Building Pragmatic AI: How Andy MacMillan, CEO at Alteryx, is Empowering Enterprises to Harness Data Effectively

0:00I think that's where a lot of people are still at. They're thinking about how we build on top of these things. I think we're heading in a very different direction. AI works a lot differently than these traditional applications. And so I think building sort of alongside is a big piece of advice I would give people.

0:14Welcome to Data and AI Mastery, the podcast where we bring you cutting edge insights, practical advice and inspiring stories from the leaders shaping the future of data and AI across the globe. I'm your host, Raoul Gabriel-Urma, founder of Cambridge Spark, the leader in transformational data and AI upskilling, career development and progression. In each episode, I will be diving into real world case studies of companies harnessing the power of AI to drive innovation, reduce costs and create new business opportunities. So whether you are an aspiring data scientist, AI engineer, or seasoned executive, this show is designed to give you the tools and knowledge to stay ahead in a world where data is transforming every aspect of business.

0:59Stay ahead, stay inspired, stay masterful. Welcome to Data and AI Mastery. Hi Andy, how's it going? It's going very well. It's nice to have a little time here with you, Raul. I appreciate it. 100 % welcome to the Data and Inai Mastery show.

1:22So Andy, to kick us off, you're the CEO of Altrix, big company, big job. Can you tell us a little bit about Altrix first of all? Sure. Altrix is a platform that helps people that understand data, but aren't necessarily developers. So think of your, maybe yourselves or friends, you know, who are very good at things like Microsoft Excel or working in Google Sheets helps them build really incredible automations by being able to bring together data from all kinds of different sources, kind of prep, blend, and build solutions with the data, and then maybe output it to reports, dashboards, or now more and more often AI solutions.

2:00So we really think of it as a way to help people solve some of the messiest, most challenging problems around data by kind of applying their knowledge about the business and about the data. And then again, setting up automation so that that can run at scale and create productivity. Awesome. I look forward to deep dive in this topic with you. Before we do that, Andy, can you tell us a little bit about how did you become a CEO of Altrix and what led you to take the role? Well, my background, I've been in tech a long time. I graduated from Michigan State with a degree in information technology and a minor in computer science.

2:35I actually started as a developer. And I joke with people that I wrote such good code, people tell me maybe be a product manager. So I shifted into product management, which I really enjoyed. Product management sort of being thinking about how to solve problems for users and working with engineering teams and designers to do that. I came up through the ranks at companies like Oracle and Salesforce, where I ran fairly large product lines. And then ultimately took my first CEO job at Acton software, which is in the marketing automation space. I recently ran a company called User Testing, which was all around getting feedback from users and helping people build better, more engaging products.

3:11And then I joined Alteryx as CEO this past December. Amazing. And how has your background in product management led you to think about the data and AI space and think about innovation in this industry? Well, I spent quite a bit of time just around data in general. As I mentioned, I worked at Oracle at one point, which is pretty big in the data space. I then actually ran a product line at Salesforce called data.com. So I've been in and around sort of helping people solve problems with data as a product person. But I'd also say as a CEO, a big part of your job is working with data across the span of the company and trying to figure out how to solve problems using data to help you make informed decisions.

3:55And I find as a CEO, one of the things that's really interesting is a CEO doesn't run any part of the company. You're not the head of sales. You're not the head of product. You're not the head of finance. You work across those teams. And I often find that that means the data that I want is sort of the intersection of where these teams are working together. And so I sometimes find the CEO job in particular works with that sort of messy intersection of data of sort of like, well, how is our sales data and our finance metrics sort of maybe coming together the right way? Or, you know, how does our product analytics and our customer success motion come together?

4:29And so I've always had a passion around understanding the data that drives the business. And so it was a pretty big opportunity for me personally to come work at Alteryx, a company that I think is very focused on helping people use data to solve problems. That sounds great. You've clearly been swimming the data world all your life. And I can really resonate with COSB, like Chief Everything Officer. You need to have your fingers everywhere and, you know, be in touch with every department. So that's great to hear.

5:04I'd love to take you now into the data analytics space and, you know, more focus on the work that Alteryx is doing. So today we talk a lot about agentic AI, right? Every conference now, every show, you hear about agents everywhere. So could you give us your perspective around how do you think the data analytics field is evolving? Maybe traditionally you'd think of a data analyst, data scientist writing SQL queries or working with an interface to do sort of data analysis and aggregation. How do you think about agentic AI in this world now? Well, I think one of the big promises of agentic AI is the ability to put AI to work to solve kind of scale business problems.

5:53One of the challenges in all of that is in order for an agentic AI system to understand your business problems and your customers, your users, whatever you're trying to solve, it does have to understand your data and how your data relates to your business processes. And I think this is one of the things that a lot of folks are discovering when they bring AI into their enterprise environment is, yes, it's a really kind of smart tool that helps me maybe make an email more concise or something like that. But if I turn around and ask it, you know, hey, I'm in sales, what will my commission be if I close this deal?

6:28It needs to understand a lot more about your specific environment. And what we're hearing a lot about is companies struggling to think about how to bring their data into an AI environment in a way that the AI understands their data that's predictable, that gives the right answer, because a lot of their systems aren't set up to do this. And so to me, the promise of agentic AI is massive. There are real challenges in how companies are going to inform that AI to make decisions in line with their business in line with their numbers. It's not as simple as just, you know, bringing in chat GPT and letting it crawl through your cloud data warehouse, like it will not make sense of your massive, you know, pile of data, you have to be deliberate and think through how to provide context and the right calculations in the right context.

7:12And so I think that's the big opportunity. I do think it's real. I think we started using it even internally here to solve real problems to give correct answers to kind of business queries, but it's a path we're all going to have to go down. And I think companies are going to have to make very deliberate decisions about how they're going to empower not only their employees with AI, but their AI with their data. Makes a lot of sense. So what sort of guidance would you give other CEOs, you know, if they're looking at this space and they think you're investing in agentic AI, would you recommend they start to think a bit more about the data infrastructure, the data quality, the data assets, or different perspective?

7:53What I think you can do as a CEO is start to say, how do we pick discrete problems, things that AI can help us solve? How do I empower a team to put together a data pipeline that gets the right data with the right context to inform that AI? How do I put in place an approval process? So as an executive, I probably have a board that cares about how our data is being used. I have an executive team that cares about how our data is being used. And I'm sure I have a data privacy team and a CISO and an InfoSec team that cares about how our data is being used. So how do I put together a process where I can get a set of data and run it through an approval process that like, yes, this data flow can inform AI.

8:34We often refer to that at Alteryx as setting up what we call an AI data clearinghouse. So a construct that the company can work around for bringing data into that environment. And how do I work with the reality that my data is messy today? You know, I talk a lot about Alteryx likes to work with messy data. So when you show up and tell my team, you know, hey, yeah, some of it's in Databricks or Snowflake, but some of it's a report we run out of Salesforce and some of it's a spreadsheet that sits in the finance team's office. We're happy to have you pull that stuff into Alteryx and build a data pipeline to inform your AI.

9:05And I think that's sort of the crawl, walk, run strategy that I would give to most executives is, you know, don't wait for your all of your data challenges to be solved. That'll take a long time. Don't just let certain teams start throwing data into AI. Instead, how do you empower teams to start building that context, that data pipeline? In the environment, do you think specifically conversational AI makes data analytics more accessible to the wider mass? Or is there like a place for some more specialist skill set in the environment? I think it's both. what I'm finding when I work with it, I'm spending a lot of time right now building some of my own GPTs and things internally.

9:49Sometimes there's a narrowness that helps, right? So I just built one where I'm analyzing my pipeline, like my sales and marketing pipeline. And having it be really clear on what it is that I'm gonna ask it about in a very specific data set, I think yields really good answers. It doesn't sound like I'm asking a generalist. The flip side is, you know, I think there's also the ability to have sort of general chat interface AI, know a lot more about your business to help people that maybe wouldn't go through an analytical process be able to get an answer informed by analytics. There are just a lot of people in this world that aren't necessarily data people, and they're not going to say, hey, I'd like to go out to my BI dashboard and investigate this.

10:33They're going to go ask around to get the answer. And I think there's a possibility that AI could be the way they kind of ask around. And they go into this chat interface and they say, hey, what's going on with this? And if it can use data to provide an informed answer, I think that's really helpful as well. I think it's sort of both. I think there's a role for people that want to get into the data, want to make sure that there's a very analytical approach to doing AI can do that. I also think there's a more general purpose, like how do you make data informed answers part of how the average employee just interacts with your systems.

11:06Great, great. So there's definitely a place for both. But I like that you're so hands-on as a CEO and you kind of get stuck in the data yourself. I think that's really refreshing. it's it's uh it's something i tell every ceo like i think rolling up your sleeves and getting involved in in your product involved with your customers um like i said it's a weird and hard job you you don't run any part of the company but you're sort of responsible for all of it but the job gets a lot easier if you talk to your customers you talk to your employees and you'll actually understand whatever the thing is that you're building for the world like those core principles sort of help make the job a little bit easier 100 you've got a passion but you also walk to work.

11:44So that's really great. So if we take to Alteryx specifically, can you share a couple of examples in terms of ROI that customers can expect from investing in data analytics? Well, I think a big one is, frankly, just better outcomes. And those can be measured, you know, the three big ways you always measure value is sort of, you know, revenue, managing costs or managing risk. But I think it can really drive to your core outcomes. A great example, you know, our partners at McLaren, we work with McLaren Racing quite a bit. They're a great partner of ours. They use Ulteryx extensively in their business.

12:23They're very data driven. When you listen to them talk about, for those who maybe don't follow F1, they're currently winning the F1 race, both for the driver and for the team. And they talk a lot about being very data driven. And you think about things like the reliability of the car and all the parts and the systems and the supply chain that make up the car. You know, they have over 1.8 billion data points that make up their race data set from the car to the reliability to all the sensors on the vehicle. Bringing all of that together and making decisions is how they win races. So I think it could be a little bit bigger than simply revenue, cost and risk.

13:02Like, what is your core goal? that can use data to get to those outcomes. I love partnering with our customers and our big partners to go do that. We see some really incredible outcomes. So that's at the company level. But we also, even in our founding, sort of go all the way down to the personal level. I mean, how many of us have repetitive data tasks that we have to deal with where every week, every month, if I'm in, I don't know, the accounting team, do I have to reconcile everybody's, I don't know, budget versus actuals. You know, what was their budget? What did they actually spend? And for decades, people did that either by hand or on spreadsheets, and they did it as a repetitive task every month.

13:43And one of the things that we focus on is automating that kind of data work. And so we can save individuals incredible amounts of time doing that kind of work. And instead, they can be then evaluating, like, how do I work with the outcome? How do I help the team? I don't know, maybe meet their budget better versus spending all my time sort of reconciling the data. So everything from the, you know, the highest goals of, you know, McLaren winning races down to, you know, can we help accountants close out the month and get their job done more systematically? All of that can really be about using data to solve problems.

14:18And I really enjoy that we're kind of right in the middle of that. You speak my language, Andy. I'm a big F1 fan. So it's great that you partnered with McLaren. So if I play this back, so on one side, investing in data analytics and the platform allows you to optimize your business, allows you to execute on your strategy, being data-driven. On the other side, every business has a lot of inertia when it comes to working with data. Things like doing reconciliation, like you mentioned, and you can automate those processes and maybe make them a bit more intelligent to save people's time and hours so they can focus on the business.

14:59So that's great. So in this context, what's your vision around how AI can support those goals specifically? I think there's a couple areas where AI will be really interesting in our space. One is every one of these automations that folks build today with Alteryx, you can imagine putting, instead of a dashboard on the end of it, you could put an AI analysis agent. So don't just show me the, in this example we were using earlier of reconciling budget versus actuals. Typically what companies do, they run a report and you create little red, yellow, green, like who met their budget, who didn't. And then everybody goes, well, why is that red?

15:42And then you spend all this time sort of digging in, right? Or was that red last month? Or how long has that one been red? And what I think AI can help us do is that first pass of the analysis. So now instead of getting a dashboard, Maybe what you're getting is a fully fleshed out report saying, hey, this team is still red, but they're actually trending in the right direction. Maybe this other team is green, but they're actually trending closer to red. Like, actually, you need to dig in and understand what's going on here. AI is really good at doing pattern matching, looking at trends over time.

16:11So one is I think we can use AI to sort of get a level deeper into analytics. I think another is a lot of the automation that we do today around data, we could start to do as well to include more semi-structured and unstructured formats because AI is also really good at working in those formats. And so, for example, at Alteryx, we've been building tools that you can use right in Alteryx to bring in the large language model of your choice as part of a step in a data-driven process. So imagine reconciling, you know, a large set of data, for example, but maybe one step in the process, I want to look at all the maybe the open form fills and categorize that into six categories.

16:53That's something AI is really good at. Now, I still want the rest of my kind of prescriptive workflow to happen, right, with the tools that a data analyst would need. But I think AI can be really powerful in these workflows as well. And so I think both sort of informing large language models and agents with data is one piece. But the other is really changing the way we work with data to think about, you know, a whole new set of tools and capabilities to let us sort of, you know, wrangle our information in a consistent way. And so doing that in a kind of process container like Alteryx, I think, is really interesting, too.

17:30So that sounds really cool. So I guess one question that came to mind, the space of AI can be quite overwhelming because it's moving so fast, right? We're talking about LLMs, this new version of models coming through. You can integrate them in your workflow and so on. Do you have any recommendations for business users and clients? How can they upscale or how they can get on that journey? Because probably a lot of people need a bit of education to be on board. I think part of it is not trying to tackle it as this giant thing, but to start solving discrete problems. And the other thing I would say is, as much as you can get to kind of a new architecture versus just trying to build everything on top of your old stuff, I think the easier it gets.

18:18And so a lot of what we've been helping customers with is, you know, maybe they have a whole portfolio of business applications they're using. They probably have a bunch of business processes orchestrated across those applications. They've been building these for 20 years. You know, it can have a little bit of a nobody touch that. It might not work anymore if anybody touches it kind of feel. But they also replicate, for example, all the data from those systems into their big new cloud data warehouse that has that data in it. And so you can actually go to that cloud data platform and bring in a tool like Alteryx to pull in a subset of that data, you know, pull it together, have it make sense, give it some context, hand it to something like a GPT or Gemini or Anthropic and build an agent that can support your business.

19:06and what you're not doing is building it all on top of that 20 years of people building stuff in those apps. I think the first pass at creating agents, everybody thought, well, I'll just string together all my legacy apps and have them get smarter. And I think that's really hard. Your legacy apps are not expecting to feed AI. It's a really narrow straw to consume through these APIs. And so what we're helping people with is like, hey, those apps are still running. I'm not saying to turn those things off today, but maybe you create a set of agents that complement those apps where somebody's just going over to, you know, a chat GPT and maybe they're, you know, configuring a Salesforce opportunity, but they're doing it all through a chat GPT interface and they're doing it right off their data.

19:48And when it's all done, it gets put into Salesforce, but you didn't have to build it on top of that. Or maybe they're, you know, opening up, asking some benefits questions and they're doing it off of the data you have in a workday or an SAP, but you're just using the data to answer the question. And I think that gives people this feeling of being able to have kind of an emerging architecture they're moving to that, again, supports these apps, but you're not trying to build on top of everything. I think that's really hard. And I think that's where a lot of people are still at. They're thinking about how do we build on top of these things.

20:17I think we're heading in a very different direction. AI works a lot differently than these traditional applications. And so I think building sort of alongside is a big piece of advice I would give people. And we're doing it today. I'm doing it inside Altrex. We're building a whole set of complimentary kind of GPTs that work alongside our enterprise applications. We're building them quickly. We're building them with all the correct data. They give the right answers. And we can do it at scale and at speed. That's super interesting. That's a great, great advice. So building alongside, thinking about reimagining rather than patching up your legacy system it's hard to let go right um but i i'm with you in this context where the technology is so transformative and uh the user interface is so different it makes sense to kind like start on the side and and start with with new because it's so productive and you can get resolved fairly quickly um so if that's the case then can you walk us through i guess some of the challenge that you're seeing from customers that are wanting to invest in agentic AI?

21:25I think one of the challenges you've described here is at what point do you do incremental improvement of your existing application versus you just start a new one, you know, maybe AI native from the get-go. What sort of other challenges are you hearing? Well, I think one of them is I think it can be such a transformative technology that it's a little bit like the more you work with it, the more you realize you would change your business process quite a bit. And so I think that's sort of a learning curve a lot of folks go through. And again, I think that challenges a lot of your business processes are encapsulated in these line of business applications.

22:02And so you end up with this tension of, you know, how do I re-engineer my business process, but I probably still want to support it by these business applications, but I want to have an agentic model. And so I think that core tension is something that I see a lot of organizations trying to work through. So I'd say That's one. So how do you give yourself the technology footprint and the flexibility to really iterate on your business process as you figure out what AI can do to help your team be more productive? I think the second is the cycle times are so different. You know, this is not a traditional IT project that you're going to roll out over 12 months.

22:36Not only the AI changes so regularly, but also, you know, you can move really quite quickly when you think about AI coding capabilities and the speed at which you can build things. Again, I mean, I built my little GPT for analyzing our pipeline in a weekend. You know, I mean, that is something you just didn't do three or four years ago. And so you have to, as a business, think about what's the pace of iteration you can do, not on the tech, but on the way the business works, on how the team of people you're supporting interact with this. And so I think that's another challenge that we see organizations working through and thinking about how to do that.

23:12I think there's an employee upskilling challenge. So how do you get people comfortable with the idea that the AI is going to help them be more productive, not just do the job that they're doing? And I think the reality is it will maybe shift the ratios of jobs. So when you think about a team of people that are working together from three or four different types of roles, AI might do some of those differently. And so you might sort of be adjusting how teams are structured. So I think that's part of it. And I think the last one is maybe where we started, which is, you know, getting comfortable with what data are you putting into these AI tools.

23:46And there's a lot of things to think about from, you know, privacy, you know, information leakage, data security. You know, that's an interesting one as well. But your final point around security, guardrails, governance is a great one. And I'd love to get your perspective around how do you think about responsible AI at Alteryx? I know you guys have thought quite hard about it and you've published your principles on your website. So could you walk us through a little bit? What's your perspective on it? Well, I think it's a very wide-ranging topic. So I think there's sort of responsible AI when it comes to the large language models themselves.

24:25So you think of these big companies doing this training and how do you think about, you know, bias and things like that in those models. We thought a lot about with our customers, how do we give them flexibility and choice so they can pick the models that are right for them? Because I think that's going to be an ongoing evolution as people work through that. I think a lot internally in our company about what data we put to work in AI. And so we have created a small group, our AI strategy team, and we've created an AI data clearinghouse model. And so when people want to use any Alteryx internal data in any AI system, they have to bring their Alteryx workflow to that team and includes me and my CFO and my head of legal and my head of privacy and my head of InfoSec.

25:13And we get a look at that and we have some questions like, is this, you know, is this the right level of data to be in this system? You know, what could the negative side effects be of having this in this system? I think those negative side effects could be if this data was ever made public for whatever reason, there could be, you know, negative side effects. Who can query this data? What kinds of questions might we not want being asked of this data? You know, I'm sure, take for example, like, let's say you posted all of your sales results in there, but you didn't aggregate them up at all. It's just a list.