Adel Nehme: Hello, everyone. Welcome to DataFramed. I'm Adel, Data Evangelist and Educator at Datacamp. And if you're new here, DataFramed is a weekly podcast in which we explore how individuals and organizations can succeed with data and AI. Today is the first episode in our Data and AI Literacy Month special series of episodes.

In case you missed our announcement, Every year on September 8th, we celebrate International Literacy Day. Given the rise of data and AI into the mainstream, it's only natural we dedicate all of September to discuss data and AI literacy. So throughout the month of September, we are hosting a bunch of events all related to data and AI literacy.

Whether you are a practitioner or a decision maker, part of the data team or the learning team, There's something here for everyone. Moreover, we're capping the month with an awesome day of events during a special edition of Datacamp Radar, our signature conference, where we'll hear from data and AI leaders at McKinsey, AWS, New York Life Insurance, and a lot more.

Now, of course, we also have an awesome set of guests and topics for you throughout September on DataFramed, so make sure to tune in for this month's episodes. Now as we kick off Data AI Literacy Month, there's no better place to start than discuss the really urgent need to build AI literacy within organizations and individuals.

For the past few years, we've been beating the drum about the importance of data literacy and why organizations need to invest in a data driven culture. Mindset and skillset. However, as generative AI quickly rose to prominence and have amassed large amounts of users at every major organization today, the concept of AI literacy has never been more important.

So where does AI literacy fit into the equation? Is it an extension of the illiteracy, a complement, or a new paradigm altogether? How should you AI literacy ambitions? There's arguably no better person to speak around this than Cindy Howson, Chief Data Strategy Officer at ThoughtSpot and host of the Data Chief podcast.

As Chief Data Strategy Officer at ThoughtSpot, she advises top clients on data strategy, best practices to become data driven, and speaks internationally on top trends such as AI ethics, data literacy, generative AI, and a lot more. Cindy was previously a Gartner Research Vice President as the lead author for the data analytics maturity model, Analytics MBI Magic Quadrant, and a prolific thinker and writer in this space.

Throughout the episode, we discussed how generative AI accelerates an organization's data literacy, how leaders can think beyond data literacy and start thinking about AI literacy, the importance of responsible use of AI, how to best communicate about the value of AI within the organization, what generative AI means for the data team, and a lot more.

If you enjoyed this episode, make sure to let us know in the comments, on social, or wherever you listen to your podcasts. And now, on to today's episode. Cindy Howson, welcome DataFramed.

Cindi Howson: Thank you for having me, Adel. Good to always talk with you.

Adel Nehme: Likewise, so it's great to have you on for Data and AI Literacy Month. You are the Chief Data Strategy Officer at ThoughtSpot. In a previous life, you were also a VP at Gartner. So maybe putting your Gartner hat on I'd love from your perspective to learn, you know, where are we on the generative AI hype cycle today?

And where do you think we're headed in the next year?

Cindi Howson: Yeah. So if I were still a Gartner I would probably put us at the innovation trigger phase. The one thing I don't like about that phase though, is that Gartner describes it as really not having real production use cases. Whereas I think there are many, many, many production use cases already.

So I, think we're a little past that, but if we think about as an industry, there is still a lot of work to do to figure out how to trust these models when not to trust them. Getting consistency across both the tech companies and the data sets being used.

Adel Nehme: That's great. And definitely want to unpack a lot of that with you on use cases, where we are today, what's possible, what's not possible, but maybe also focusing on, the generative AI innovation cycle. you know, Looking back, you've been in the industry for quite a long time. Looking back on how AI has evolved in the past decade.

How has this trajectory surprised you, if at all, in 2022, 2023?

Cindi Howson: Yeah. So GPT from a ThoughtSpot perspective, of course, the platform started with natural language processing. And you could even say that we had our own. internal large language model that we were using. So there, if you look at the earlier GPT models, so let's just focus on that from the beginning, they were not particularly accurate and that's why nobody was really talking about them.

It was more in academic circles and it was really the release of GPT 3. 5 back in the fall that showed a greater degree of accuracy. and what was the trigger for that? Open AI getting funding from Microsoft, running these large language models, so having access to the additional compute in the cloud.

And that is what really took everyone by surprise because they've been out there for a while, but it was in the fall, the degree of accuracy and the chat interface also, I think, allowed the public to actually look at these things. In ways that wasn't possible before.

Adel Nehme: And in a lot of ways, Chachapiti is really the first mainstream interaction that we have with AI at such a large scale. We've always had AI in different tools. Think from Google Translate, but it's really Operates invisibly, right? Where kind of puts an AI and fingertip that makes it so effective.

Cindi Howson: It does. and I also think AI. So some people have gotten afraid of AI, like, is this going to destroy humanity or make it better? And to your point, AI as an umbrella term, if we parse that out, look at even on your phone, if you want to look up a picture of a cat, a dog, a person, that is AI. So these machine learning models.

I think we have to differentiate between what's happening now with generative AI and AI has been used in so many use cases for years now. So the generative AI is the part that I think has exploded and where there's a lot of potential. But a lot of things that we need to pay attention to as a society and as an industry.

Adel Nehme: You're such a prolific thinker in this space. It's pretty challenging actually, to find a set of topics to discuss in preparation for today's podcast. You mentioned here on the culture side, for example, of, there's a lot of fear on AI and displacement, for example. I wanna really cover that cultural aspect with you.

Especially for data and AI Literacy month, a lot of organizations now, a lot of leaders, We're trying to think about how to address the culture and skills component of the AI revolution, of generative AI revolution. But maybe before we go into that as well, maybe trying to set the stage a bit more and try to understand the lay of the land.

Maybe try to first outline what is possible and impossible to do with AI. I'd love to learn from you now. Where do you think AI can be used today within an enterprise setting? And where do you think we're still really far off from usage?

Cindi Howson: let's just focus on data and analytics and then drill down on the data literacy side because I actually think generative AI in particular will help improve data literacy. So if you think the data and analytics space, the whole workflow process From capturing your data, to modeling your data, to uncovering patterns, to actually describing the visualizations, generative AI will impact that entire workflow.

And for your data and analytics professionals who are listening, some I think are like, okay, this is cool. But they have not yet fully imagined how it will dramatically change their jobs for the better, but only if they reskill and understand this. And that's why we like partnering with DataCamp because you help with the education side.

But if you take data modeling, for example, we have a lot of manual data modeling processes that are really based on physical representations in a source system or in a data platform. And so being able to present a model that starts more from the business side and uses business terms, I think it help with that and then including things like metadata, descriptions.

I don't know, who likes documentation? I feel like I should say I like it given I've written so many books, but does anyone really like it?

Adel Nehme: Yeah, don't think so.

Cindi Howson: You don't think so? I don't know. I like writing. I like writing. Okay, well, for the people that hate writing and documenting, this documentation is going to really be automated and integrated. In that data modeling process and, I find that super exciting. Even the data creation side, if you think about maybe how, let's take a CPG company and the way you engage on a website, the ability to have a more intelligent bot and capturing those conversations.

So this is bringing in the semi structured data and even images. Increasingly possible in a generative AI world. So now if we go to the insight side of things, so if you think about Adele, there's been a mantra in the data and analytics space for decades, mainstream BI, BI for everyone. generative AI.

We'll actually just make data part of every worker and every citizen's interactions. The same way we never think twice now about looking up a Google map we use those intuitively and it's infused in when you order an Uber, when you order your food. Data has been locked away because we've needed experts.

To be able to answer questions and be that middle person with generative AI, we're going to be able to take the expert out of the core middle, and people will just be able to ask questions. What were my sales last month? And maybe not even have to ask. the question, maybe just be alerted to say, you have an unusual pattern.

It's the summer of Barbie. Pink everything is on fire and there's an outage in this particular store, a stockout. So let's replenish there. And AI will be able to actively tell you that. And for maybe a let's say a college grad, fresh out in the workforce. They're not so good at interpreting anomalies or outliers.

Generative AI will actually be able to describe what this alert is telling them and what the next best action is taken. So that's the whole workflow. Any questions on that?

Adel Nehme: Yeah, that's wonderful. There's a lot to unpack and what I love what you mentioned is you're painting this picture of how generative AI can really accelerate data literacy within the organization. And we always used to think of data literacy as not just a skills transformation, but also.

a change management, a cultural transformation as well. When you add that AI accelerant on the data literacy program, how do you think organizational culture needs to adapt as AI becomes more ubiquitous? How do you think AI will challenge organizational culture? Yeah, so we know in general, creating a data culture is a huge part of people change management and it needs executives on board as well. And if you think before we could survive on an experience based culture because there were mainly an analog interactions. So I could see to go back to my retail sales.

Cindi Howson: I could see the customer walking into the store and I could see that they're hanging out more in the brightly colored areas or what have you. Now, when those interactions are digital, we lose that visibility. And this is where we have more data to give us the visibility, but we need the skills to be able to interpret that.

And it has to start with a common language. So a mistake that many organizations make, is that they confuse technical literacy with data literacy. Technical literacy is learning a particular platform, whether it's data science or analytics. and really ThoughtSpot, our view is let's flip the requirements for the degree of technical literacy for every business user to be able to ask questions.

Right now we spend 80% of our time on the technology, 20% on the data. And we want to flip it and have it 20% make the technology super, super easy, as easy as a consumer grade app. And then that way you can focus more on what are the business definitions. In a digital world, even the definition of customer, there's not a consistent definition of customer.

A salesperson will define a customer as it could be a prospect, somebody who went to your website. But then in accounting, somebody may say a customer is only somebody who actually transacted and bought. So that, that data literacy has to start with a common language, agreeing on a common language.

Adel Nehme: that's really great. And, you know, you mentioned something as well, like in a previous discussion we've had on a DataCamp webinar, You mentioned something very poignant that I still think about to this

Adel Nehme: day, which is the importance of not using data to punish, right, when communicating about the importance of data literacy and that psychological safety.

I think a palpable parallel that we see now with AI becoming more ubiquitous is, as you mentioned, the A. I. Taking over a job, the fear of displacement, maybe when you talk to leaders and try to advise them on how to approach that communication and kind of that change management program and trying to integrate an A.

I. A. I. Into the work processes. What is the message that you think leaders should be leading with when talking about A. I.

Cindi Howson: They have to tie it to the business strategy and business value. If they are approaching it from purely a technical. Approach, it's not going to succeed. So if I think about healthcare, I'm going to switch industries, healthcare, it's, we are going to use AI to improve patient outcomes and lower the cost to care for those patients.

They have to tie it to that. Otherwise. Employees often think of this as just a shiny new toy and the latest fad. And if I just stick my head in the sand, I can wait until this fad passes away. and there's that resistance to change.

Adel Nehme: And how do you approach that resistance to change What are good patterns of communication here that you've seen leaders succeed with, you know, outside of communicating to the business value, there is like, how do you address the fear component maybe of using AI?

Cindi Howson: So we have this model of best practices and it starts with the why, and then there's, four tactics around the why. So the why, I talk about with them. What is in it for me? And leaders need to understand that there will be a different why or with them for every level in the organization.

So for some, it might be for the data analyst, let's say, I just want to, I like learning new things that might be there with them, or they might. Be fed up with the drudgery. Somebody described themselves to me as a dashboard monkey, and that's not what they went to school to get a business analytics degree in.

So speaking to their why from both the head and the heart is super important. And then the four tactics, communication and communicate using consistency and different media. For some, it will be this podcast. There's one customer I work with, they actually use TikTok as part of their communication strategy internally.

The second thing, super, super important incentives, and I wanna come back to the incentives because you also mentioned this fear using data to punish because we have something playing out in the industry now, but incentives can be a career path, a promotion, a bonus, or even just badging. Related to the badging skills we have to give people time to re-skill.

And there's so many good learning platforms out there to learn the new way of working, to understand what generative AI is. But leaders are not creating the time for their employees to upskill and re-skill. So I think of another organization I work with, they have something called Focus Fridays. And every person on their data team, they are given four hours a week.

Focused varieties are expected to reskill. The fourth element is tribes and role models. And what is difficult here is whenever we have a new innovation trigger, such as generative AI, is who are the role models? And I think we have to go back and look. In our industry, in our society, the last major innovation trigger was the internet and how it changed the way we work, the way we did things, the way we deploy technology, the way we shopped, the way we even got patient care.

And so we do not want to be the blockbusters. We want to be the Netflix's of the world. And so I think this is where the role models are changing so, so quickly, but organizations really want to surround themselves with other people that are embracing AI and framing it from a business viewpoint and communicating well.

If you're hanging out with the organizations that are taking a wait and see approach. This is where I think the pace of change, we will see more destruction and more businesses caught short footed than ever before.

Adel Nehme: This is brilliant. And, one thing that you mentioned here is the rescaling component, right? Something, of course, Datacamp, we're Very invested in and, we've definitely worked with a lot of organizations on data literacy programs. But one thing that we're starting to see now as, tools like chat GPT or generative AI not only are being used within the data team, but within the marketing function within the HR function, everyone needs to learn how to use a generative AI tool somehow.

We're seeing more of this concept of AI literacy emerging. So how would you maybe define AI literacy? Do you see it as an extension of data literacy, a complement? do these terms intersect or diverge?

Cindi Howson: I do see it as an extension and also, so Adele, if I can also confirm, we use the, some people use the term chat GPT interchangeably with GPT. And ChatGPT is very different from GPT. So GPT is the API that is getting infused in so many products. The chat part is just another way of interacting with it.

So, AI literacy is understanding what is the training data set, and then how is that being used to generate Either the output from GPT or from an AI generated insight. So even if you think we know generated AI has a problem with hallucinations, understanding what a hallucination is and why that may happen is part of AI literacy.

Data literacy is related to that. And so I'll give you maybe a funny example. A couple months ago when I was first playing with ChatGPT back in December or January, I asked Who is Cindy Housen? And it comes back, I think it had me as CEO of ThoughtSpot, which I had to chuckle with that.

that's, that was totally false, but it also had me, here was the thing, it had me as the author of big data for dummies, something like that. so understanding AI literacy then is why did it come out with that totally factually incorrect information? And it's really just putting together words.

Based on predictions, what is the most likely next word? The series of books, anything related to dummies is way more popular than the particular books that I wrote. So it, just took popularity and topics that I write about and stuck them together. And that's really where a hallucination comes from is just getting the prediction wrong.

So AI literacy goes to what was the data trained on? What biases does that introduce? In the machine learning algorithms, and then being able to interpret the AI results. So we, we would always say. AI plus human has to go together. risk with poor AI literacy is when somebody takes an AI generated result as fact and a hundred percent, nothing is a hundred percent with AI because all data is biased.

Adel Nehme: And you mentioned here, like, on the importance of understanding bias and responsible AI concepts is so important. Also another great collaborator of ours, Valerie Logan, right, was on DataFrame, talked about how we're Almost three years ahead in a 10 year cycle when it comes to data literacy. And I think we're starting to seeing a bit of maturity in the data literacy conversation where you have within an organization dedicated data literacy leads, for example, under the chief data officer, where we see more and more mature data literacy programs within an organization.

Do you think that a similar effort is needed on AI literacy, or do you think we'll be able to cover the gaps more quickly? Maybe to reframe the question, what do you think a successful AI literacy program looks like?

Cindi Howson: So of course, love Valerie Logan, the data lodge work both when we were at Gartner and, and we partner today as well. And I do think we are what I am pleased to see. Is that data literacy or data fluency, as I like to call it, is now a boardroom conversation that executive teams are saying, how do we create a more data fluent workforce?

So I think we definitely are making progress on the maturity front, the data literacy and AI literacy needs to work hand in hand, because again, if, if you don't have that common language. And now you're interpreting things, maybe an AI generated insight says the propensity for this customer to churn is zero and yet, or low 20%.

And then they churn and you're like, wait a minute, that's not what AI told me. and that's because you didn't know how to interpret a propensity score.

Adel Nehme: That's really great. I love that example that you mentioned. And, I think this also connects back to your earlier point on how AI can really accelerate data literacy. Right? So I'd like to kind of return back and really park on that particular topic here. Maybe because you've been so involved with this at ThoughtSpot, maybe walk us through some of the AI use cases that you've seen in the data space particularly that have really excited you and that make you very hopeful.

I'm excited for the future of accelerated data insights and data literacy.

Cindi Howson: so if I can talk about ThoughtSpot Sage, we use generative AI. In multiple ways, so one to make data more approachable is that nobody likes to start with a blank screen. And so actually giving people AI generated questions based on the data and the popularity of other questions asked just helps with that approachability.

The other thing is that we use it as a way of asking questions of your data in natural language, and it's on top of our patented search. So that we have always human in the loop and greater accuracy. The third way is explaining what you're actually seeing. So giving descriptions about what that chart on the page is telling you.

Or the forecast or the anomaly. And then the fourth way is in app support. So, actually working with an AI assisted tech support person. If you're asking a question, How do I create this chart or add this calculation? Then it's through support. So, there's four different ways we use it in our platform.

When you think about how we use it in terms of data literacy there's actually an interesting finding. from our program manager is that because there's a lower technical risk of asking questions, people actually are asking questions more. So we did a before and after test. People that are using ThoughtSpot Sage are asking more questions than people who are using traditional ThoughtSpot Search.

And so I think what's happening here, Adele, is that. They are able to understand their data better. So saying something like, okay, what, what are my customers? And I get a huge number and it's, oh, wait, that included prospects. And so then I can ask another question and another question. And so it's allowing that iteration.

The other aspect is these descriptions. So business definitions. It's bringing in business definitions and we have these knowledge cards so that people in context can actually see what is that definition. How do you define customer or how do you calculate day sales outstanding things like that. And, I like to comment on one of our partners.

Atlin, for example, is using generative AI to create that business definition and then we surface it inside our knowledge cards. So I, I think all of these capabilities will have the opportunity to have something more like a data Sherpa that will let you say, explain this data to me.

Why is this performing that way? And it, just will be a, much more human like interface.

Adel Nehme: What I love here is just really how much extra context you get when interacting with data. You mentioned the business definitions. I've been on, a lot of meetings where people are like, Can you define what that metric is? My creating that shared business definition, it's such an important way to smoothen the conversation with data.

Now, while we discuss here the impact of generative AI on data, science as well as, data literacy in general. In many ways, it seems that generative AI can help scale self service analytics and reduce the strain on a central data analytics team, right? for one, these types of tools will really change how we think about what it means to be data literate for the wider organization.

But they will also change the makeup of a strong data team. So maybe starting off with the latter, how do you think the skills of the data teams or the skills data leaders should be focusing on today when building out their team, how will they be impacted by the rise of generative AI tools?

Cindi Howson: Yeah. So I think that's where it enables the central data team to focus more on the harder problems, which is really about provisioning the data, governing the data, ensuring data quality. And it will allow the decentralized teams to create data products more quickly, but still in a trusted way. So that shifts the expertise away from having to be highly technical.

And that steep learning curve of the domains. So taking the domains and letting them use low code tools is better on this trusted governed data and with information about. The degree that it's, quality data. So I think the centralized data team will focus more on evangelizing, provisioning, upskilling.

So upskilling on data storytelling. Being also the organization that facilitates data fluency upleveling and then, we've really been stuck at diagnostic and descriptive analytics. This will free the data team to say, how do we double down on the harder things, predictive and prescriptive analytics that there just is never bandwidth to get to,

Adel Nehme: I couldn't agree more and maybe focusing here on one particular role, because the data analyst or the business analyst within the organization is really focused on that diagnostic and descriptive analytics side of the equation. How do you think that particular role will evolve as, Charity of AI is able to take some of the load off of that particular rolls back.

Cindi Howson: yeah, I think they're going to be able to do more with less. Thanks. And they'll be able to hone, focus on those skills that have been lacking like data storytelling skills. So when you're so bogged down on just, Oh my God, get me this data. Where is this data? How do I create this dashboard and report?

Actually getting to what is the data trying to tell me and how can I best communicate this? There almost isn't the headspace or brain space. To do that. So again, I think we're going to be able to flip the equation from gathering data, creating dashboards. To really, what are the insights and what is the data trying to tell me?

Adel Nehme: And Cindy, as we close up our conversation, I'd love to learn from you, you know, how you see the space is headed. So maybe just focusing on the data space and intersection of data science with AI. Where do you think we'll be in six months, 12 months when it comes to generative AI impacting this data tooling stack and accelerating data literacy?

Cindi Howson: So if I look six to 12 months, that's like even a short timeframe, but this is where there's only a handful of tools that I would say are in production already with generative AI, but we will see more across that whole life cycle. And then we're going to see the mistakes. And so that's where people will, come back and learn.

And maybe that's when it moves into peak of inflated expectations or whatever that cycle is. but I, I think this is where we're also going to see shifts in

the jobs for those early adopters. and that's a good thing, that will be a good thing. And I, I already think if, if I'm allowed to mention at least two customers that have said how generative AI has helped them with the data fluency, I look at Chick fil A. You don't think of a restaurant manager, for example, as being your traditional data person, and yet data is part of their jobs, or Pepsi has said.

How ThoughtSpot Sage has enabled greater data fluency. So it's this shift in who we think of as a data person that will just continue to further accelerate. also, I think what's really important, we've talked about GPT. The other thing is the other large language models. And this includes industry specific.

Large language models. So Bloomberg, for example, has a financial services, large language model. Truvetta has a healthcare large language model because we're not paying enough attention to the cost differences of there's GPT, there's Palm, there's others. But we need to pay attention to not everyone needs training on such a large corpus of data.

And sometimes having smaller models will bring more efficient compute, but also greater accuracy. So I think that's going to be another big thing, the battle of the LLMs.

Adel Nehme: Yeah, the battle of the LLMs is definitely something to keep an eye out for. And you mentioned something here, Cindy, on, you know, how jobs will be transformed for early adopters. as we close out our conversation, I'd love to also like understand from your perspective here, given that you've seen different tectonic changes in technology over time.

How do you view the conversation on automation of roles versus augmentation of roles? Do you think that AI will be able to displace many roles in the future? Or do how do you view that particular risk?

Cindi Howson: So to me, the question is, is it a net job creator and then how does it affect me personally to your listeners? And it will affect every role, but it will be a net job creator. and we can go back in time, taking an example of a bank teller, we have fewer bank tellers because we have mobile banking and we have ATMs.

And so it reduced the number, but it changed the job of now what a bank teller does more a trusted financial advisor. And so this is where sometimes want people to freak out. Because I want them to act and start reinvesting in themselves today. So, will the role of a data analyst job go away? It will be changed and really everyone will be a data person.

Every citizen, every worker will be a data person. How you interact with that data. And the degree that you're authoring dashboards, well, we already say dashboards are dead. That is what will change. And if you are unprepared and unwilling to look at what is the composition of my job, you will be left behind.

Adel Nehme: That is a great way to end today's episode. And maybe one final question, Cindy, what is your call to action to listeners today as we AI Literacy Month? And as they get started thinking about how to future proof their careers.

Cindi Howson: Be a continuous learner and also educate those around you. So I, last week, was doing trainings with my family in ages from 12 to 91. On what is generative AI and how it matters in their daily lives and how it matters in their work lives. You listeners are the data educators and evangelists as well as the data practitioners.

Adel Nehme: That is awesome. Cindy Howson, it was great having you on DataFramed.

Cindi Howson: Thank you for having me, Adele. Always a pleasure.