying to solve and ask more questions, I think the better it's going to be. I'll leave it with saying don't let technology be the lead in this instance. Really focus on, what is the experience you're trying to deliver; what is the problem you're trying to solve? Let technology really just do its role as an enabler.
Michael Krigsman: Okay. Well, with that, what an interesting and very, very fast discussion. I would like to say thank you to Milind Kamkolkar, who is the chief data officer at Sanofi. Milind, thanks. It's great having you here, and I hope you'll come back and do it again.
Milind Kamkolkar: Absolutely. My pleasure.
Michael Krigsman: I would like to also say thank you to Hicham Oudghiri, who is the co-founder and CEO of Enigma Data. Hicham, thank you for being here. I hope that you'll come back and do it again as well.
Hicham Oudghiri: My pleasure. Thanks so much for having me.
Michael Krigsman: You have been watching Episode #275 of CxOTalk. Please like us on Facebook and tell a friend right now that they should also like us. Thanks so much, everybody. I hope you have a great day.
Next week, we are back, and we have amazing shows. We'll see you soon. Have a great one. Bye-bye.
Published Date: Feb 02, 2018
Author: Michael Krigsman
Episode ID: 500