Ep 46: Inside TransUnion's AI Evolution with CIO Munir Hafez

‍Evan Reiser:  Hi there, and welcome to Enterprise Software Innovators, a show where top tech executives share how they innovate at scale. In each episode, enterprise CIOs share how they've applied exciting new technologies, and what they've learned along the way. I'm Evan Reiser, the CEO and founder of Abnormal Security.

Saam Motamedi: I'm Saam Motamedi, a general partner at Greylock Partners.

 Evan:  Today on the show, we’re bringing you a conversation with Munir Hafez, Chief Information Officer at TransUnion. TransUnion specializes in credit reporting, fraud detection, and data analytics.  With a team of over 19,000 employees and over $3.6 billion in annual revenue the company impacts more than a billion people worldwide.

In this conversation, Munir shares his thoughts on AI transforming business operations at TransUnion, balancing the potential of AI against the current hype cycle, and the future impact of AI in the enterprise.

So Munir, thank you so much for taking time to chat with us today. I was really looking forward to this conversation. We'd love to obviously talk a little bit about TransUnion, but before we dive in, do you mind sharing a little bit about kind of your background, your career and how you got to where you are today? 

Munir Hafez: So I started off several decades ago as a software developer. So I spent my first 10 years really kind of cutting my teeth on, uh, hardcore software development in telecom industry. Uh, not to date myself, but the first product I worked on was caller ID. And then from there, I went into the big four, PwC, Deloitte, and EY, spent about 15 years across them doing large scale implementation, strategy, really trying across as I'm moving from role to role, work towards my, Future state or end state where that I wanted to be.

I knew for a while that I wanted to get into a CIO, CTO role where I could have large scale influence. But also, I really enjoy working with teams and helping them grow and seeing them reach their potential. So I had sort of a vision where I wanted to be and I tried to chart, sort of, having different experiences from operational roles, strategy roles, engineering roles, PMO roles, to try and round out my experience to hopefully get me ready to, um, you know, where, where I got to a few years ago as a CIO. 

From there I went back into industry. I was an automotive for a few years working on self driving cars. So that was very, uh, interesting work. You don't always get the opportunity to potentially change society, uh, with, with capabilities like that. Of course, that hasn't quite materialized as everybody had hoped for. Uh, but certainly 

Evan: Going in the right direction. 

Munir: Yeah. So it was very cool technology to work on for sure.

And then from there, I joined TransUnion about three years ago. TransUnion is a global credit bureau, data analytics company that provides data, fraud, marketing, in about 36 countries, I think, uh, around the world. And, um, you know, we have data on over a billion people worldwide. So, very heavy technology footprint, very heavy data analytics footprint. Our organization is probably, 5, 000 plus developers. So, you know, we're more of a tech company than really anything else and, rely on technology, to be cutting edge and be competitive in the market.

Evan: There's a cliche in technology that bigger companies, especially in, you know, regulated areas are kind of slow and like, put all the red tape as possible. So like, you know, nothing can happen, right? 

This side of the spectrum, the cliche is startups are the wild West. They move quickly, but everything is messed up, if not illegal, right? The example you gave is a good example of kind of the balance, right? 

And so, you know, given just the nature of, you know, the, the sensitive nature of your guys’ work, right. It affects businesses, affects a lot of people. There's obviously a bunch of regulatory privacy issues, right. That I know you guys take super seriously. A lot of those kind of thing could be seen as impediments to innovation, but the example you just gave is actually a great example of where there's kind of like a best of all worlds, right?

If you figure out how to innovate to get both, do you mind talking a little bit about how do you kind of create that culture, especially an organization that's been around for a long time when technology is changing so quickly.

Munir: Yeah. So a couple of things. One is I think the partnership with information security organization is super important. So I have a very close relationship with our CISO, Bill Shields. We work very closely together to find that balance, right? Under no circumstances are we going to undermine our cyber security posture. At the same time, we want to be able to do things in a way that is secure. 

So give an example, when I came into TransUnion three years ago, we didn't have BYOD for mobile devices. It was deemed insecure. Technology has moved, uh, along. Microsoft with Intune has provided really good ways to containerize the applications and be able to protect them. And in a BYOD mobile setting, allow that access so you can have productivity on the go. So, only about 40 percent of the organization had a TU phone. So you can imagine 60 percent of the population. We have about 19, 000 people worldwide. 60 percent of the population had no mobile access. So unless they're sitting in front of a laptop, or somebody's calling them, they had no idea what was going on or could be productive for work, right? 

So we worked very closely with information security to find that balance. What compensating controls can we put in and not compromise on the security, but allow BYOD devices in the mobile space. So that partnership, I think, is really, really critical with information security in particular. It's a very special partnership. 

But the other piece, we brought in Frog, which is a design firm originally from Germany, bought by Capgemini not too long ago, to come in and actually interview. And we, we created an advisory board from our associates. Different levels, so from the entry level to SVPs, uh, people who travel a lot, like sales, people who never travel, people who are always in the office, people who are remote, different countries. So we created a group of about 40 people, that had all of the different mixes, and then we interviewed them to understand what are the pain points, what are the things that are an impediment to them moving as fast as they possibly can to be productive. 

And from there, we created a multi-year roadmap for how do we deliver and actually, uh, Gartner did a, did a, uh, use, uh, a study on the work that we've done, in terms of the extent that we went to make it human centric and how do we enable and empower associates to be productive.

So from there, it was really a constant hammering, you know, what is DAX? What is the digital associate experience associated with what you're doing here and here? So, you can shout from the top, right? Culture comes from the top. So you can certainly set that from the top. But where it really comes in is every time something. You do it with your organization is doing something you have to make sure to remind them of the impact of what they're doing. 

An example is patching, right? Patching is super critical. We are rolling out patching with a new tool called manage engine and the way it was interacting with the user, not allowing them to postpone the patching. I'm like, think about that as a user, right? Like if you're in the middle of a meeting and we're forcing you to, to reboot, like that's not a good experience, right? So you really have to have to hammer it time and time again until people start, it just becomes second nature, right? So it's part of cultural change. Change management is always very difficult. So you have to set, it. A good North star, but you have to follow up with the examples. You have to stay on top of it. It probably took 12 to 18 months to make DAX sort of like people having nightmares about them, you know, at night in my team, right? Like everything has to be from that lens because that's what we're here to do. Right. 

Our job as a technology organization is to enable the rest of the business, right? That's what we're here for. And what I've seen in a lot of organization is that it becomes very IT centric, right? And self serving, right? Like this, well, this is the easiest way we can do this. It's the cheapest way. And, you know, this is all the people that we have. So we have to do it that way. It's not really looked at from the end user perspective. And I think that, that becomes really critical to be able to constantly hammer that home. Every time you're rolling something out, what is the impact?

Another example, we did guest Wi Fi, and they were allowing people to create guest accounts. But what about the use case of, you know, you're a user, you're bringing in a vendor, and there's 20 people. You know, is everybody going to create their own account, right? So we had our associates sitting down creating 20 accounts, every one of those could, right, versus create a group account that can be shared, right? So it's, it's, It's the little things that make the difference. 

Yes, you have guest wifi, but is it user friendly, maybe not so much. So I think, you have to pay attention to the little, little, little details and that's what takes a good experience to a great experience.

Evan: We are in like, you know, arguably the, you know, the biggest hype cycle right around AI. A lot of people we've had on the show and a lot of your peers agree with that. They also agree like, Hey, we're getting real wins in some kind of tactical areas. And so there's some, there's like needles in the haystack, right? There's a lot of like nonsense hay. Where do you think we are in that kind of hype cycle, as you sit here in July, 2024?

Munir: I think we're really at the peak of it. Right. So, so we we've started down the path of AI as I'm sure every other company has. And we've started down a few areas. We are piloting, uh, being co pilot as well as Microsoft 365 co pilot, right? And Microsoft 365 co pilot takes really good notes. It gives you action items, summaries, right? Um, at the end of calls. So we're like, aha, why don't we use that to do knowledge transfer session? We don't need tech writers, right? Makes sense. So, you know, you go through it and then it's like, well, it's not, not quite there yet. Right. 

So what we've done with Microsoft 365 is created pilots for six weeks so people can come in, teams can come in and say, Hey, I want the co pilot license. We give it to them. We do training. We have, uh, lunch hours with Microsoft at the table where we answer questions, and then at the end of the six weeks, we give them a survey like, okay, help us understand how did this help you? Right. 

I love the tool. It helps me prepare for the next meeting. It helps me to find things. It helps me to do action items. Is it transformative? No. Does it buy me 10 percent productivity? Sure. Right. 

I think the use cases that have worked really well are things that are around help desk, things that generate documentation, right? The most common one you, you hear is, editors and writers are getting fired, right, left and center, right? Some companies are making big claims. IBM, we're going to fire everybody. We don't need anybody, just AI. I think the truth is somewhere in the middle. I do think there are use cases, where it's really good.

We're doing the same thing. We're using it both for internal as well as for external, but I'll give you an example. Amazon went very heavy on that, right? I've been a long time Amazon user. Used to love the customer support, be able to call in, talk to somebody, get issue resolved. Now it's almost impossible to get to a person and you go into an AI chatbot. Super frustrating. Doesn't really give me what I need quickly enough. So I think 

Evan: And that was like a marquee strength. There's, they were renowned for having like the world's best, you know 

Munir: Customer support, yeah. So, is it good for that? Yes. Is it going to replace people? Not yet. Same thing for me. Like, does it help a tech writer? Yes. Uh, is it going to help me get rid of tech writers? No. So, so I think, um, it really depends on how far the technology.

Now the technology is moving very fast. Um, I do think it also depends on whether you create, um, a lot of grounding data and do your own LLM, right? So the LLMs that you get from ChatGPT, Microsoft, um, Certainly the, the, the ones that are externally facing, right, they're using internet data that are not grounded in your data, right? Uh, when you're using the Microsoft 365 Copa, it uses the Microsoft Graph, so it does have some of your emails, you know, SharePoint, etc. Um, but I do think we've overestimated that it's, that it's ready for primetime. 

Evan: Talk a little about kind of like the organizational consequences, right? And like, you know, just talking about like, you know, even this, even like the. The trade, you know, these kind of GPT training sessions, right? That was not a construct, right? Two years ago, right? 

So if you, as you look through like the next couple of years, let me hear your perspective on kind of what you think Gen AI's impact is going to be on the, the workforce, right? Like, is there, you know, presumably there's new training. There's obviously gonna be new tools. Do you see kind of like new organizational responsibilities? Are we going to have a GPT knowledge manager in the future is that, or knowledge management function, right? In the future?

Munir: So, you know, I mentioned kind of the pilots that we're driving with Microsoft 365 to see whether we can realize the value to justify the cost.

Um, we're doing similar and other use cases in the business, um, in Salesforce, DocuSign and many other tools where we're looking at business processes. We've identified like 88 business cases. Um. We brought in Deloitte to help us because what I wanted to find is what is the use case that's going to provide me the best value. Number one. Number two is where is the technology? My team supports over 500 applications. Everything has AI, including toothbrushes and trash cans has AI in it. So which which tools are actually far enough along that it provides value versus Everybody has AI, right? 

So we were looking for the triangulation of the technology is advanced enough. The business case makes sense and the business users are engaged and are willing to learn and help support. So we've triangulated on 88 use cases. Obviously there's still some skepticism in the organization. We said, okay, what's the one use case that we think we can bring to market in a couple of months, where then we can get testimonials from the teams in terms of the value that's added.

So we're looking to basically do a proof of concept so that we can start to generate excitement around what this tool can do so that we can start to reap the benefits of it, right? Start small, prove it out, and then start snowballing it, right? So that's kind of the approach we're taking. 

I do think prompt engineering is going to be real. I also think the LLM and the grounding of LLMs and having custom LLMs for the organization is how most organizations are going to get the most value out of generative AI. So I do think there's going to be centers of excellence, whatever you want to call it, right, that are responsible for bringing in the data. You said HR data, engineering data, whatever it may be, put it into an internal LLM, train ChatGPT on it or whatever solution that you use, and be able to get the most value. So I think there will be change management, where you have sessions to be able to train people, educate people, particularly on how to do the prompts and get the value, and then a central organization potentially, or a COE type where you have people matrixed to be able to create and maintain the LLM so that you can get grounded with as much data that's custom to your organization so you can make the answers that it comes up with the most meaningful for you rather than more generic answers.

If you're asking it how to make a chocolate chip cookie. You don't really need to ground that in your data. If you're trying to say, Hey, what cross-sell opportunities, can I have for this customer? You have to be able to feed it your product catalog and everything else. Right. 

Um, I do think some of the tools have need more grounding and more support, or you can get more out of it. You know, copilot, uh, Einstein Copilot and Salesforce, right. Uh, can help you pre-write emails and things of that nature, but you can also expand the prompts, right? But that takes a team that's going to do that, right? A team that understands, okay, if somebody is asking this question, then you know, here's the type of data that I need to ground, right?

So, uh, I think over time, it's going to get smarter and smarter where you need less and less grounding perhaps, or less, uh, prompt engineering that's teaching it. But certainly in the short term, I do think training, custom LLM, and then depending on the tool, like Salesforce, uh, Einstein co pilot teams that are designed and focused on extending the prompt engineering and the capabilities, um, will be the best way to, to get the value.

So I don't think, certainly not in the very short term, you're just going to plop in ChatGPT and voila, you're going to get rid of 50 percent of your population. Um, at least in my mind, unless they're all sitting, writing, writing articles all day long. 

Evan: One challenge, a lot of, I think organizations struggle with including our own organization is kind of like, when do you like the technology evolving so quickly and like there's new, you know, there's new startups like building these really cool things. You got kind of the bigger platforms, you know, doing things a little bit slower, but like they're making new things, uh, or making, you know, new enhance, enhance, you know, AI capabilities. How do you kind of know when to like build, when to buy, to partner, right? In an age where like, again, every week there's a new technology, right?

We've had a couple of new, I mean, even take the, You know, I know maybe it's kind of like more your bearish case, but on the customer support example, like today, right, you know, I'm sure your team go hack together your own data, right. To, you know, build, you know, fine tune a model on a ChatGPT, you know, take a fine tune, ChatGPT model, or open AI model. You could build something. You could go to like some new startup that does that, like as a very focused thing, or I can wait for sales for a service now to kind of mature their thing. 

And so like, today, right, the answer may be obvious, but when we've kind of, when you look through, like, what's the right decision to make today, based on where you think we're going to be in two years, you know, what type, you know, what type of use cases or what, what conditions in the environment would cause you to take one path over the other?

Munir: So there's a couple of things. so mentioning TransUnion's high degree of sensitivity across our cyber security posture, etc. So, for example, we can't have any LLM be trained on any customer data as an example. 

Evan: Makes sense. Yeah. 

Munir: So the challenge that you have with a lot of these products. To be used in a public company, for example, uh, they may not have SOX compliance. They may train their LLM based on your data, right? 

So, generally, right? In my mind, things that are competitive advantage, you build things that are more commodity. You buy, right? Nobody's going to build their own workday HCM, right? You, you buy that, right? 

Evan: That makes sense. Yeah. 

Munir: um, for our internal, we use Converso, uh, a product, uh, that has GPT integrations. We hooked that up to our SharePoint sites. We hooked that up to Converso so it can consume that data, and then when somebody is asking, they can give it a better answer.

So in this space, certainly in the foreseeable future, I would look for, kind of the, buy the base and build the extensions, build the LLM and build the prompt engineering, I think there's going to be a bit of marrying of those two. I think as the technology gets better, you'll probably get more and more of that off the shelf, but certainly in the short term, I would see, uh, buying that and then being able to extend that with things that you build internally.

Now, this is very much, kind of the enterprise IT position, right? TransUnion on the product side um, has built a lot of generative AI capabilities and has for quite a while. I'm not going to speak for that side of the house, but certainly on the enterprise IT side, the buy versus build, we generally err on the side of, we buy it unless the cost factor starts to come in, but generally in enterprise IT, it makes more sense to buy and then extend.

Evan: I'm just curious about the kind of the use cases. You're more, you know, AI use cases. You're more bullish or bearish about, especially relative to your peers. Is there anything like, you know, if you met one of your kind of, you know, fellow CIOs, right. If you know, Fortune 500 company, like, what would be the thing you'd say? Like, Hey, if you're not thinking about using AI to this, you know, for this use case in the short term, you're kind of missing out.

And is there another side of that where you're like, if you're trying to apply AI to this use case in the short term, you're, you're kind of, you know, it's like, I just, it's not a good use of time right now. Like what would be kind of when you're like must do and don't do list? 

Munir: I think on the most do is activities that have a lot of, either text generation or consumption of text. So the customer service example, or generating a draft legal document, reviewing a legal document, summarizing documents, right? Sometimes you get 50 pages and you need to summarize the key points, right? So I think anything that, that it's really good at, right? Which is probably why a lot of what you hear out there is, well, it's generating stories and all editors are getting fired everywhere, right? Because it is pretty good at generating text, as well as consuming texts and giving you summaries. I think it is super on that. 

In teams, it does a really good job with Microsoft 365. Sometimes it's amazing. Like when it gives you the summary and the action items, it's fantastic, right? Like, you can just send it out and it's easy, but I think the other use cases where maybe there's a lot of grounding and there's still a lot of human thought that needs to go into the creativity of it. So would you want Salesforce to pre generate an email for you? Sure. That's going to help you. Would you want it right now to just go send stuff to customers without your salespeople interacting with it? No, you probably don't. Right. Will it eventually get there. No doubt it will. But I think in some of these cases, it will be more of a productivity enhancer, right? 10, 15 percent 20 percent not the I'm going to replace a lot of people. I do think there's going to be, you know, some highly manual repetitive work, accounts payable accounts, receivable, matching POs, bills, those kinds of things. Image recognition, extracting content, comparing, I think it does really, really well, on that type of, uh, use case. 

Evan: If you kind of imagine where the world might be, you know, in the three to five year timeframe, what are some of the ways you think that, or maybe what's the most exciting way you think AI is going to transform, you know, how you operate the business, the TransUnion that, you know, you're, you know, you're just excited about, or you think could have a major impact.

Munir: I think that over the next three to five years, it's going to go from a helper of developers writing test cases, documentation, certain functions to writing entire capabilities. I do think in the next three to five years, it is going to exponentially improve, um, and it's probably going to. But maybe that's a bit bullish three to five years that I wouldn't say that developers would be out of out of business in three to five years. But I do think it is going to go from a productivity tool to a massive boost to development efforts, or to quality of applications right chaos engineering right fighting fighting errors in in software that humans, in normal testing cycle will take too long. So I think in those kinds of scenarios, uh, I think AI will get really, really good at in the next three to five years where it will replace a lot of manual, uh, activity today.

Evan: Okay. So I know we got limited time and, maybe, with our kind of last five minutes, love to do a lightning round. So looking for kind of the one tweet answer here. So, first question, how should companies measure the success of a CIO? 

Munir: Um, how much they're able to expand margin. 

Evan: How should a CIO best position themselves to best collaborate with the rest of the C suite? 

Munir: I think you've got to build yourself as a trusted advisor. And to do that, you really have to understand the business, and build the right relationships. 

Evan: And so maybe switching gears to the personal side, what's a book that's had an impact on you? I'd love to hear what it is and, you know, why? 

Munir: I'm a big fan of Malcolm Gladwell. Um, he, you know, he, he has, um, uh, I'm so bad with names, but I, uh, I had outliers and he had several other books, but Outliers was one of the ones that, um, I've read, I reread recently. I, I read it when it first came out years ago, but I've, I've reread it recently and it's really interesting in terms of the lessons that it teaches you and how much in life is actually outside of your control.

Evan: What is an upcoming technology or what is an upcoming new technology that you're personally most excited about? 

Munir: Yeah. I mean, it's a, it's a bit of a trick question, right? Because you kind of have to say GenAI, right? Uh, I mean, uh, it's the, the most hyped technology out there. Um, so I, I think that's going to hog the airwaves for a long time, just like digital transformation did for years and service oriented architecture did for years, right? There's always these, you know, the big buzzwords for a couple of years that just kind of hogs the airwaves. I think it's, it's hard to say anything about Gen AI probably for the next couple of years as that then starts to get down into reality. Right. And we actually able to realize the value from the tools. Right. 

Evan: What do you believe will be true about technology's future impact on the world that most people would consider science fiction today? Like I'm looking for your kind of contrarian view. Like how do you think technology will change the world in ways that maybe other people don't?

Munir: I think technology and going back to my days of working self diving cars. I think one of the things that really excited me about how that can be transformational to society. Think about somebody who's disabled. Think about somebody who's elderly that has lost the ability to be able to, um, have freedoms, right? Going to going to a grocery store or what have you, right? So there's a lot of scenarios, cases, um.

50, 000, uh, deaths every year from, from, from automotive accidents, right? So I see technology, certainly in an automotive space, I'll use that as an example, because that's one that, that is near and dear to my heart, having worked on it in several years, that can really save lives, as well as empower Uh, population that, uh, today is not able to take advantage of the freedoms of having a car that most of us to kind of take for granted.

Evan: That sounds like an exciting future. I imagine it's going to be, uh, probably even more crazy than we think. So, uh, really appreciate you making time to chat, chat. Um, hopefully get a chance to chat, uh, talk in soon and, uh, yeah, appreciate making the time. 

Munir: Thanks Evan. I appreciate you having me on.

Evan: That was Munir Hafez, Chief Information Officer at TransUnion.

‍Saam: Thanks for listening to the Enterprise Software Innovators podcast. I’m Saam Motamedi, a general partner at Greylock Partners.

‍Evan: And I’m Evan Reiser, the CEO and founder of Abnormal Security. Please be sure to subscribe, so you never miss an episode. You can find more great lessons from technology leaders and other enterprise software experts at enterprisesoftware.blog.

‍Saam: This show is produced by Luke Reiser and Josh Meer. See you next time!