For true enterprise AI readiness, organizations — especially now in the era of Generative AI — need to advocate for organization-wide data literacy. According to Harvard Business Review, data literacy is “the ability to parse and organize complex data, interpret and summarize information, develop predictions, or appreciate the ethical implications of algorithms. Like math, it can be learned in beginner and advanced modes, spans multiple disciplines, and is often more practical than academic.”
Recently, our very own Claire Gubian, global vice president of business transformation, was on the Data Humanized podcast for a great conversation around how exactly Dataiku champions data literacy for analytics and AI transformation. Feel free to watch or listen to the episode here, but if that’s not your vibe, feel free to read our favorite parts in the truncated transcription below!
Wait, Why Does Data Literacy Matter?
Mark Palmer: How is the workforce reacting to the accelerated use of AI across all industries and how can leaders use data literacy to mitigate those fears and concerns while using AI to drive digital transformation? In this episode of Data Humanized, we'll explore these questions with Claire Gubian, the global vice president of business transformation at Dataiku.
In each episode, we bring you the unique perspective from enterprise leaders at the intersection of technology in humanity who are leading cultural transformation through the power of data. You'll also hear the real-life stories of learners who have graduated from the data science for all program and are embarking on new career pathways, creating a more inclusive collaborative and effective workforce. I'm your host Mark Palmer, please visit the Correlation One website for more about how data literacy transforms enterprises and tell your friends about the Data Humanized podcast.
So why is data literacy important? AI is projected to replace 85 million jobs by 2025 but create 97 million jobs at the same time. But we have a massive skills gap — over three quarters of workers don't feel ready to thrive in the digital-first world. The problem is going to continue to grow: Also by 2025, 70% of us will work heavily with data so we have some work to do. Data literacy is the key to closing the skills gap, but only 15% of leaders say they're ready to transform their workforce. So today we're going to learn from Claire Gubian about how she's helping organizations accelerate transformation through data literacy, upskilling, and best practices.
Prioritizing AI-Driven Cultural Change
Mark Palmer: I’m here with Claire Gubian for the Data Humanized podcast! Claire is from Dataiku, one of my favorite named companies in the data science space. She is the global vice president of digital transformation at Dataiku and a lot of people listening to this will probably know of them — they're one of the best data science platforms in the industry, called a leader by all sorts of analysts, and a fantastic tool. I know a lot of people that have a lot of affection for it, so I'm really excited to have you here Claire to talk to us on Data Humanized, welcome.
Claire Gubian: Thank you Mark, I'm very excited to be here as well.
Mark Palmer: So, we have to start with the name of the company which I'm told by my friends at Dataiku is a combination of data and haiku, the form of poetry, which I think is one of the most elegant names of a company I've heard. I was curious if you had a Dataiku haiku to share with us to start off.
Claire Gubian: Yes I do Mark. Before I share it with you, it was actually the tradition a few years ago that every new joiner would have to compose a haiku. I actually asked our little friend ChatGPT for some assistance and I thought let's do a Dataiku haiku with digital transformation and this is what it did:
Dataiku guides
Change, business transforms, blossoms
AI steers new course
I think it's pretty spot on with this podcast, right?
Mark Palmer: Ah it's dangerously good, in fact that’s how I thought we'd jump in, so that's excellent. Kudos for bringing ChatGPT into it — it is the intersection of data and humanity, that's why we call it the Data Humanized podcast. Your role is really interesting because your job isn't to build the tools, it’s to help your customers go through their digital transformations using data science. Maybe you could start off talking about your role there and I know one of your passions is about the best practices associated, so maybe you can take us through what you see as some of the best practices you've seen people use Dataiku to solve and go from there.
Claire Gubian: Indeed, I lead the business transformation team at Dataiku. We are a team of about 15 senior advisors and we all have a background in consulting and data & analytics. What we like to tell our customers is that they've invested in a great technology which is Dataiku — they are embarked on a digital transformation journey and it goes beyond the tech. It's about the people, about the process, about driving a data-driven culture. We work with over 500 customers across multiple geographies, multiple industries, and this gives us a very unique perspective on what works and what are the pitfalls also to avoid when you're driving transformation.
We've put together an Everyday AI blueprint in order to flesh out in a more comprehensive way how to go about it. There are two points that we like to highlight specifically where we see AI winners really invest. The first one is ensuring a stringent alignment between data science and the strategic priorities in the company — really thinking through what use cases you're investing in starting with a business problem. The second area where we see very successful companies invest mostly is change management and upskilling of employees — not just giving them a new tool or a new technology but really investing on how to change the way decisions are made, changing the way people work together, embracing data analytics and that doesn't stop at the data scientist or data analyst level. It is throughout the organization, so it incorporates both the subject matter experts and the data experts.
Mark Palmer: So well said. We talked about a recent survey from NewVantage Partners, Tom Davenport's group, that says that 90% of leaders identify culture or change as the biggest obstacle to doing better AI, however only 1.9% of them invest as their number one priority change in AI culture. How do you talk about some of those ways that you help clients get that alignment, I think that's one of the big challenges. What do you find with your clients that that tends to work, how do you get that alignment done?
Claire Gubian: The answer is in people, process, and tech. First of all the tech — investing in a tool like Dataiku which helps bring different level of code knowledge and domain expertise on one single platform to be able to collaborate and crack business problems. We really believe in the power of that, where the best projects are when you have the domain expert, so let's say a person from marketing, a person from supply chain, and the people who are the data experts who know where the datasets are, who know how to build models, are put together in one platform. So that's one piece, which is really breaking the silos between the teams.
Then from a process perspective, we really believe in the importance of spending time reviewing use cases, qualifying them, and prioritizing them against what value are they going to bring, what costs are they going to incur, and what are the risks attached to them. Then, the third piece is looking at how you drive culture, so let's spend a few seconds understanding what is behind the word culture. Culture is the system of belief, the common system of belief, that helps people best work together and make decisions. The word that strikes me as really important in this definition is beliefs. When we're talking about beliefs, we are in the domain of the irrational and looking at barriers to adoption of data analytics, one very concrete barrier to adoption is fear of losing my job, of not being relevant anymore.
That's the fear that we love the most working on. Why? Because we actually believe that it's the opposite: AI, data, and analytics actually help make people extraordinary. That's our tagline: Everyday AI, Extraordinary People. Everyday AI, what do we mean by that? AI can be applied to any business decision to make it better or any business process to make it faster and improve it and it can help make people extraordinary by giving them more capabilities and making them faster.
We love to talk about these examples, so one example for instance one of our big customers Standard Chartered Bank. They embarked on a transformation journey and they have equipped their financial analysts with Dataiku and thanks to that, they've been able to cut down the time that they were spending on P&L forecasting from 2,000 hours per month to about 70 hours. How have they been able to do that? By automating the manual tasks, not only have they been able to reduce the time they spend, but they are also improving the accuracy of their prediction because they're able to compute much more data in a quicker amount of time. We're not talking about sending people to the moon here, we're talking about what we call the mundane use cases, the everyday processes and where people can really augment their jobs thanks to data analytics.
So, how do you drive change? Well you drive change by showcasing these kinds of examples. Tangibly speaking, we're talking about spending less time on manual tasks, going faster, and being able to make better decisions.
Mark Palmer: Love it. One of my favorite examples with this fear is what David Arturo from MIT talked about in a TED Talk that he did where he compared the fear of automation to what happened with introducing automated teller machines about 20 plus years ago and everybody's said that's the end of tellers in the banks, that’s the end of people in the in the branch. Actually what happened is jobs increased and tellers increased but they just weren't doing the cash dispensing anymore, they were actually doing higher value tasks. They weren't doling out money, they were doing customer service, they were providing an elevated form of service which it sounds like your example with Standard Chartered is a great example of where AI is making their job better. Presumably, the net impact was that those people are doing better jobs and probably more interesting work, right?
Tracking Digital Transformation
Claire Gubian: Exactly, we hear that every day. So, the second thing that my team does is value engineering and we document how customers are using the Dataiku technology and how much value they're creating with it. Honestly, it's the best job ever because you hear about value creation every single day but also just amazing use cases. The number of times where I hear this technology is making me love my job again or more because it does remove the time spent on manual, repeat tasks to focus on what people love the most doing, which is creating or in the case for instance a banker spending more time building the relationship with their customer instead of researching.
Mark Palmer: That is a great part of the job, seeing that intersection with humanity. Of course, the only problem is you can't hang an ROI on loving my job better, can you? I mean you certainly can with the subsequent knock-on effects. I do think that's the irony of it is that the fear is that they'll be automated away. The reality is that you'll do better and actually, in a lot of cases, elevate your work. Let's talk about measurement, that was something that you brought up in your principles. In my background, I've worked with companies that measured the percentage of employees engaged (as you pointed out, you made a great point though not everybody has to be engaged with data science). So, what have you found there best practice wise about measurement methodology or even frameworks or how do you advise companies to look at that aspect of tracking what you’re up to with data?
Claire Gubian: A great question. I believe that tracking is the only way to measure progress on a transformation initiative, but what are the KPIs that you can use to track digital transformation? Asking customers and looking again at some very successful programs, being able to track: The example that comes to my mind is Mercado Libre who has defined a specific set of KPIs in terms of percentage of employees who are using the tools, how frequently do they use the tools, and then what value is being generated by the usage of these tools, so it's both from an adoption play and from a business outcome play.
Mark Palmer: One of my favorites there is the Scottish EPA, a customer I've worked with in the past (Environmental Protection Agency of Scotland) and with 1,200 employees, 80% were touching data all the time. Now that's the EPA though, because their job is to understand the country and the environment and that's a very heavy data job. The basic nature of the business is about data, but that's a similar methodology, so usage, who's using it, how often, and then of course the art of it is trying to measure the impact. I guess that's why you get a good value consultant and partners to help you through that.
When I first started getting into business intelligence, it was a new skill that I found myself reaching for just about every day, even just to track my daily routine. Then it becomes a tool that's just natural like breathing. Let’s talk Responsible AI — of course the blessing of AI is all the things that we've talked about, all the power you can get from it, but of course the other side of that is the responsibility of using governance. I've done a lot of research in the AI bias mitigation area based on work by Daniel Kahneman, a social scientist on how to mitigate risk.
Risk Mitigation and Responsible AI
Kahneman himself says since humans can't identify bias themselves and they program algorithms, it's really hard for them to spot the bias that they might place into algorithms or what they're exhibiting. He talked about the concept of decision observers (i.e., the best way to spot bias is to ask somebody else to identify it for you). I'm guessing that with your collaborative approach in what you do with your clients that that's in line with some of the things that you talk about, which is to get more people involved, cohort style. What are the best practices you've seen in the Dataiku universe for people doing more responsible AI?
Claire Gubian: Responsible AI is really built in the principles of our product. I think the one you alluded to which is collaboration of different stakeholders, balancing points of view, but also bringing information that everyone possesses, that is just one very clear way to echo what you were saying. The second piece is we've invested in developing specific features that enable us to detect bias in datasets, bias in models, and to be able to make decisions at least knowing what the bias is. We do a lot of evangelization on the topic as well to raise awareness about it and make sure it's at the forefront of decisions.
Mark Palmer: Within the tool or is it a procedural thing, heuristics or the algorithms detect potential bias in an algorithm? Is that something that's a a procedural thing where you flag and then broadcast, “Hey, we think there's some bias or we've detected something” because of course bias or discrimination are pretty straightforward, but some are very gray areas. How do you think about that and how do you coach your clients to deal with the grayness of this really super important area?
Claire Gubian: We do have specific product features such as bias detection that are built in, and then we do have a team within our business solutions department that owns notably all our governance principles and that's where we can actually do hands-on work with customers to help them better define governance principles and governance processes that can then also be integrated within the product, within the Govern console.
That console helps looking at a given project, so you can monitor all projects from the Govern console and identify what are the potential risks tied to these projects in a very detailed way.
Mark Palmer: I've always railed against the citizen data scientist phrase just because my crack on this is that nobody really wants a citizen doctor to operate on them, they wouldn't want a citizen data scientist to design algorithms. However, in this context it seems that this is really where the combination about getting more people sort of citizen data science aware and literate so that they can interpret. My other side of the joke about the doctors is you don't want a citizen any citizen operating on you, but everybody should be able to understand and interpret their chart or their vital signs, and their basic health — so I suppose that's where you get that artful intersection between something that's very technical and having more people be able to see where bias might be and be able to read their chart.
Claire Gubian: We don't really like using the citizen data scientist phrase either because it doesn't bring the idea that we're trying to convey, we prefer talking about accelerating the impact of data experts and upskilling domain experts. Then bringing these two profiles data experts and domain experts on a common platform to be able to get the right insights.
Mark Palmer: That's where the Data Humanized term came from for the podcast, this idea of data humanist, so there's intersection of the humanist, the subject matter expert, and data — the more data literate.
Key Takeaways
Mark Palmer: If you were to give people three things to walk away from on this topic what would they be?
Claire Gubian: If I start with a quote, you must have picked up on my accent that I am French and so I was very inspired by a French quote which is “Paris was not built in one day.” I like this quote especially now where everything just goes so fast and we talk about speed all the time and we are all about delivering AI quick wins, but also creating change takes time. It is a journey and understanding that and understanding that long-term investments are important.
For key takeaways, you know we touch upon these themes, but really the first one is don't think about data as just a data expert field, invest in data upskilling to bring your domain experts up to speed on these topics. It will just make the whole company better. The second one is a temptation to think that any data project is created equal but no, start with the business problem in mind and then spend the adequate time part qualifying and prioritizing use cases to make sure that you're working on the right ones that will actually deliver business value. Understand what is the business value tied to the data project.
Last but not least, we're talking about change management here and I believe in a very basic technique which is success calls for more success. Identify those flagship use cases that are going to create adhesion and understanding of what AI can do for your company and also inspire others to follow suit. When you communicate about this, avoid any sort of data science jargon — focus on the business language that can be understood by anyone because again, AI can be really applied to any parts of the business.