Welcome back to the Harvard Data Science Review podcast. I'm Liberty Vittert Capito, feature editor of the Harvard Data Science Review, and along with me is my co-host and editor-in-chief, Shaoli May. Every month, we sit down with people shaping how data science actually gets talked about. And today, we're going behind the scenes of one of HDSR's most read columns, active industrial learning. If you've spent any time on the HTSR site, you've probably run into this column already. It's the place where the gap between what data science theory says and what actually happens inside a company gets bridged. Practical, sometimes messy, always grounded in real organizations trying to get value out of their data. Our guests today are the two people steering that ship. Hamid Hamouchou brings three decades of industry and consulting experience in data and AI, and has also spent years thinking about how organizations build data literacy at scale. He's also the co-founder of the Initiative for Analytics and Data Science Standards. Alongside him is Miguel Perdi, his co-editor who's helped shape the column's recent run of conversations with senior data leaders across major companies. Miguel works as a partner at Kearney, helping enterprises with their AI strategies and transformations. He is an executive fellow at Harvard Business School, where he does research on AI and strategy and boards and the impact of AI training. Together, they've brought readers interviews with executives like Todd James at 84.51 Degrees, the analytics engine behind Kroger, and Ashok Srivastava into its chief data officer, plus deep dives into how data science teams actually get built and what happens to a model after it ships. Today, we're going to ask them to pull back the curtain. What's it like editing this column? What have they learned from the people they've interviewed? And where do they want to take it next? Let's get into it. First of all, Hamid and Miguel, thank you so much for doing this podcast. And as the editor chief of Harvard Data Science Review, I obviously particularly appreciate that both of you serve as a co-editor for this column called Active Industry Learning. So the first question we have is for listeners who haven't read your column. How do you describe the Active Industry Learning column in terms of its mission, in your words? and what kind of gaps, you know, were you trying to feel when you took over this column? Maybe I'll kick this off, Miguel. So I've been with the column, I think, for about three, three and a half years, roughly. And of course, the world has changed tremendously in those three and a half years. And I'm sure we're going to get into that. But obviously, a lot of data science has been since I guess the term came about over a decade, a decade and a half ago has been applied, right? So organizations using the variety of tools, techniques, technologies that we maybe collectively call data science to make better decisions, to drive their organizations more efficiently, more effectively. It really is an applied discipline from a lot of different perspectives. So I think the column aims to bring that learning from applying data science globally across organizations, sectors, individuals, practitioners, and see how we can share that with a broader ecosystem so people can benefit from each other's learnings. So not only from all the great research that is shared across different parts of HDSR, can we bring a lot of insight from the field in how it's being applied in real life? And of course, AI has changed a lot of things for us. But Miguel, you came on board, I think, about a year ago. So maybe it was slightly different from your perspective. I don't know what you saw when you first came on board. Yeah, Hamid, thank you. I think when I came on board, first of all, I was very honored to be invited. And I think what I saw was an opportunity to really highlight some of the elements that many times are not discussed in traditional publications that are geared towards enterprise, corporate business leaders. which are usually where things break, right? So like our active industrial learning column focuses on applications of data science to businesses. And I like the phrase we have as a part of the description, which is with an eye toward practical considerations that make data science initiatives more likely to succeed. And obviously, AI initiatives more likely to succeed. And I think this is very important because you can talk about the technology, but really what usually is not discussed is those nuances that, again, make things break. So I think the focus of those insights from the people that are publishing and the highlight of those insights for people that are reading, I think is what really was appealing to me and which I see in my work with executives, with companies that becomes the key success factor to increase the probabilities of value capture with AI and data. The interviews you've published now as co-editors, you know, so Todd James at 84.51 degrees and Ashok Srivastava at Intuit, was there a moment, I guess, in either conversation that genuinely changed how either of you think about applied data science? Yeah, I mean, I was fortunate enough to, you know, conduct those two interviews. They were really illuminating for me in different ways. You know, with Todd, this is a very large retailer, right? There's a lot of challenges, a lot of data. They've been around for a long time. So there was a lot of deep questions that they were dealing with. And Todd just did a great job. In his new role, he tackled those. So it was very kind of real world, you know, large organization. If you're a chief data officer, if you're a chief AI officer, I think his interview was very illuminating because I felt like a lot of people could relate to that very easily. Ashok's was a bit different. They'd been working on AI, working with AI since before the chat GPT moment. So there was a lot of accumulated expertise in using AI technologies for better decision making. So that conversation revolved around AI. And I tried to get a lot of really specific, detailed, initiative-level information from Oshok, and he was very generous with that. So we have figures, you know, a number of decisions being made with machine learning algorithms at Intuit, which is another huge organization. I think that was probably useful for our audience, HDSR audience, because maybe it gave them some benchmarks almost that they could use or compare themselves against. Yeah, maybe what I'll add, I thought both interviews were really good, and Hamid did a great job at really getting a lot of the insights and the details of what makes things work and what are like some of those double clicks that leaders need to know that are important. And what I really enjoyed about Ashok's interview was he mentioned and highlighted the need for ethical AI practices to prevent negative consequences like misinformation And I think responsible AI ethical AI is not something that is talked about as much And many times when it is sometimes some organizations are talking about it because they know it's important, but they don't necessarily have a good framework or don't know what to do about it. But I think Ashok had a really good view on this as some practical considerations that are worth looking into. I mean, you're the one who really brought the topic of data literacy to this column. You also organized a great workshop for HDSR led to a special theme. So I want you to share with our listeners a bit about your motivation, the visions, and what you see are the challenges and why particularly you emphasize that for the industry and the business sector. I mean, that experience, you know, from my perspective was great. We basically got to experiment to see what happens when you put about 50 to 60 really smart and thoughtful people in a room and give them a challenge of how they define, they use, they try to improve data literacy in a variety of situations from higher education to K-12 to workforce to society more broadly. So that experiment itself, I think, was valuable. Of course, we were able to get about five or six different articles as part of a special theme. But beyond that, I think that format was very useful. And it's one that we think we could repeat in other instances because we have an ability to focus people on a specific topic and then continue their collaboration afterwards. And what's great about data literacy or AI literacy is it concerns the entire organization. So we're not talking about a department here necessarily, right? We're not talking about challenges that are important for, I don't know, one or two percent of a company. But when you say data literacy or now more and more AI literacy, it really is about 100 percent of the company from your front line to your back office, to your operations, to your marketing and all of that. You know, in terms of providing content that is relevant at such a scale, I think should be an important focus of the column. And that exercise gave us that opportunity. I don't know, Miguel, what you would want to add from an AI literacy perspective. Yeah, I know. I think the whole literacy space is super important and it's becoming more and more important now. In the past organizations, when machine learning, like data science, optimization, or like the thing, and it wasn't as prevalent as AI is today, I think people didn't really think that much about training because your stakeholder technical base, the engineers, was small. And you really didn't have to train and educate the whole organization. I mean, some progressive forward-looking companies did because they knew that if they really helped the business people that were not technical understand what data science was and what could do, that would be the most beneficial way of creating value. Because you basically put glasses on someone and they can start identifying opportunities to bring in the technical team. with AIs lowering the bar to entrance and everyone being able to use these tools, that just made the AI literacy imperative extremely important. And that's why we see today like so many organizations really thinking of like, how do we train people? And again, they have in some ways because everyone has access to a ChatGPT, a Claude, a Gemini. And that basically requires that there's some training and some literacy around, maybe not necessarily what is going on under the hood, the data science, the AI, but yes, at least on how to use these tools responsibly, how to use them effectively. Miguel, you know, you recently co-authored a piece with Thomas Davenport. What makes this piece different? Why does it work with this column? What kind of columns in general are you all looking for to really run a through line through these pieces? So we, together with Tom Davenport, who was a co-editor before me, and actually thanks to Tom, I met Shali and then became a part of HDSR. We explored basically the question of, there's always been talks about, oh, AI, robots, automation will take our jobs. And we humans, very, very smart people, have been predicting what jobs and how many jobs will be lost. So we basically looked at the literature and looked at the main sources of organizations that were predicting and what they said and how they updated these predictions. And we were trying to understand, like, do we really know? Can we really know? And especially in this new way, can we really figure out what jobs is AI going to take or will the robot take? Our conclusion was basically like, we never get it right. But I think that's like, everyone knows that. Like, everyone knows that predictions are hard. I think our point was more like, there are things we know. there are things we don't know. And if you look at history, you will see that certain things, depending on the technology, can start being replaced or accelerated. And we focused on that. To your question of like, what are the type of articles that we want to have in our column is we want to be the source of truth. We want to be able to help approximate that. We don't claim that we know the source of truth, but we want to bring in this objective source of truth as much as we can. So when executives, when business people with data science leaders, AI leaders, they're trying to figure out, hey, I need to understand some implication of a technology of AI to the business, they know that they can turn to HDSR in our column and say, we know that we're going to get very good, very credible information from top people that are actually doing this in the field or in academics working with enterprises and therefore really have a good domain knowledge of how the technology is implemented and the complications of that. One thing I can add, and I think what we've been trying a lot to do, and the interview is a good example of it, we just want to create different ways of engaging also with our column, right? Because not everybody thinks or contributes the same way. So if it's through an interview, that's great. If it's through an article submission, that's great. If it's something that either as co-authors we can take the initiative on or we partner with someone to write something. So there could be a lot of different formats so that people find it not only important to contribute to HDSR, but also engaging and also from a process perspective, easy to contribute. Miguel, you actually give a lecture on the paper you work with, Tom. And so I would like to follow up a bit with you on that. Can you tell or listen a little bit about that experience? That's obviously is the form of engaged industry and the business, you know, directly. What are the lessons you have learned there? What are the things you can bring back to the column? For example, your column article becomes lecture material, which is terrific. Is there some way you think about it goes both ways and it's all about enhancing the column? Yeah. Part of the Harvard Data Science Review and Initiative online courses, we have some courses. The one Shelley is referring to where I give a lecture on the future work is called agentic AI. And then I help out with a couple of others leadership and AI and some new ones we coming out with And in that lecture basically idea is like what the future of work in general What the future of industries of insurance and banking And I think one of the core propositions was AI is a general purpose technology, a GPT, a term that economists have been using for a long time to describe technologies like the steam engine, like electricity, like Wi-Fi, the personal computer. and similar to other GPTs, they will transform, AI will transform society, the world. We don't know the pace. It's probably going to be more accelerated. We don't know the distribution of the transformation. There's a lot that will remain the same, but the truth is the world will change. It is already changing. A lot of people use AI in their daily work and it transformed at least some tasks by that. So we explore that in the course. And I think one of the most interesting things is that these courses are for practitioners. There are people that are not AI experts, not technical people. There are functional leaders. There are people in the business, marketing, finance, product managers, supply chain people. And the beauty of this is that they provide data and perspectives, which there's a term that's like the, it sounds like a cliche, but it's true. Like whenever one is lecturing or giving a talk or sharing some knowledge about something. One learns a ton from the students. These are very experienced people that are facing realities and complexities and can help put in some additional, a lot of information that then comes back and reinforces and improves the content of the course. And I think the interesting thing about the course is that it's designed with AI. And it really is trying to transform the way education is provided with AI. And honestly, this is a feedback loop that just helps really improve the content. As you're growing this and as you're moving through thinking about what you want to write about, is there a company or industry or type of data leader that you haven't featured yet, but that you're actively trying to get for the column? Yeah, it's a great question. I mean, More recently, I've been focusing applications of AI in education, right? If you think of education as a sector, of course, HDSR covers education in a variety of ways. But when I look at the applications of AI in education, it's so broad, the obvious applications in teaching and learning, but also educational institutions are organizations. Some of them are quite large, and they share a lot of the challenges that other industries are going through because they have operations departments. They have marketing departments. They have large physical facilities. So one really interesting thing might be to think of education also as a sector, as an industry, and maybe start talking about those challenges. And there are a lot of really passionate and experienced, qualified people to be talking about that, because it would be interesting to see maybe some of the common elements we might not be thinking about, since we think of education as different from our financial services or from a retailer. That might be one interesting angle to cover and it would kind of fits the different types of things I've been focusing on over the last few years. Yeah, I would complement that with, I think there's many domains that are not business domains that business people would benefit tremendously from. So I'm thinking like the arts, music, philosophy, so many scientific, physical spaces. I think there tends to be silos in human knowledge and understanding. One of the things I've been thinking is how do we bring in people that have nothing to do with business, but they are basically struggling with a lot of the same things that business people struggle with, organizational questions, leadership questions, data questions. And I think that one today, AI really makes that even more relevant. And one of the beauties about AI is that AI is so good at solving certain problems or giving advice on certain things because it extrapolates from other sectors and so people within one domain supply chain or even like i don't know like uh food and beverage will maybe find with ai a solution or a mental model or approach that they're not familiar with in their industry because ai is gotten so much data from different domains so i think bringing in that would be great um the other one is I think the article has done a good job of like gearing towards like leaders that are pushing and really leading in the AI and data science space and readers who are interested in that, both technical, non-technical. I think with AI, there's a phrase that I don't know who coined and like, and this has been true with big data, with data science, with a lot of different technologies or innovation, CEO and the C-level, they become the chief AI officer. So honestly, this is a push from the top, from the board, from the CEOs. I think having some board members, having some CEOs, having more C-levels write and contribute, I think will be super important. And I'll just throw in a plug there. Hamid and I are relaunching a lot of new things. So keep tuned because a lot of really interesting things are coming in the column. I truly love the vision you both have. I think there is something you're saying here is essentially to turn this term, active industry learning, your column term into a real action piece, right? You basically say there are things we can learn from industry, how they deal with all these organizational complications, human team works, all these complications, that there are a lot to be learned there. So I would actually really love to see how your column is going to venture into that dimension goes beyond the traditional notion of industry and the business really provide maybe a very distinct feature to separate yourself from others. So how do you do that in a sense, for example, if you're going to now pitch potential authors to get them to write, how would you go about it to really start to realize that grand vision there? There's two ways that come to mind. One is identify those common problems that an artist or an ex, whatever ex is, that's traditionally considered outside of the typical business space and industry space. But that overlaps and is similar. So you basically identify those typologies of problems. And I think one can try to depict it and try to have the person that's outside of business describe the problem and with some guidance come up with something that becomes very educational for people outside of that space, but applicable to and with lessons for the enterprise, for the corporates, etc. The second way, which probably is more effective, is you pair an enterprise industry person with a non-industry person. And you have them work collaboratively on an article, on a piece, on a case study, where they are getting to understand each other's mental models, understand each other's context. And that allows them to identify the common points and the non points which are very important I think because when they describe the problem and how they approached it they will also understand the limitations of solutions or the considerations when one tackles it So I think that those are two things that come to mind Hamid what do you think Yeah no those sound great. I mean, maybe to add from a more mechanical perspective, I mean, this is also a push-pull model, right? Actively seeking out diversity in terms of the types of people, industries, areas of application. I mean, you mentioned music, for example. I very recently met a professor from Berkeley College of Music in Boston, they actually had a full day conference working session on the impact of AI in the music business slash education, right? So if you're a producer, or if you're a performer, right, I mean, AI might be impacting you in different ways. I think engaging with different communities, different groups of, you know, people, different institutions, so we can more proactively seek those types of contributions. And the other is, you know, obviously this podcast, One of the goals is hopefully to get the word out. If we can create that kind of awareness, if people know that there are a variety of ways to engage and to contribute, hopefully we'll be receiving also submissions from the different parts of an organization in addition to the more traditional data science, chief data officer, chief AI officer, personas, or their departments. Well, I think we would have 8,000 more questions, but we are going to begin the beginning of the end and wrap up with some rapid fire questions. First, if you had to describe the ideal active industrial learning article to a prospective contributor in two sentences, what would you tell them? We could provide an opportunity for people who are in industry to do two things simultaneously, which might not be very easy to do elsewhere. One is to take a high-level view as to why they're doing what they're doing and what the impact is, but also nerd out a little bit so that we can get into the types of tools and technologies and how they're actually implementing those types of things. Balancing those out in a way so that it's interesting to a variety of people is something that we can uniquely provide. Okay. Favorite interview you've published so far? Todd James. Pomet? Yeah, and I'll pick Ashok. Most overused buzzword in industry data science right now? Agents. You have to agree with that. So, yes. Agents. Dashboards or narratives? I mean, we're all human beings. I think narratives always win, especially if you're trying to drive a decision. That's the narratives. One skill every data science leader underrates. Narratives. Hamid? Fully agree. Fully agree. I mean, a lot of people are now realizing that those types of skills are exactly what's needed in the AI era. Best piece of career advice you've gotten from a guest. Or, I guess, in general. When I was in consulting, the CEO of the company that I was with, he said, if you're coming to work or whatever you're doing and you don't have some butterflies in your stomach, that means that you're not learning, you're not challenging yourself. Not too much, but some amount of butterflies is a good thing. Life is short. Work with people you like that challenge you and make you better. These are all great advice. Now we come to the time where we always end our podcast with a magical one question. So I hope you're ready. This will be for both of you. If you could wave a magical one and get one type of article, data sets, or interview subject for the column, no constraints, budget or access issues, what would it be and why? You know, given data science and AI has now become such global, even geopolitical issues, having a group of guests who can illuminate us maybe or share their wisdom as to how we can collectively win as people, broadly regardless of where we are and what profession we're in engaged in, et cetera. What's that vision of AI that we need to actively seek and maybe some realistic ways to try and get there over the next several years, maybe a couple of decades. I think that would be great. For me, that would be super helpful because it would help me maybe understand this transition a bit better. Magil? I would love to have a data set of cognitive and emotional states of executives so that we can understand, correlate decision-making. I would love to have that for the entire world so we can better understand each other and have like an empathizer algorithm that interprets that and connects us with those external stimuli. Because I think we humans don't understand each other. And many times we make mistakes or make assumptions that lead to suboptimality and we don't get along because there's so much hidden data. So cognitive and emotional states would be my magic one data set. Well, thank you to both of you. And I think your answers are essentially point to the same direction. Both of you mentioned being global. This is truly very inspiring. What you're saying is that what we want is to have these technologies really help us to understand each other. I again appreciate your really grand vision here I want to thank you both of you again but also encourage all the listeners there that are trying to contribute to this particular column there are many other columns that have the science review there are 10 of them with a variety of different topics some are facing industry, other government last time we featured the recreation randomness just for the general public But whatever it is, please share your wisdom, your experience through these articles. You can take a look at Harvard Data Science Review's website and see how you can contribute. Thank you. That's a wrap on this month's conversation. Huge thanks to Hamid and Miguel for letting us into the editorial room of Active Industrial Learning. It's clear this column exists because the two of you keep chasing the question of what actually works inside real organizations, not just what sounds good in theory. If today's conversation made you curious, go check out the column on the HDSR site. There are links in the show notes. Start with the Todd James and the Ashok Srastava interviews if you haven't already. And if you're a practitioner sitting on a story about building a data team, deploying a model or fixing something that broke, Hamid and Miguel want to hear from you. Thanks for listening to the Harvard Data Science Review. Everything data science and data science for everyone. We'll see you next time.
