The AI Optimist: On AI in the Enterprise, Technical/Non-Technical Leaders, and Human Flourishing

Khullani Abdullahi (00:01.646) Hi, and welcome to the AI in Chicago podcast. I'm your host, Khullani Abdullahi. I am the founder of Techné AI, an AI governance consulting literacy and enablement training firm based here in Chicago. I live in Wicker Park and my office is in Fulton Market. I'm very excited to announce today's guest, William Benjamin. He's a strategic AI leader with over 20 years of experience, both research and on the development side, building enterprise scale artificial intelligence initiatives across multiple sectors and finance, insurance, manufacturing, and transportation. Currently, he's a principal generative AI machine learning expert and works with large teams deploying advanced AI solutions. including large language models and agentic frameworks. Principally, he works on the cost savings and operational efficiencies, value use cases of large language models. He has founded AI centers of excellence, has multiple patents, and mentored high performing engineering teams. Dr. Benjamin William, thank you and welcome. William Benjamin (01:23.162) Thank you so much. I'm very excited to be on this podcast. Khullani Abdullahi (01:27.786) Excellent. So you have a PhD in computer vision and machine learning from Purdue University. Can you share with us why you went into the field? And when you went into the field, did you know or anticipate kind of the deep learning revolution that would then actually take machine learning from this like very research based universe into being embedded in everything and being like the number one driver of value in the U S stock market, right? That transition period was very, very quick, right? So I was a transformers, paper from, Google out when you were doing your PhD, walk us through kind of that academic journey. William Benjamin (02:04.953) Yeah. William Benjamin (02:17.956) Yeah, definitely. So yeah, people are surprised. My undergrad is actually in mechanical engineering. So the revolution, when I joined Purdue for my PhD, my research kept getting pulled into machine learning. I took all my courses where computer science and machine learning based and always felt this pull. And you could feel the energy where the field was moving really fast compared to all the other fields. And that really made me like sense that this thing is going to be great once I graduate. And that's how my research sort of really got started. And during my research, again, I think there was no one moment, but I could always feel the energy, the way the research was progressing. Like every time, especially when deep learning came in, like the first time I implemented neural network, and built this code which actually for project where it was deployed onto a self-driving car and it was moving that automatically recognizing signs. I was really feeling that energy. This was like 10 years ago and now we are seeing Tesla, Waymo out and it's really, you can feel like, those dreams are really coming true. And then of course, like with Chad GPT, I think... Khullani Abdullahi (03:32.046) Yes. William Benjamin (03:45.936) I was a little late to adopt, but I think the moments came when, for example, when you start doing wipe coding and the agents just autonomously act like a collaborator, not just like you're querying an engine, but it's like telling you what to do, like proactively and solving problems for you. So you are in the designer seat now instead of being the coder. So that really, I would say, Khullani Abdullahi (03:57.965) Yes. William Benjamin (04:14.766) made me a full believer and like made me drop everything and jump into this field with like full energy. Khullani Abdullahi (04:22.808) I love that. So there has been in the digital spaces, I won't name a social media platform unless they're advertising on my podcast, but in the digital social media spaces, there has been an ongoing debate about the continued value of advanced degrees even in computer science. So I don't know if you saw, but even computer science majors are having a very hard time landing some of those entry level. software engineering roles, whether they're machine learning or specific or not. Based on where you sit in the marketplace as an AI leader in the enterprise sector, if you were deciding on the study, taking a pause to go get a PhD in machine learning, what advice would you give yourself or somebody who's considering that? Does, does a PhD in computer science still make sense? on the eve of chat GPT-5 and Cloud5 being dropped and with their coding abilities, especially as they increase their ability to be a machine learning engineer themselves. William Benjamin (05:33.912) Yeah, that's I think a very deep question. I've thought about it a lot and I've come actually on both sides. On one end you have like with PhD, you get, it's not just the research that you do, but the ability to think and solve a problem like learning about different new areas and combining that to really build something that is novel and make it like a tiny dent in the research universe. But the field is moving so fast. It's amazing how you can start a billion dollar startup company within, with like one or two people and still be successful. So given the time it takes to do a PhD, would give myself the advice like, hey, learn by doing like code is the truth in the end. Like if you're able to code something and build it out. You learn a lot just through building those things out. And learning is also much easier with ChatGPD. You can ask the right questions. So I would probably learn myself at this point. Khullani Abdullahi (06:51.084) Yeah. Speaking of learning, so I talk to lot of executives and here is, I think, the hidden, not so hidden reality. The hidden, not so hidden reality is most leaders have a full-time job unrelated to AI. Simultaneously, their C-suite and their boards have said, what is our AI innovation strategy? What is our AI implementation strategy? What is our AI integration product operations? What is your AI blank strategy? so executives who have a, you know, they have a P &L responsibility already in some non-AI related business function are now required to connect the dots with AI. And that is requiring them individually outside of their traditional work to upskill and to educate and to enable themselves very, very rapidly. How are you, even though you work in this space, as you said, there's a thousand machine learning papers published a week, right? There's a conference every other day. There's new models dropping left and right. William Benjamin (08:00.751) Yeah. Khullani Abdullahi (08:08.206) How are you staying abreast of the latest developments in machine learning? What filters are you using to decide what is and is not valuable? Do you have dedicated time every week where you're reading or watching videos or taking courses? How are you also enabling your teams to do that continuing education? Because I think people are tactically feeling overwhelmed with information overload and there's a lot of uncertainty about how do I find and where do I spend the time to continue to increase my own capacity in this arena so I can then turn around and apply it to my specific role. William Benjamin (08:51.608) Yeah. Yeah. I think I'll break it down into two parts. As you asked me, like one is what's my own personal strategy and then on an enterprise scale or within my teams, like how should we look at it? So I think on my own end, as we discussed earlier, I just became a believer in this technology. Khullani Abdullahi (09:18.36) Mm-hmm. William Benjamin (09:21.136) The PhD also, it's basically a training where you learn how to read research papers. So I try to stay abreast with the latest research. So I have some time carved out every week, probably around three to four hours to read and stay on top of it. But I really believe in code is the truth kind of idea where... Khullani Abdullahi (09:27.022) Yes. Khullani Abdullahi (09:40.791) Okay. William Benjamin (09:49.162) I think Andre Karpathy, he's a great teacher. He talks about like, if you're able to like, yeah, build out something, like nobody can question that. So I'm right now working on like just automating my own workflow, like my data science, machine learning, data analysis. I'm trying to build an agent tech system which will automate this data analysis for me. Khullani Abdullahi (09:53.976) Yeah. William Benjamin (10:16.75) And while doing it, you learn, that learning is much deeper than the, what the AI influencers are talking about. I think once you are firmly grounded on the basic concepts, then all the new stuff is just like incremental additions and you immediately get the crock-sock, okay, this was some missing functionality and MCP, for example, fixes that connectivity or A2A agentic protocols. Khullani Abdullahi (10:25.631) Yes. William Benjamin (10:45.936) it's gonna add this additional. So you just build on top of, get your, I would say basics really strong and then any new learning is just an incremental add. So you're not like constantly trying to keep up with everything. So yeah, that's my own strategy. And then I would say for my team, as well as a greater team beyond the team I work with, We have like a, I talked about like a center of excellence idea where you have a group of SMEs from different parts of the company, as well as experts in generative AI who gather like once a week and do discussions on some latest papers, as well as do it in an applied sense. So there's a lot of research, but let's talk about what can be applied to our current problems. And that's how it gets you started to think about it. And then if it's a really interesting idea, then build a prototype. there's a lot of hands-on ideation and discussion that really gets people excited. And then they work. If you're excited yourself, then you'll go learn, figure it out on your own. Khullani Abdullahi (11:42.99) Mm-hmm. William Benjamin (12:11.066) So it's not like a mandated thing. Khullani Abdullahi (12:14.062) And then doesn't feel burdensome, right? Once they realize the value that they can derive from it. Let me ask you this. there's, you know, people think that I hear this phrase often, especially among non-technical executives. Well, I'm not technical. So how deeply can I potentially understand AI? William Benjamin (12:17.113) Yeah. Khullani Abdullahi (12:39.874) But I agree with your earlier point that true comprehension in this universe means using these tools, right? Like you have to have a very like almost intuitive feel for the technology, which only comes from repeatedly like interfacing with it. So even at my level, I'm very intentional. William Benjamin (12:51.248) Yeah. Yes. Khullani Abdullahi (13:04.02) about doing vibe coding, right? Like I could easily, when I started this new consulting company, I could have easily outsourced building my course website and my website and my newsletter, all of that. But I don't need to anymore, right? I don't need to hire a designer and a backend developer and a front end developer because I know how to get a require. I know how to put together a product requirements document. I know how to have. William Benjamin (13:16.666) Yeah, totally. William Benjamin (13:29.231) Yeah. Khullani Abdullahi (13:29.826) you know, chat, GPT or Gemini, improve it. And I know how to have Claude develop the code. And I know how to go into GitHub and like launch a GitHub pages. Right. And, it's faster that, and I will tell you most people with a philosophy and law background who are in strategy aren't launching their own GitHub sites. Right. But I can't translate AI capabilities. William Benjamin (13:39.833) Yeah. Khullani Abdullahi (13:55.938) to the business strategy, unless I know what is this technology, how does it work? Where does it operate? Where can it be used? How do I prompt it? How do I interface with it? And then go deeper. I share that to say, I've come to the understanding that everybody, all senior leaders have to develop some technical fluency of this technology in a way that they didn't have to go develop for IOT, right? William Benjamin (14:02.127) Yes. Khullani Abdullahi (14:24.482) They didn't have to understand there are plenty of executives who have no idea how to go into Azure or AWS or Google cloud. have no, they have no notion of navigating it. have no notion of an API and what it means. They know Zepierre, but they don't understand that it's an API technology. I don't think they can remain as distant in the past as they were in the past from. William Benjamin (14:38.724) Yeah. Khullani Abdullahi (14:53.162) APIs today, which are going to be the backbone of, of agents and model context protocol, right? Like they have to get their hands a little bit more dirty. Are you finding that to be true in the enterprise context? And, and how are you partnering with executives to kind of bridge this business technical gap? Or do you think, like, do you disagree? Do you think maybe that they can remain? a little bit more distant and continue to speak about it in generalities, but without having like an intuitive understanding of using the technology, like on a day in and day out basis. William Benjamin (15:32.704) Yeah, that's a great question. think, let me start with the preface, right? Compared to all the other technologies, generative AI, I feel, is almost more of a revolution in UX, like the user interaction, like the bandwidth that we had. I mean, why do we need software engineers is because the bandwidth of communication is through code. Previously it was assembly code. Now, then they went to C++, which was a higher level language. And now people love like working with Python because it's even easier to code. And now you can use natural language. Khullani Abdullahi (16:12.014) Yeah, the abstraction level just going up and up and up, right? Yeah. William Benjamin (16:17.048) Yeah. But like it's increased the bandwidth. So anyone who can talk English or you don't even need to know English, like it's bilingual too. So with language now you can prompt the model and it will generate code, generate websites. So it's, I would say given the benefits that are there and the ease of this interaction, think there is no reason not to do this. where you can really get a sense of what is the value of this product, even in different parts of, depending on where you come from. For example, your area, security, law, and then, for example, with marketing, there's tremendous possibilities. And it would really help knowing how this technology works. So you know the ins and outs too, especially from the perspective of like, okay, what are the limitations? Otherwise, it's very easy to get carried away by the hype. But when you actually play with something, you know, you understand the intuitive limits of the technology and it will really help in building robust systems which leaders understand. Khullani Abdullahi (17:27.405) Yeah. Khullani Abdullahi (17:43.79) In 10 years, do you think that there will be any senior leaders who don't have technical skills? Do you think that they will be relevant in 10 years? William Benjamin (17:55.76) Yeah, I think we are at, I would say, early, similar to internet when we were in early 90s when there were people, even I didn't have internet until my school days. And then now it's like impossible to think about someone without internet. We are at the cusp of this technology. are just figuring out like back in nineties, like you were trying to figure out, okay, what can, internet do? And they made these direct, like they were just transforming whatever was available, like directory structures and made it into websites. And nobody really thought about like apps and the idea of like cell phones. And now it's like impossible to imagine. So I don't think there will be anybody. it will just be. Khullani Abdullahi (18:37.741) Yes. Khullani Abdullahi (18:43.83) Right. William Benjamin (18:50.924) another layer that we interact with in the next 10 years. Khullani Abdullahi (18:54.061) Yes. I think that's fascinating. I also think, I think you and I are probably in the same generation. I remember when internet came onto the scene. Like I remember when cell phones, right? I remember when YouTube was launched, right? Like I remember those. So I'm familiar and like aware of those like technology fluctuation points. And I recall that there are people at each of those fluctuations that were left behind, right? There are attorneys that hated electronic filing of cases. There are now attorneys that have no idea you had to hire a messenger service to hand deliver your filings to courts around the country. That was just a role. And so now everything is electronically filed, but there are people who didn't develop that muscle. I suspect similarly, there will be people who don't develop this muscle, right? Like I've done trainings and people have asked me, like, how do you prompt it? And, their approach is very transactional. They ask a question and they get an output and then they just stop. And the recursive back and forth, digging in deeper, adding in layers, providing context, asking questions like, What am I missing? How would you improve this? Those epistemic humility questions, those mental models, I don't think those are things that you can easily transfer and teach. It's almost like, you a native post-AI leader or are you not? And that dividing line is... William Benjamin (20:34.82) Yeah. Khullani Abdullahi (20:44.43) has yet to be determined, but when it is determined, there will be clear people on one or the other side. And from a labor standpoint, like, do you see as a leader in enterprise, do you see these technologies impacting how you allocate budget to headcount? And if so, how is it, how are you thinking about budget implications, labor and headcount implications? and similarly, if you were to invest in a new human being who's a recent graduate, what skills are you looking for that would continue to be meaningful and valuable when you have, you know, agents that are powered by Chachi VTA? Like who would you still need to hire? Who do you think would still have some relevance? William Benjamin (21:37.164) Yeah, I'm in the camp of actually human centered AI. the technology, like all technology, right? Like you can use it for good and you can use it for something that's not good. technology should be built for human flourishing. In terms of like... Khullani Abdullahi (21:44.045) Okay. William Benjamin (22:05.366) looking at AI, you can actually look at it as making you 10x productive, right? So you can do a lot more now with the power of AI. So in that terms, given, let's say someone who's already there, I think there's a big question of like, what are folks going to do? And that's where I think re-skilling and really helping people understand the power of AI and then really helping them. at different stages, there are people who are early adopters like you, you got in like the beta release. And then there are people who are skeptical, but the beauty about AI is it's a much, it's basically a UX revolution. So it's like, it's very intuitive when you design, let's say interfaces, they need to be intuitive. And this is one example where the more sort of research we do, for example, there are voice agents now which sound like human and you're just basically have, we could be in a podcast and you could be talking to an AI clone of me and it really sounds real. So the interfacing with AI will, I would say is becoming less and less challenging. So as I said, there shouldn't be any reason why you don't upscale. depending on like even your level of technical ability. think all humans have inherent skill sets and inherent creativity, which cannot be replaced by AI. So you could end up being the manager of like three or a team of AI experts, but you're driving and we should sort of build AI responsibly so that we are Khullani Abdullahi (23:41.899) Right. William Benjamin (23:58.444) not dehumanizing people, really bringing out the value of people, their skill set, their curiosity. You brought up, for example, a young graduate. mean, we could have a young graduate who's very curious and is asking the right questions. Now you have the power of AI to augment their capability and do a lot more. So I feel AI should be... built to empower people and that's what technology should do at the end. Khullani Abdullahi (24:34.936) So I love the philosophy that you just articulated and I love that you framed it as a choice, like we should, because it is a choice, right? And I think there's two competing, several competing philosophies and competing interests. as we're building AI and there is very much a position where people are working to just gain efficiencies, reduce costs, increase revenue, irrespective of the human impact, whether it's at the individual, organization, society, country or global level, right? And then to have positions like yours articulate And to say that as an AI leader in an enterprise setting to say that we should be building to augment and extend human capacity and creativity, I think is a very powerful statement. And I hope that that philosophy and that position wins, right? Like in the marketplace of ideas, but I think it's going to take work, right? Like we are going to have to articulate and say, We want this technology to work for us in our communities and our societies. And this is how we think, excuse me, this is how we think we can do that by putting humans at the center. So I want to switch gears a little bit to how you ensure alignment, right? And so McKinsey had a recent report on... William Benjamin (26:03.267) Yo. Khullani Abdullahi (26:14.67) the very high failure rates of AI initiatives. And that's not surprising because large scale change management and technology implementation initiatives, even pre AI always had a fairly high failure rate, right? And so you layer on AI and of course there's still a number of challenges. So you are leading major. Generative AI initiatives before generative AI, you're doing predictive analytics. I'm assuming at John Deere because the big data was a big thing. so you're doing large data sets. You're doing cross-functional teams. you're doing different business functions. What are some like very tactical practices and activities that you find effective in keeping? William Benjamin (26:46.746) Yes. Khullani Abdullahi (27:05.408) everyone aligned so that they can actually ship impactful AI solutions at a cadence that meets business objectives. What are you doing in a day in and day out basis, tactically? Some of those strategies, a lot of leaders are in your position, but maybe haven't gone through a decade of shipping products and solutions. Do you have some best practices that you've developed and some go-to strategies to just keep everyone aligned? William Benjamin (27:39.47) Yeah, that's a great question. And I think that's kind of the central learning that I got actually during my PhD. So most people do the product building like they do it in the opposite way where it's the industrial model. You build a product first and then throw money on advertising and hope people start using it. Yeah. And that's where I think the statistic comes from. Khullani Abdullahi (28:01.826) Yes, and push-up. William Benjamin (28:09.346) let's say what Mackenzie is talking about, all the AI products. There's usually a buzz, like maybe 10 years ago, it was deep learning, now it's AI. But what people forget is like the problem solving aspect of it, identify what's the problem, who are the people affected, and then solve that. So there's a framework that I really love, and I learned it during my PhD. My professor was actually a Stanford graduate and Stanford is big on what they call design thinking. And it's developed in the design area. And since I'm actually from mechanical engineering, that really helped me apply this to AI now. The underlying idea is basically you empathize and understand what the real problem is. And there's research which is shown like, if you solve the right problem and identify it, Khullani Abdullahi (28:44.12) Mm-hmm. Mm-hmm. William Benjamin (29:06.656) early in the design stage, you save a lot more in terms of rework and redesign much later and advertising much later. So if you get the product right and getting the product right is what? Like knowing your customers. So really empathize, spend time with your customers and really define the problem. What is the problem that we are trying to solve? Is it like the friction when let's say a call comes in and let's say a representative is just looking up the context instead of like listening to the customer. So really identifying where the problem, where the friction is. And then ideation. So the third step is ideation where you basically use ideas like paper napkins as much, throw in as much as possible. It's a cycle of like convergence and divergence. So you're at divergence where you have a lot of ideas. identify the correct idea which could help and then do again, so that's the convergence and then you do prototyping, converge on one idea and then build out again, diverge, build out some prototypes and see how your prototype works, test it. And then second part again is alignment. where getting alignment is a lot easier when you have a prototype working. So early iPhone, they just built like very, they didn't build the full thing, but they just built simple prototypes, gave it to people and asked them what it felt like using it. And you get early feedback. That's the most valuable thing because it helps you remove the noise and build the right product. And once you have the right product, people will come, like people will buy it. Khullani Abdullahi (30:46.35) Right. William Benjamin (31:00.82) Once you get a loyal set of customers internally or externally, that's what has been my experience, they will advocate for your product. You just bring them in into meetings and they'll be like, yeah, we love this. Like it has helped us save money or save our time, improved our efficiency. And that's how I've, that's the, the, would say my secret sauce. Khullani Abdullahi (31:12.419) And William Benjamin (31:29.104) to getting like alignment. The second or the third thing is building relationships. That's something I realized after my PhD getting into the industry. In PhD program, everything depends on you, your caliber, and how fast you can do research. But then in the real world, it's all about community and building relationships both with senior leadership Khullani Abdullahi (31:31.585) Yeah. William Benjamin (31:58.34) where I'll have one-on-ones with them, where they can actually, it's more of like helping them. You're not really trying to sell something, but like asking them what, how can I help you? And that's when the real problems are discussed. Similarly, like with my team, understanding like where they are getting blocked and removing the major blockers, acting like a coach, that's really... I think helped me get the trust from leadership as well as my team to help build out great products. Khullani Abdullahi (32:35.448) Do you find, there a specific mental model or like just interpersonal approach or mind state that you have found to be particularly helpful in being an AI first leader that you think other leaders might consider adopting themselves? William Benjamin (33:02.192) I would say being AI first again helps me in being able to sort of have the credibility to say when I say something, they know it's backed up by years of research or understanding of the field, definitely. But on the other end, I think it's basically the kind of gravitas you can carry. If you don't have the sort of, I'm more of an introvert, so I work better one-on-one in one-on-one meetings. So I set up a lot of one-on-ones on my calendar with leaders. But if you are the kind of extroverted leader and you can build a compelling vision for the team, I think that's also pretty, very helpful where you can develop a vision, inspire people and really get the senior leadership excited about the possibilities and the ROI behind these products. Khullani Abdullahi (34:10.882) I love that. So a big topic, a subtopic in the AI world is governance, policy, risk. And it's an area I'm very familiar with because that's where a lot of my work is in. A question that I often ask prospects and clients is at what level are you beginning to have, at what stage of the product development life cycle? Are you considering governance, risk, compliance, model maintenance, et cetera? So as you are building proof of concepts and scaling in large enterprises, where do you, when do you start having those conversations? What are those conversations like? What roles are important for that? And how are you embedding, embedding responsible and safe? to use cases during the development of these AI solutions. William Benjamin (35:15.148) Yeah, I think depending on the company, I would say it really matters. For example, if you're selling insurance, you are basically the product is reliability and dependability. So, and especially, again, learning from design, from the design process, the earlier in the design, you make a change and get the right change in, it's a lot. cheaper to make that change compared to later down the line when the product is in production and now I'm trying to add a layer of security. So early in the design, if you could include governance and security and put in all the processes in place, the more I would say it helps and it's sort of a built-in process. I call it like My framework for basically building any product is three things, right? Like we talked about alignment. The second thing is data infrastructure. You need to have the right one. But the third one is governance and security and building in the privacy into the product right from the start. Because that's where then you kind of assume that someone is going to break the product and then you make those early design decisions. based on that. Khullani Abdullahi (36:46.124) It's almost as if you have to be more defensive in the development of AI solutions than you did in traditional software solutions, right? Because the stakes and the potential negative consequences are so much higher, companies and I think engineering leaders have to build it as if someone is gonna adverse, like the adverse, William Benjamin (36:59.919) Yeah. Khullani Abdullahi (37:15.64) prompt injections, And so, and I feel like that that's new and different. So if I hear you correctly, there's like three pillars, right? There's data and infrastructure, there's alignment use case and design thinking development, and then other infrastructure for new product AI development initiatives is governance, risk, cybersecurity, safety. Okay. I love that. It's a great framing because I think, William Benjamin (37:16.162) Adversarial, exactly. William Benjamin (37:38.222) governance. Yes. Khullani Abdullahi (37:44.992) It means that when you're doing your product road mapping, you need to address all three of those pillars as you're kind of laying out those stages. I, you know, I won't say it's a bad thing because it's why I have clients, but there are certainly clients who have built it and said, Hey, hi, Kumani. Can you help us develop policies? William Benjamin (37:51.258) Yes. William Benjamin (38:10.53) Yeah, it can be done, it's expensive, yeah. Yeah. Khullani Abdullahi (38:10.84) for a fully film solution. It is expensive and it is also very time consuming, right? Build it right from the beginning. I have, there are companies that have full product AI solutions that are like, we need a model card. The solution is deployed, right? Like the audit needs to happen from like... William Benjamin (38:19.941) Yeah. William Benjamin (38:25.754) Yeah. Khullani Abdullahi (38:35.542) the inception of the product and now we have to work our way backwards and it is a much more difficult process. So I love this three pillar framing and I may point people to you and say, here's how an enterprise leader is ensuring that this gets built in from the beginning. So I love to think, want to talk about balance and how you're balancing a few things. So a lot of leaders are struggling with prioritizing AI solutions. to accelerate the productivity and optimize the workflow of employees, especially in large enterprise organizations, right? The thinking is, if we can get 25%, right, workflow optimization or productivity efficiency gains in an organization of 20,000 employees, that's gonna be very significant. Versus, how do you prioritize building and integrating AIs solutions into your products and services for external customers. So in many organizations, that's not two different teams. That is a single organization that shares resources, engineering strategy, product, et cetera. So when you are an enterprise leader and you're thinking about what is our AI innovation strategy at the 30,000 foot view. How are you thinking about and how do you recommend other leaders consider thinking about the trade-offs between prioritizing initiatives in internal employee efficiency, workflow optimization, and building those solutions, proof of concept, scaling them, and or simultaneously or maybe consecutively building solutions to increase revenue? directly by integrating it into new products and services. William Benjamin (40:38.316) Yeah, that's a great question. The framework I use is basically identify what's your biggest part of your expense or revenue, Like John Deere, it's mainly about operations, like the whole supply chain of manufacturing. So any AI that could really help improve efficiency, that has a huge impact on your bottom line. So I would like basically break it down and see where do we have the biggest bang for the buck for improving the operations. But if you're, say like here, technology is a software company that where you could quickly prototype a gen AI product and people are like already trying it out and you get early feedback and you could go build a totally new area of products. then there's a huge value there. So depending on where you think is the biggest ROI. The third thing is R &D. Just as a technologist, think R &D is like the fuel that is, it feels hard in the short term. It's like, what are we doing? But if R &D is harness really well, it can like really push you like. over let's say multiple years, it can really push the innovation and your leadership in this field. Like for example, Google has been a monopoly in search for so many years. Just recently now, there's perplexity trying to challenge Google. But Google really pays a lot of attention into like the innovation and hiring the right really smart people. So. The R &D budget, would say, should probably be around 10 to 20 % of the time that you're spending versus depending on if you're heavily operations focused, maybe you spend more time on improving the operations. And then products would, again, give you more market share if you build the right products. Khullani Abdullahi (42:59.99) It's very interesting because I think I haven't heard or read anything recently, I don't think, where someone suggested what you're suggesting, which is to have dollars and labor allocated. to R &D where you didn't choose whether it's for internal or external purposes, but you're doing research and development and innovation for its own sake and giving them like a green field to kind of almost freestyle. And I think that's a very powerful statement because a lot of companies have a clear deliverable and end state in mind. William Benjamin (43:34.554) Yeah. Khullani Abdullahi (43:48.47) when they are building centers of excellence or AI innovation centers, right? They have a very clear mandate. So you are supposed to do X percentage of efficiencies, launch X number of products. There are very few organizations. And I think this is a testament to, think, whether it's a tech-centric or non-tech-centric. companies, so like in the tech, I think in the large tech companies, they have their labs, right? Where they give researchers a budget and free reign to freestyle and follow what they're calling research taste, right? Like what is your intuition telling you, giving them the time and the space to really follow different nuances. and follow where the research takes them without having a deliverable by the end of the quarter and a deadline that says in three months, what will you have shipped and what will its impact be on our balance sheet? So I think I love that idea. And to your point, it doesn't have to be much. It can just be 10 to 20 % of the dollars and time that is being allocated to AI innovation for either external or external purposes into this just R &D. William Benjamin (44:55.343) Yeah. Khullani Abdullahi (45:04.492) Let's just figure out in our market, in our enterprise, in our fields, like what are optimal ways in which we can approach AI innovation. And that's different than the AI center of excellence model, right? So I think there's some interesting business model changes that will come as a result of the need to. to build in an environment where we still don't know what we don't know. And what we don't know, we don't know is very large right now. We have a lot less, I think, certainty. William Benjamin (45:37.27) Yeah, we are at the cusp of an AI revolution and yet nobody knows how it's going to turn out. Everyone's speculating, but that's where I think that you need to be really in the weeds to understand. And I think the AI budget will help there. Khullani Abdullahi (45:58.017) If you take a step back and think to yourself, 18 months from now, 24 months from now, and you look at where we were 18 months ago, 24 months ago, do you expect the rate of change to continue to accelerate? Or do you expect the changes to continue at a rapid-fire pace? for us to have seen like you don't expect the models to to improve at the rate that they were improving say over the last 24 months where do you fall on that continuum? William Benjamin (46:38.416) In terms of models, think we are kind of at least reaching a point where we have, you're basically training the whole internet, getting the data from the entire internet and then training these large models. Where I see the growth is the real monetization and improvement in quality. With AI, you can do hyper-personalization. Khullani Abdullahi (47:00.651) Right. William Benjamin (47:04.964) You can totally improve user interfaces. You can find hidden insights within, let's say, call transcripts or design documents that these, let's say, a manufacturing company had design documents. You can now go read those without a person actually going through those. And then you can derive these insights and then apply them to new problems. And that's where the real value of AI is going to come from. And a lot of companies are basically doing that right now. So next few years, I think you'll see a lot of growth where the applications are improving. And especially with like agent tech, there will be a lot of automation that will bring out these insights and help improve customer experiences, help improve efficiencies across the board. Khullani Abdullahi (48:03.032) Do you think... Khullani Abdullahi (48:08.238) that as the vendors in AI, I think, have much more power than traditional vendors and traditional, to say, even cloud providers. Are you evaluating vendors for? your AI stack differently than you did traditionally? What are you looking for when you are selecting vendors at those various touch points in the product cycle? Do you recommend companies build in-house and just kind of rely on their own engineering resources? Are partnerships becoming more important or just straight up? implementing and integrating third party solutions and then just building your own layer, intelligence layer on it. How are you considering, what are the trade-offs that you're considering? What are the factors that you're looking at? What is the decision-making model that you're using to kind of evaluate the vendor space? William Benjamin (49:09.432) Yeah, that's a great question. And often I've found myself thinking about it. I would say my framework is threefold. One is speed to market. Like how fast can you get to the market? And is that the most important thing? The budget. Like do we have the money? And to like do our own R &D build out this product? Or if we can throw a lot of money, we can hire smartest people and build out. And the third thing is security, right? A lot of the security problems actually happen at these interfaces or someone forgot to really vet this vendor and they were using some really outdated authentication and that's how we lost all our data or there was data breach. So evaluating, like thoroughly evaluating vendors for, because they are part of your ecosystem if you're building. Khullani Abdullahi (49:57.794) Mm-hmm. William Benjamin (50:07.03) out a AI system and then you need to have the same standards and vet them thoroughly at a similar level that you would do your own systems. So yeah, those were the three things. if, yes, speed to market is really important and they have the right security credentials, systems that they do the right red teaming, have the right testing for these security breaches. And then, yeah, it's a lot lower cost to pay a subscription or contract fees rather than spending money on building an AI team. On the other hand, with an AI team, you get full visibility into your customer. That insight could help you build other products or really understand the customer needs. So that's a trade-off where if I know more... Khullani Abdullahi (50:53.176) Thank you. William Benjamin (51:05.048) about my customers, I could build a really good customer experience. You kind of lose that with vendors. So I would say a hybrid model somewhere where the key insights and key security aspects, we would really like to build ourselves. But then the less, I would say, less critical aspects where instead of my team spending time, like six months building it out, We could just pay subscription or contract fees to just get the vendor on. I think that's kind of my framework. Khullani Abdullahi (51:42.264) Do you think, Chicago is the third largest city in America and you are hiring in this space and enabling individuals. Do you as a leader in AI and an enterprise org headquartered in Illinois, do you think that Chicago provides you with the labor and expertise that you foresee needing in both the present and in the future? Is there anything that you would like universities in Chicago, in Illinois broadly to be more focused on? How do you feel about our AI expertise in our state and the kinds of students and workforce we're producing? William Benjamin (52:28.908) Yeah, I believe in the firepower of Illinois in terms of AI. Like we have like great universities. We have University of Chicago, Kellogg, there's Purdue where I came from. So there is like tremendous AI talent available like, and they're hungry for, hungry for like these new ideas and AI implementation. And then you don't have to worry about Khullani Abdullahi (52:43.214) Mm-hmm. William Benjamin (52:58.256) paying that $4,000 rent like you would in San Francisco. So I think we are really, yeah. So Chicago is really set up for the AI revolution. It's only, people, think the community can grow. if some people take people like you who are building this community of AI professionals and doing a lot of meetups, like that's where I think the melding of minds will do. Khullani Abdullahi (53:03.222) Yeah, or Boston, right? Yeah. William Benjamin (53:27.522) As well as I think we have great MBA schools, again, Booth and Kellogg. No other city has so many good schools. So getting the right product mindset along with the right technical background. I think we have all the bones. And Chicago has a great history in the past in innovation. So yeah, I would definitely count on Chicago. Khullani Abdullahi (53:51.118) Excellent. So in the last part of the interview, I always like to do a little bit of Chicago culture. I always remind my audience that I was born in Somalia. I'm Somali and Ethiopian and I grew up in Minnesota, but my daughter was born here. So I do have ties to the city. But what is your, if you have to pick a pizza, are you a deep dish fan or are you a thin crust fan? William Benjamin (54:17.754) there's no question about it. It's the deep dish. Like there are two kinds of people. There's like the thick and the loyal people. And then there's like the thin and flaky. I'm just kidding. The people who ditch you on your second date. Khullani Abdullahi (54:21.07) you Khullani Abdullahi (54:31.278) Wow, you have a strong point of view. That's a strong point of view. William Benjamin (54:37.496) I'm just kidding, I like both but yeah, definitely preference for the deep dish. Khullani Abdullahi (54:40.91) I love it. last podcast guest, Sergio from AFVI, he said paella because he's from Spain. That's not the question, Sergio. So I love that you had a strong point of view. Do you have, I know you have children. So do you have some favorite activities that you and your wife like to do with your kids, like summertime in Chicago? William Benjamin (54:50.938) Ha ha ha. William Benjamin (55:00.73) Yeah. Khullani Abdullahi (55:09.87) What does that look like for you guys when you're not working? William Benjamin (55:13.848) Yeah, that's basically me being an unofficial unpaid lifeguard for my two toddlers. Basically think Baywatch, but minus the slow-mo running, but more like chasing with diapers. Khullani Abdullahi (55:26.414) So that they don't kill themselves, Yes. I remember that stage. Okay, do you have a favorite restaurant, whether it's for adult night or for when you have your children in tow? William Benjamin (55:31.182) Yeah. William Benjamin (55:47.584) it's, Billy's barbecue right outside in my backyard where I smoke like the best brisket. And then sometimes there's bourbon flavored, yeah, totally. You're invited. Yeah. Yeah. I have my whole smoker and a grill. that's my favorite. COVID just made me a great cook, I think. Khullani Abdullahi (55:55.372) Whaaaa- Khullani Abdullahi (55:58.956) That sounds like an invitation. That sounds like an invitation, I'm bringing my daughter. You have a little smoker in the backyard? Khullani Abdullahi (56:15.502) I love that. And you went out and bought all the gear, so now you intend to use it. William Benjamin (56:20.17) Yeah, yeah, totally. I mean, it's a professional activity. Khullani Abdullahi (56:24.952) I love it. I'm glad to see machine learning leaders have great hobbies outside of their work. I'd love for you to share any last thoughts with our audience about what's happening in this space. You sound very hopeful and optimistic about what an AI-enabled future could look like. So we'd love for you to share any last thoughts or comments on that. William Benjamin (56:51.692) Yeah, of course. think I want to just reiterate, I think we just talked about it, right? It's a choice where we can be like scared, one scared of AI and like really clamp down on it, or we can use it for human flourishment and like really improving customer experiences, giving more personalized, let's say teaching to our kids, improving the literacy rate. helping reduce crime. So there's a ton of opportunity and I think as technology leaders how we use it and use it responsibly and ethically, I think that's the key. REST I think is just human endeavor and adventure and it's going to be a great adventure. Khullani Abdullahi (57:41.652) I agree. I agree. think we're, I feel very fortunate to be in this space, in this city, in this time period. So thank you, William. I really appreciate you taking the time to, out of your day, which I know is incredibly busy, to share your machine learning expertise and insights as an enterprise leader. So I appreciate you joining and to our audience, you'll be able to catch William's episode on Spotify, Apple podcasts, and on the website ChicagoPodcast.ai. So thank you guys so much again.