Episode 3: Behind the Buzz—How Starbucks Approaches AI

0:00:07.3 Shashank Garg: Hello, everyone, and welcome to the Intelligent Leader podcast. I’m your host, Shashank. And today we have a very, very exciting leader with us, Megan Brown from a very coveted, loved brand globally, Starbucks. I just got my coffee. Megan is the director for the Global Center of Excellence for Advanced Analytics and Data Science at Starbucks. She’s an analytics leader with about 20 years of experience applying all the scientific methods, statistics, machine learning, to inform decisions in a variety of fields. And I’m gonna let her talk about what she does today. They are curious and passionate about solving hard problems, making it easier to use data and analytics to influence business strategy and tactics. And they have earned their PhD in cognitive psychology. Is that right, Megan?

0:01:05.0 Megan Brown: Yeah, yeah.

0:01:06.3 Shashank Garg: From the University of Wisconsin. So let’s kick things off. First of all, thank you so much for joining in. And could you maybe just describe your role a little bit more. What are the stakeholders? What exactly do you do?

0:01:19.3 Megan Brown: So my role is the part of our data analytics and data science insights org that faces internationally. And so when we say international, it’s actually a very different business model from the rest of Starbucks. So, it’s more like a business-to-business consulting role where we take our most effective and incremental use cases and apply them to other markets. There’s a lot of variation between what we do in North America and what we do internationally because there’s so much difference between, say, customers, the partners who work in those markets, the data systems that we have in place, and also just the way of work in international is necessarily different ’cause you’re not in the same time zone. So everything’s a little wonky. Yeah.

0:02:10.0 Shashank Garg: Yeah, yeah.

0:02:11.0 Megan Brown: And then in terms of my personal life, I am a parent of three kids, they are six, almost five, he would want me to say that, and 10 months. I do some modern dance in my limited free time as a parent of three. And then I also write here and there, but not very well.

0:02:30.0 Shashank Garg: Awesome, awesome. Thanks for sharing. So you’re both a dancer and you make numbers dance, I’m assuming, so.

0:02:37.4 Megan Brown: You have to use all the parts of your brain.

0:02:40.0 Shashank Garg: Yes. I can understand your role, especially with the scope of international, with so many countries and different cultures and different people, and of course, different systems of data. So I can imagine the complexity of what you’re trying to juggle. So just to move forward, needless to say, I’m assuming you, with Starbucks international, you deal with vast amounts of data. What are some of the challenges you are encountering in your business? If you wanted to share?

0:03:14.0 Megan Brown: Yeah, I think one challenge is actually getting to an enterprise data platform. When push comes to shove, if your enterprise data platform isn’t in the top, whatever priorities, then you’re going to stand up multiple data platforms. Because everyone needs data and they’re going to do what they have to do to get there. And then you’re going to look back at this map of data platforms and you’re gonna say, huh, no one has a view of the whole company. And you’re gonna have to start to make changes on top of that. I think the other thing is we’re in an age of a lot of hype. And I think data science has generally got a really strong hype cycle to it where it’s like the new thing is so exciting, oh, it didn’t do what we thought. Oh, new thing, oh, so cool. Oh, it didn’t do what we thought.

0:04:07.0 Megan Brown: And I think that happens on small projects and large things like Generative AI is the obvious example right now. But the smaller things, for example, are data science use cases that they read about. Like we all read about personalization and all of these mainstream business things. And you’re like, oh, that’s great. And they say, it’s easy. You just need your data and you need a model and then it’s done. And then you work with teams who are like, yeah, let’s do that thing, let’s do it tomorrow. And you’re like, not tomorrow, actually. But one of the things we run into with that hype is they want the whole shebang, they want everything right away. But what they don’t realize is it’s going to fundamentally change their ways of work. It changes their business process, their marketing strategy, whether they trust data or not.

0:05:05.8 Megan Brown: And to go from one extreme to the other just sets you up to fail because there’s so many steps. Like, do you trust the data? Do you trust your loyalty tools? Do you trust your strategy to use this the right way? Is it returning the incrementality that you want? And if the answer to any of those is no, you’re just… This is gonna bomb, it’s just not gonna work out, so.

0:05:29.4 Shashank Garg: I think you bring up a really good point. And you talked about the hype that we see in the whole data science and AI, and especially fueled with the whole generative AI way, which came our way 18 months ago and two years ago, and then all the renewed interest in AI, which has always existed, but now everybody wants it. And the very real challenge of how do you help your business stakeholders navigate through the change? Do they trust themselves? And are they gonna let an algorithm take control and be accountable for what the algorithm is doing?

0:06:06.7 Shashank Garg: Yeah.

0:06:07.4 Shashank Garg: You’re bang on. Do you wanna, maybe, for the sake of our audiences, and this is a very real challenge. And I love the fact that you brought that up. Maybe take an example right there and see what exactly do you mean? Why is it so hard to trust and just go by it and in a fully automated fashion?

0:06:28.8 Megan Brown: Yeah. I can give a great example. So let’s say you have a brand new data platform. And this data platform is intended to take in your point of sale data, and one of your marketing platform data. And there wasn’t really any plan to engineer the data on top of that. There was a little bit of cleaning, but really, like, we just wanna show it as it is. Then when people go in to use the data, the data is still structured for machines. So to get your regular business metrics, you have to join across 12 tables. And like, there’s three people in your regional office that can do that. So the first step really is, if your data aren’t usable by humans, if they’re not structured for humans in a way that will make sense to how humans think, they’re not gonna be trusted. They’re going to always feel or look wrong.

0:07:28.0 Megan Brown: And then they’ll have their one or two people in the business that they think can do it the right way and they’ll pull different numbers. And then you have this conflict. So that’s one foundation is like, the data aren’t trusted, maybe we’re scared of the data because it’s new. And I don’t like no one would sign up and say, yes, I’m scared of new data. But we do see a lot of like fear behaviors. Like avoidance, attack whenever like the data doesn’t look exactly like what you want. Favored sources that are like, maybe not the main source. So if you don’t trust your data, and you’re trying to build some fancy models on top of it, let’s say for personalization, you’re not gonna… It’s kind of like the garbage in garbage out, but it’s like, not trusted in, not trusted out.

0:08:15.7 Megan Brown: Like, even if your data scientists are relatively sure that they’ve got the data the right way and the right shape with the right metrics and the right slices, if the business doesn’t believe that, because there’s just something fundamentally off. Then no matter what you build on top of it, it doesn’t matter if it’s a dashboard, doesn’t matter if it’s a pivot table, or a fancy model, the business is not gonna follow along the way you need them to. So you actually have to take them through a few stages of development. And you can do that at the same time as you’re building the fancy models. I don’t think these are necessarily serial, but they do have to have a source of data that they can trust, that is reliable, that is the main shared source of data.

0:09:04.0 Megan Brown: And they have to have a few people who can get the data into the right shape for them when they need it. But ideally, they’re going to dashboards. And then if they have new questions, or new metrics, or a new way of looking at some of their data, let’s say a driver’s model, they need to be able to see that and push it around. And then they’ll start to trust the models that are less explained, and less easy to explain.

0:09:29.8 Shashank Garg: Yeah, yeah. Well, you bring up a good point. You’re basically saying trust issues with data have existed ever since the field of data has existed, which is 3D case now. And data science and advanced analytics, it just puts another level of complexity and abstracts, and sometimes in a black box kind of way, which makes it even harder. And there is no way way around it. So you walk before you can run, and you take your users along and help them build that trust and there’s just no shortcut.

0:10:09.0 Megan Brown: There’s no shortcut and it’s kind of a culture challenge. It’s about changing how people make decisions for their jobs. And that’s very… Like, we as adults have a lot of ego. If someone said, we’re gonna do your job with AI now, you just sit there and watch it, I would really be concerned. So it’s probably fairly similar to that. I also think that once you have data in a model, I think what the business side doesn’t typically expect is that it’s gonna change how they do their job. It’s gonna change everything from strategy to their daily lived life. And I think if you think the data science solution is just like you flip a switch, and you get all this money, which sometimes you get lucky. But really, your team, your business team, like my counterparts in the business are going to have to change how they think and how they work pretty drastically to achieve those results. ‘Cause optimization, when you have humans involved, involves humans making different decisions than they would have without the data.

0:11:17.5 Shashank Garg: Megan, we spoke about challenges, and these are very, very valid challenges. Do you wanna also touch upon a little bit about some of the successes you may have had, either in your current role or prior roles, and sort of what worked, maybe use cases or tactics, whatever you wanna share?

0:11:37.0 Megan Brown: Yeah, I think in my current role, one of our big successes is actually having a use case that we have built over time with the participation of our stakeholders and our market leaders. So let’s pretend it’s personalization. So we started out fairly simple. And we have to go to these markets and say, “Alright, we’ve got something new for you. It’s not the fanciest thing, but it will get you to the fanciest thing. And we’ve got to start here.” And there were a few markets that are like, Oh, I don’t… We’re not, I don’t know about that, we’re not, like… That’s different. And we had to take it around to like 15 different markets. And we found two that were really excited about it and they were like, I’m not sure how we’re going to do that, but sounds interesting. Sign me up.

0:12:27.3 Megan Brown: And those two markets have really taken the lead in many ways and they’ve been nimble, and they’ve adjusted quite well. And what we’ve done with those markets, as we’ve built out this use cases, we’ve built it with them. So we listened to them about what is the next thing you want to see from this. What is the fanciest thing you can imagine? And would we actually do it. Our mission and values might actually, let’s say, not do dynamic pricing. Maybe you walk it back a little bit, what is the end point of this game and what are we building towards together? So you get the satisfaction of new features coming out fairly routinely. And then you also get the satisfaction of having this relatively big end goal. That’s been pretty exciting. And I think the other thing that I’m really proud of that’s completely different is from my prior role, which was about knowledge management and data literacy.

0:13:29.0 Megan Brown: Yeah. Completely different. You put a data scientist in charge of knowledge management, data literacy, what do they do? They make a spreadsheet of everything you’ve produced in the last three years, and they put it behind a dashboard. And they make it searchable and it’s a very silly search, it’s not smart, just a string search, nothing too fancy. But we were able to get into the APIs for everything we used. So we had a customer facing one that was like, internal Starbucks partners. What decks have we created? What videos have we created? What dashboards exist and what’s in them? So just making things a little more obvious, a little more transparent and easy to find. Because before that, used to have people, they would email their friend in Daibo and say, oh, have you heard about this thing?

0:14:18.0 Megan Brown: And the friend would say, let me go ask a friend. And that friend would… It’s like a two week chain of asking friends and we’re like, well, let’s just make it one search, use two words, get the deck, and you have what you need. We also had an internal facing one. So it went through the code, it went through Confluence, it went through Jira, it really did, it had everything from the external facing one and then also all of the ways that we’ve built these things. And I’m still really proud of that, I think it made a lot of people’s lives much easier and it sped up quite a bit. There was a bunch of other work we did around that that was very… It felt very much like public affairs, public relations. But that was like the data science kind of heart behind public affairs. Yeah.

0:15:08.4 Shashank Garg: Yeah, yeah. So if you were redoing it today in today’s world, would it be fair to assume that you would be tempted to put a Generative AI led chatbot or narrative summarization engine on top of it, right?

0:15:24.4 Megan Brown: I would be tempted to do that. Especially, the reasons I would be tempted are probably not the big hype reasons. The reason I would be tempted is there’s no PI. Like, I could get that through our data use review really easily, and it would make it more convenient. People would probably be excited to engage with that chatbot. What do you know about transactions per sales day?

0:15:47.0 Shashank Garg: Exactly.

0:15:47.2 Megan Brown: Gosh, there’s two paragraphs and you put in your slide deck, suddenly, everyone has the same metrics definitions that would be bad. Yeah.

0:15:55.0 Shashank Garg: Yeah. Megan, on that topic there’s a lot of people trying to experiment and push the envelope when it comes to use of generative AI models or large language models on top of structured data. And we know we had these challenges and it’s just really hard, not that easy. What are your thoughts on, like, do you see your users conversing using a chatbot with not even the models, yeah, but like just basic reporting? Yeah. Metadata or actual data, both actually, yeah.

0:16:32.5 Megan Brown: We’re not there yet at Starbucks. And I think one of the things that slows us down is really our mission and values. So we have to come at things not just through, like, what’s the most efficient, but also what works for our customers and our partners. And we’ve been through great change, so who knows? This could change in the future, someone could get very, very, very excited. What I can tell you is that business leaders in international are very excited about it. They want it everywhere all the time, no matter what and I think I’m a little bit more of a naysayer, it’s kind of a part of my personality. So like, as soon as there’s a hype cycle, I’m like, hmm, I don’t know. I don’t know, it’s not ready yet. I do think generative AI will change a lot of what data science and analytics does a lot.

0:17:25.0 Shashank Garg: Yeah.

0:17:25.8 Megan Brown: So I know that is coming, I do believe that, it’s not really there yet. And so I get more excited about tiny use cases that are more about efficiency. Like, Copilots. Yeah, let’s go. Like, if it gets the first draft to the code, great. We can do the fun stuff like making it fit our code base. I have a use case that basically creates, applies a seven layer hierarchy to new data from old data. And so like, let’s pretend you have a bunch of products that you sell in your stores. And there’s an enterprise way of looking at those products, but it’s not quite enterprise, it really only works for North America. And let’s pretend you finally have some data platforms and international, and you have SKU level information. How do you apply that same hierarchy or a similar hierarchy with maybe some new categories?

0:18:23.8 Shashank Garg: Yes.

0:18:24.0 Megan Brown: To the new data quickly, without taking a month of a person’s time to go through line by line, fill it out.

0:18:30.7 Shashank Garg: Yes. Absolutely.

0:18:32.4 Megan Brown: Yeah. You train a Generative AI model to do each layer of that hierarchy and you have a human check it because someone’s got to use these eventually and they have to trust it.

0:18:41.0 Shashank Garg: Yes.

0:18:41.4 Megan Brown: So that’s, I’m really excited about that, but it’s so small. And it’s really is about efficiency ’cause we’re gonna have SKU level categorized data for every single one of our markets in the near future because of this work. But it’s also not like it’s not driving incremental value, it’s not going to be life changing, it’s just gonna be something that goes in the background someday. Oh, you’ve got a new item, run it through the thing, and then it shows up in EPH. Great. Wonderful.

0:19:09.2 Shashank Garg: Yeah. Yeah.

0:19:11.0 Megan Brown: But it’s like, I actually, what I want from generative AI is I want it to do my dishes and laundry. I don’t want it to do the fun parts of my job. The creativity, I like writing. I don’t, I think it has a pretty bland voice. I don’t know that I’m gonna use it to make like a strong pitch deck. I want it to do the stuff that’s really boring and takes too much time and everyone rolls their eyes at, but does it anyway? Like, let’s do that. That would make my job so much better.

0:19:41.4 Shashank Garg: Absolutely. Megan. The way I am seeing a lot of our clients that we work for so sort of use of generative AI for all the backend stuff, which never gets seen. And sometimes it’s really hard to justify value and get a project approved there. But the beauty of it that I’m seeing, you don’t even have to get a project approved, the incremental investments in training, something like what you just said. It’s something that we as data leaders or analytics leaders should be able to find the bandwidth to work on. And if you take, the example that you just took, not that it wouldn’t have gotten done earlier, but somebody would be, nobody likes that kind of work.

0:20:28.3 Megan Brown: It would’ve made six people’s jobs really terrible for a few months. Would’ve.

0:20:35.2 Shashank Garg: And if that’s the case, then humans are very good at not doing a good job there. And if they did not categorize them properly, the repercussions are far greater. So if you’re making the boring work easier or cutting it down by 80, 90%, I think that’s a huge win. We know all the hype is around the user side of things that can I talk to data and these chatbots, and I think those will also evolve. And we’re starting to see, we’re starting to see platform players like Snowflake and Databricks and AWS Azure. Everybody get out there and say, here’s the fun way of you look at the fabric architecture and you can talk to your data. But there are challenges. And I think as a consulting firm, as an AI solutions firm, we often actually get to focus on all the backend stuff. And I think that’s where some of the real wins are in the short term, while we all work towards let’s talk to the data and the metadata, and then everything will be very easy.

0:21:45.0 Megan Brown: I think a lot of the same sort of thought processes that go into this smaller, less exciting backend work, will train the business to do the really exciting stuff. You’re not gonna have one bot to rule them all. You’re gonna have a bunch of little bots doing their specific jobs. Maybe at some point you’ll have a secondary layer, tertiary layer. Great. Cool. But you need to build those little bots first and you need to figure out if they need to talk to each other, how you need to track them, like how you’re maintaining them. All of these basic tech things that we don’t know yet. And so you need that in place, whatever AIOps is going to be, before you really have thousands of agents talking to each other, potentially making prescriptive decisions for your company. Like where is your checks and balances in that?

0:22:34.8 Shashank Garg: I really appreciate you bringing this up. I think when the hype is important, the Salesforce agent force is important, and all the agent tick AI is gonna, what you just described. It’s gonna happen. We all know it’s gonna happen, but before it can happen, there is so much work left on, even you just talked about something really, really basic as number one challenge being having a unified data platform for what you would call is a brand and a firm that we all love and probably has good systems already.

0:23:11.1 Megan Brown: They do. They’re intelligent.

0:23:11.3 Shashank Garg: And even there, the work is not done. And we’ve been doing this work for many years now. And that’s not uncommon across any of our clients. The setup of the data platform and ensuring the right glossary and metadata and quality and consistency. That’s hard work. And if you apply AI in general, and especially generative AI to reduce the treasury of people who manage it I think that’s a huge win. So thank you for bringing this up.

0:23:46.6 Megan Brown: Yeah. Well I think someone like me is really useful when you pair me with a pitch person. We have to work together because a pitch person left to their own devices will promise the sky, the sun, the moon, the stars, everything really fast. But to actually live through whatever the pitch person gets funding for you actually need a there there. And so working together with a pitch person actually makes their pitch a little more down to earth, which they don’t always enjoy, but also means that you have somewhere to start. Like you have a plan for how you’re going to get there, even if it’s not turn the switch and make all the money. In reality, we all know it’s a longer road than that.

0:24:30.3 Shashank Garg: Makes sense. Makes sense. Megan, one more question before we talk about some of the future stuff as well, which is what we’re discussing. Like if you apply whatever you’ve learned at Starbucks and we talk industry in general, maybe it’s retail, maybe it’s CPG, maybe it’s the whole, I don’t wanna call it fast food, but you get the point. What are some of the things that you would recommend people really look at in terms of use cases? Any advice there?

0:25:02.8 Megan Brown: One of the first places you should look to train up your teams are the least exciting ones. ’cause no one wants to build their first agent for the first time in something that’s super high profile. Maybe you have one person who’s like, ready to go with that and they’re working on that. Your other data scientists for whom this is new, like we all put AI on our resume, but this is new. And we’ve taken classes, but have we built the thing? Probably not. So in heading in that direction, like give them projects that are your boring cleanup metadata projects, data quality, and that’s like just a great thing because it makes their work better in the long run. It makes people’s lives easier in the long run and they can mess it up a couple times and still get to the end point.

0:25:49.1 Megan Brown: That just seems a reasonable place for a learning data scientist who really wants to get into AI to start and they can try the fancy things, by all means go for it, but it’s contained. It’s not gonna raise any, no one in the press is gonna be like, you used an agent to do meta. No. So it’s the risk is really low and the reward is still there. Even if it’s not incremental money. I still think there should be an investment in whatever’s coming next. You still need your seed funding, you still need your proof of concepts, you have to, they’re not, I think the other side is that in business we are successful all the time. All of us, no matter what are successful all the time. So having those things that may not be successful is feels bad.

0:26:38.6 Megan Brown: But you have to have some tolerance for failing fast, some tolerance for what happens if we put 200K into this thing and it didn’t work out, is that gonna be okay? Are we gonna fire people or are we just gonna say, alright, let’s try it again this way? That’s the kind of room that people will need to do the big stuff. And some cultures have that and some don’t. And so if you don’t have it, you have to figure out how to build that at least in a small corner of your company.

0:27:06.3 Shashank Garg: That’s great advice. Thank you so much. And one of the things that we often talk about with our clients is, especially now, and at least start with an AI first mindset where there are traditional ways of doing things and at least start there. By that we also mean that failure is a given. And some things are just gonna break. So that whole culture of experimentation failing fast, that’s really possible with AI now. I don’t have to wait nine months to build out something and realize that this didn’t work. I can actually produce something in six weeks and throw it if it didn’t work.

0:27:45.0 Megan Brown: Indeed.

0:27:45.9 Shashank Garg: So I think just shifting to that AI first mindset with rapid experimentation, and I’m actually seeing, I’m quite excited that I’m actually seeing organizations start there the way exactly you said. Start with some seed funding. There’s buying from the top and the whole people becoming vulnerable and saying that, Hey, this didn’t work. It’s becoming much more of a reality than it was 10 years ago at least. Let me just say that.

0:28:08.4 Megan Brown: And I think that the truth is that there are a few big companies that maybe can make this work within their big companies and companies that aren’t tech. Tech first. You’re gonna have to mess around and find out there’s a, like you really will have to figure out what works for you, what works for your employees, what works for your processes, and then where you can find value out of it. And sometimes that value will be money and sometimes the value will be efficiency in roles. Like your clever data scientist now spends 30% of their time getting data ready for a model instead of 80%, and they can do more models. That’s amazing. So that’s what every data scientist wants. So it’s just like it all lines up. Yeah.

0:28:57.5 Shashank Garg: At some point the real definition of a data science can have less of data engineering…

0:29:00.9 Megan Brown: Yes. Less terrible data engineering. Let’s be clear about that. The data engineers do great data engineering, data scientists do data engineering.

0:29:14.4 Shashank Garg: Yeah. Megan just moving ahead. For organizations like Starbucks of that scale and many more what are some of the trends that you are seeing that you think will become real and everybody who’s listening into this podcast should really be paying attention to?

0:29:33.2 Megan Brown: Yeah. Well, I think, like I said, GenAI will be transformative, right? Like, it’s going to be, and your organization needs to be ready for that. You need some thought on how it’s going to transform your business, even if that doesn’t end up being exactly the way it happens. You should be thinking about the future because your employees are already nervous. We all know what’s going on in the job market right now. Your employees are already nervous about AI. If you don’t have a thought or a strategy, they’re just going to stay nervous. It will change things. Your employees know it’s going to change things. You should also know what it’s going to change to some extent. I think watch out for regulations like I work in international GDPR and everything that EU does is the most kind of conservative take on data, ML, AI if you are a global company, like ideally, you already have people thinking about this and they just need to train themselves on AI to get going. But that’s fast changing. Like it’s rapid and the US might not be the foremost…

0:30:48.7 Shashank Garg: Yeah. We’ll get there.

0:30:49.5 Megan Brown: Leading the charge. Like us and our very distributed non-regulation kind of world that doesn’t pan out that way other places. And then I would say make sure that everyone that you have working for you that’s gonna work with AI, which is everyone is ready. There’s just some fundamental training that needs to happen, even if you’re not building the things. Is your data trust where you need it to be? Is your data culture ready? Are people scared about their jobs or have you given them some sense of a little bit of security or stability?

0:31:27.5 Megan Brown: And then do they understand what these models do in like a conceptual sense. They don’t need to know how to build it, but do they understand what it’s doing? And then that would help them trust it a little more. And I think be very clear about whether it’s predictive or prescriptive. I think there are actually very few processes. So deep learning, for example, there are very few places where businesses are actually comfortable with deep learning because you can’t tell what it’s doing. It makes decisions. It has these layers that do different things. You don’t know what it’s doing. And mostly people are not comfortable with that.

0:32:05.0 Megan Brown: So there’s only really a couple of places where deep learning really comes in handy in business. AI’s kind of the same way. What you don’t really know what the models are churning through, you know, what you trained it on. You might have some explainability metrics, but you don’t really know where are you comfortable making decisions without a human check when you don’t know what the model’s doing? And I think the answer to that is very few. You’re probably going to lean more towards predictive and you’re probably gonna have a human somewhere in the loop, at least one place to make sure what’s coming out of the model is consistent and reliable.

0:32:41.9 Shashank Garg: Absolutely. And thank you so much for sharing that. In our clients, if I look through, what are some of the things that I’m seeing becoming very real? And you talked about you started with GenAI and those, you know, persona-based assistance, whether if you’re a B2B, the whole presales assistant, if you’re a seller, a B2C seller, then personalization or being able to personalize your marketing campaigns. I think just looking through all the roles, like even like we are a data AI solutions services firm. Then use of copilots for our developers, use for better summarization engines on our knowledge for our consultants. So just persona-based assistance. We actually, for a pharma firm, we did assistant for the CFO’s office on financials and bringing down cost of clinical trials, data management. So that, I think that’s gonna become very real.

0:33:38.3 Shashank Garg: You talked about sort of building trust in the data. I think that’s gonna continue to be very, very important. You brought up attention to regulation. And I fully agree. I think US will come in there, we have seen at least two, three regulations on AI just in the last 12 months. And I think they’ll continue to strengthen. One thing that you had spoken about earlier is the whole, right now there’s just too many tools. AI tools in all the different layers. I think when I was talking to you, you had mentioned that there will be some consolidation.

0:34:13.7 Megan Brown: There will be some consolidation. I’m actually of two minds about it. Where it’s like, often you get a lot of innovation from smaller companies, but if the smaller companies are using the larger companies platforms, I don’t know how far they can go. And there’s a big risk to the smaller companies there of just being eaten. So in some ways I’m excited to see what the smaller companies are doing, especially if they’re dishes and laundry, great, let’s go. But then I like there are so many, and companies like Starbucks I think might choose five, but they’re not gonna choose 40.

0:34:53.3 Megan Brown: And eventually after they have those five, they’re gonna wanna narrow it down to one, because that seems to be like a process I’ve seen in a couple different companies. And how is that consolidation gonna happen aside from the pressure within large enterprise? I think we’re just gonna see a lot of shifting and changing. And so I actually don’t know how people are investing other than like a peanut butter approach. ’cause it’s like, who’s gonna win? We don’t know. Maybe the big ones. Yeah. Right. So I really don’t know.

0:35:24.1 Shashank Garg: And I think what I’m seeing across all our clients, Megan, obviously this has been talked about, the higher you are in the stack, your startup, the lower the amount and risk of you getting cannibalized. And at least we are starting to see across our clients that at least at the top layer you can build, the tools are available, the platforms, APIs are available, the models are available. You can put them in your own cloud, you can put them OnPrem if you wanted that within your tight wall. So you take care of security concerns and all that. And the cost keeps coming down. At least with the generative AI piece, the cost just keeps, from the first point they came out versus today is probably a 100x decrease for maybe a 20x better output.

0:36:16.4 Shashank Garg: So I’m excited that level of innovation, and I think there’s a lot of build opportunity for everybody who’s in this field. For people who are listening in. If you’re a data engineer, if you’re a data scientist, if you’re a business analyst, application developer, I think there’s just a lot more you will be able to do. So continue to monitor all those startups. Those are great ideas. Some of them will get consolidated, but even if you don’t buy them, you may be able to build something that is specific to your business, depending on your business’ appetite is how I see it. So it’s good times, great times to be in as I say it. Just to wrap things up any sort of final piece of advice for our listeners? We’ve got people from the business side, we’ve got people from the data side listening in.

0:37:03.7 Megan Brown: I think this will be a little, this is kind of strange advice coming from a data scientist, but I’m going to give it anyway. One of the reasons I write and do modern dance is because it keeps me working on it, having an open mind. I am more hopeful and better able to envision a different future because I do those things. And knowing that the future’s gonna change rapidly. We’re going to live through a great deal of change. It’s already started. It’s coming for all of us. What do you need to do to maintain an open mind about what’s coming? Because it’s so easy to settle into fear, like, oh my God, will I have a job? It’s really easy to land there when you’re thinking about generative AI.

0:37:52.4 Megan Brown: But actually if you take a more proactive approach, or if you imagine the parts of your job you dislike going away and then imagine a path to that keeping an open mind over the next five years is gonna be a ridiculous asset. So while I’m a bit of a naysayer, I still know it’s coming and I still think about how we’re going to use it. And I try very hard to not land in the fear space about like, this is gonna change everything. Oh my gosh, I don’t know what to do. And it will change everything, but I intend to be a part of that change, not really as affected by that change. And so whatever you need to do to keep your mind open and keep your views of the future a little more optimistic is a great thing.

0:38:38.5 Shashank Garg: Yeah. That’s a great way to put it. And in my role as sort of the founder, CEO of InfoCepts, I get these questions quite a bit. We work with a lot of people both on our client side and everybody that we hire and train. And while in theory we expect 80% gains of productivity just using copilot for developers, we’re still in serious shortage of good data engineers. And that’s a reality. And even on the business side, people who become this translation layer for business leaders who are ensuring that you have some new tools now and as far as you’re staying on top and keeping an open mind, then your demand for your skills will just keep going up and up. So fully agree. Thank you so much Megan for everything that you shared, this was, I must say this was one of the most practical, realist, sort of real kind of sessions, so I’m sure our audiences enjoyed. Thank you.

0:39:42.8 Megan Brown: Thanks. Thanks for having me.

0:39:44.5 Shashank Garg: To our listeners, we hope you enjoyed this episode of The Intelligent Leader with me, Shashank. If you liked what you heard, please consider sharing it with your network, using the hashtag The Intelligent Leader. And don’t forget to subscribe to our podcast by hitting the subscribe button you wouldn’t wanna miss many more episodes. Thanks for tuning in. Thank you.