Dr. Elouise Epstein on the Future of AI in Procurement

Dr. Eloise Epson: I think as I look at the future of business, you can go function by function, and we're going to get the black boxification of every corporate function. Finance is going to be an AI black box with a handful of people that manage the black box. Welcome to Modern Business Operations, where we talk with leaders about how ops is adapting to our modern world.

Host: Hello, everyone. Welcome to another episode of Modern Business Operations. Today I have the pleasure to Talk Again with Dr. Eloise Epson. Hey, thank you for joining me today.

Dr. Eloise Epson: Thank you for having me. I am thrilled to be back.

Host: So, just for listeners that don't know, um, Dr. Eloise is a partner at Kearney, keynote speaker, you work in university in steering committee, author of books, co host of podcasts, and, and really everyone that is in the procurement space, your influence and, um, thought leadership. So thank you for joining again and you know, if you're interested, go listen to our, uh, podcast episode from more than two years ago now. I looked it up, it was July 23rd and oh my God, a lot have happened since. So, again, thank you for, uh, joining. And I guess the obvious, the obvious place to start is how do you summarize what have changed in the last two years?

Dr. Eloise Epson: I have to admit, and I may, I haven't gone back to what we talked about, but certainly I was not on the AI train back in 2023. I remember you all were the first ones that I had seen had put GPT into your solution. And I was, I was very impressed with that. But honestly, at the time, I hard, I, I was still like, AI is going to be blockchain 2.0. Um, and while was I wrong and I have, I have come full circle, I have finally, I've drunk the Kool Aid, I'm all in. And, and while it may seem like I'm jumping on the bandwagon, I've. I've also started to witness the just absolute transformation and we're both in the Bay Area. And I'll just tell you, without meandering too far off the question, what really unlocked it was taking my first Waymo, and I figured if you could have an autonomous robo taxi running around San Francisco, to me, that was when AI became real. And then the next generations of the models came out and that was it. I'm all in.

Host: Yeah, well, I mean, I mean, yeah, I think, I think we kind of like we're, we're talking about it and how impressive it feels and maybe the double click on that because I think, I think you representing a big group of people that, that might feel that way. Maybe there's some people that are still on the, on the fence of like is this real or is this uh, still hyped? Because again, if you mentioned San Francisco, there is definitely a hype too. You know, it's like yeah, oh sure, you go, you go down. People are not from this area. You know, if you guys come visit and you kind of like drive the one on one, it's pretty ridiculous. All the billboards, oh my God, for this, for that, everything is a, it's, it's.

Dr. Eloise Epson: Can I just say before you ask the question, like I take my wife to the airport and like half the billboards she's like what's that mean? Like prompt to promotion or something. And I'm like, I don't know, I kind of guessing at this. Yeah. Oh my God. The billboards on 101 are just absurd.

Host: It's absurd. It's like AI platform. The AI platform. The AI enterprise platform. Enterprise AI platform. It's like it all says the same thing too. But anyway, so with that side of it, it is definitely hype to say the least. Maybe overhype, but I think a lot of people kind of went through that progression you described where became a believer at least of saying there might be hype, but there's also a lot of real here. And so if we double click into more into the world of enterprise, even specifically procurement transformation of enterprises, where are we in that scale? Because driving a Waymo is a really crazy experience. Now where does that, where are we? How far are we from virtual? Waymo is actually taking care of business.

Dr. Eloise Epson: Well, just because, well I actually just saw something recently that uh, we all can now in the near future we're going to be able to buy our own robotaxis home. So I'm really geeked for that. Here's why I think AI is not just another blockchain or RFID or all IoT all this other nonsense. And it's very simple. And I know you asked about the enterprise. Everybody that has kids or has a vested interest in family members that are in the, any educational system right now is confronting AI. It is. I teach, uh, I'm an adjunct at a uh, university. I mean it's just. And my wife teaches, is uh, a professor at a university. And you could just see like the entire educational system, how we learn is absolutely getting turned on its head. And I think over the next five years it will be a legitimate crisis. Forget what happens at work for a moment. Everything from K12 universities, uh, unless it's a hands on program, uh, I think it's unavoidable and so I think that's going to have. Because uh, and I tell this joke a lot or not joke, it's not even a joke, it's a real story. When I was in grammar school we could not use calculators in math because it was quote unquote cheating. But in almost 2026 here, that seems so silly or stupid depending on your perspective. So you're going to tell a kid that the entire life, you know, fifth grader, tenth grader or a uh, freshman in college, their entire life is going to be using AI, uh in some shape or form that even um, like none of us that are older even know. So we're going to tell them you can't use AI because it's cheating. Uh, we have to reconcile that so. And I know there's, I'm skeptical about enterprise adoption and rightly so, but fundamentally is going to reshape the educational system so it's unavoidable in society. And then the only other thing I'll add is AI is being used as an excuse to fire a lot of people in Silicon Valley and now it's AI sort of the uh, catalyst. But you can recalibrate the workforce and what happens in Silicon Valley will eventually make its way into the corporate world. So it's unavoidable.

Host: I mean you don't need much. I agree, you just need to touch it in many ways to see, to see some of those things. I feel what's interesting to me is there's been a little bit of, if we kind of go one step below for a second, there's been a lot of back and forth in the last two years of what is a GPT wrapper, thin wrapper, actual mode, no mode, wait is GPT is even a moat? You know there's like a lot of, there's a lot of um, there's a lot of talks like that at the same time the market at least from, if you look at some of the growth, you know of a lot of companies in the space is spending and spending a lot of money on, on things that in theory are available already in chatgpt for fraction of the cost and to me that comes back to something that was true in any other variation of big technology leap which is the use case is many times an incredibly important piece of leveraging technology as well as the fact that most people are not educated enough, sophisticated enough, have the mind share or care enough to come up with leverage points for themselves with the technology they kind of needed to be packaged for them. How much do you think that's the adoption gap versus you think there might be something deeper that will kind of like transform?

Dr. Eloise Epson: Okay, so I, so I have, I think I have a view that's different than most people. I don't care about use cases. That is an old way of thinking. I believe that 100% and a lot of my work is focused on the over 40 crowd. Because my generation. I'm not going to make any assumption about your age. So let's speak about my generation. We can't sit back and wait for it to be packaged and delivered to us. And so I spend a lot of time. There's no reason anybody in the business world today can't use AI for personal productivity. You don't have to connect to your ERP system. You don't even have to connect your email. You can use it for personal productivity. I used this example of I take a lot of screenshots. Well, and then it just stamps the. You know, when I take the screenshot, it stamps the date and time and tells me nothing. So do you know how much time I waste going through opening like a hundred screenshots trying to find the one that I think I took that like in June. So I go like. Whereas I can have an action or an agent that goes through all of my images and retitles that titles the file name as what's in the content. That's a huge time saver. How many of those things? And I don't even need it tied to any of my, you know, work email. Yeah, exactly. So we start doing that, making it tangible. I just increase my personal productivity by X percent. And then you go further down the line and use something like Notebook LM from Google or where now it's creating crazy or Nano Banana crazy infographics will create PowerPoints for you or slides. So my goal is to reorient. All of my work now is on culture and people. Because especially those of us over 40, we have to get them going. Because the folks under 40, it's not something they have to learn. It's, uh, an experience they've lived with since they were young.

Host: Yeah. And I think when you talked earlier about the next round of people that will join the workforce, you know, and it's like, it's one of those things when you think about Internet and then Google search and then mobile, it wasn't like generations. Even the word generation is too big. Right. It's a matter of where were you when it came out? Literally, were you in the five years of elementary school, the uh, five years of middle school, or were you in college? Or were you the first job? Every five years, if you were earlier and earlier and earlier, it's more and more native to you. And the same applies to the ChatGPT area for, for the sake of being the first, they kind of, uh, get to claim that the era. But what does that actually look like? Even for people that just two years ago when we had, uh, our podcast, they were, you know, maybe starting college or maybe starting high school and are going to be working in companies in the next year or two or three, they're going to come completely different to it. And to give you kind of the side of leading a company for us internally, I've had to have those conversations with a company and saying you either learn how to go first to everything, first to AI, or I'm not sure you're going to have a role here for much longer. It just doesn't make sense for you to even ask those questions. Did you ask GPT? You know, like even internal stuff, you know, our documentation of our own product has been around, so GPT itself is already trained on it. You can go and ask like, how do you do accent Tonkin? And it knows ChatGPT without any sources, it knows to answer. So for a new employee to come and ask someone internally of it before they ask ChatGPT is just ridiculous. So there's so much of it and it's amazing. The fear from it was if going back for a second to kind of analyzing and predicting the fear of it in the last two years was also a lot about how is this going to replace jobs? How is this going to replace job? I'm going to just jump to a conclusion that you probably agree that it's not going to eliminate all jobs, it's going to change them and so on. I'm curious to hear, but also I'm curious to your thought of the timeline of when people not only, like you said, going to use it as an excuse to reshape, but actually would start questioning do we even need this role anymore?

Dr. Eloise Epson: So I'm not a big fan of okay, AI comes and everybody gets fired. I think that's too simplistic because I do think new jobs will come. However, I do think there's going to be a big lag because again, you see it where see, we see it in Silicon Valley. Huge numbers have been exited and more will undoubtedly come in 26 and 27 and then there will be. So it will be a big dip in the labor market, but then it will new things. The new thing will come along and I don't know what that'll be. So there will be jobs. But I think that gap, that lag between the two, uh, it, it's typically not so dire or, or so so disconnected where we're going to get exits and then a, a gap and then rehires because we. I've seen this look at uh, the turn of the century with the dot com bust. You know, everybody said, well technology and the Internet's dead. And then within two years you had Google, YouTube, Facebook, and then all of a sudden it was rehiring and then all of the network and IT infrastructure jobs. M came back, but changed. So I think that's going to happen, but I think that lag is going to be longer. But I do think that what comes next is going to be different and it actually might be more analog. It might be more the intersection of AI and physical technology like robotics. I think it's hard to separate the two. So yeah, it's going to be bumpy I think is my main takeaway. But that's why I'm trying to get everybody oriented to this change.

Host: This episode is brought to you by Tonkin. Tonkin is the operating system for business operations, providing businesses with the building blocks to orchestrate any process with no code or change management request required. Contact us@ahtonkin.com to learn how you can build complex processes fast. When we look at the area where you specialized a lot, at least, you know, kind of where I got to know you around procurement and um, supply chain and, and, and uh, maybe financing more generally. How do you see the next year? Maybe maybe in another version of it for a second. I think there's a lot of interest and a lot of excitement and a lot of people spending money on all areas of the business today on AI. But nothing comes close to how, huh, what you see with R and D, right, And anything around R and D&AI is just like orders of magnitude more than everything else, which for me this is just textbook. It happens with every new thing. There's always uh, there, there always obviously, um, because they're the technical side of it, they're always the first to kind of like get to the bottom of how can I use it the most and where I can, you know, get the most impact. So I do think that this is going to be. Look different in the next year or two. I think even in the next Six to nine months. But what are you saying?

Dr. Eloise Epson: So, so I, I'm always wrong on the timing because I always think it'll go faster than it does. I don't know that we'll see significant changes in six months or nine months or 12 months. We might, uh, but those changes are coming and I don't think it's just procurement. I think you were alluding to it. I think as I look at the future of business, you can go function by function. And uh, what's happening I absolutely convinced I'm right about this is that uh, we're going to get the black box ification of every corporate function. Finance is going to be an AI black box with a handful of people that manage the black box. Hr, same thing, uh, you know, on down the line. And so certainly for corporate functions and for procurement source to pay, most of your hot model or operating model path that we do today are going to get agentified and then black boxed. That's, that's what's. We already see tons and tons of agents in source to pay. A certain point, you have so many agents. It's really just a black box and a handful of people. If I'm in legal, finance, IT or hr, when that black box comes, most of the people have to go find something else. In procurement, I'm more bullish on this procurement supply chain. When that happens, I'm like, yes, please and thank you. I don't want to do any source to pay work. That stuff is horrible. What I want to do is the business of procurement and I want to sit with the R and D team, sit with the product developers and marketers and engineers and bring procurement practices into the business. And that is going to be a huge, uh. Like I said earlier, I focus on people. That's a huge capability because when I go sit down with those engineers and whatnot, I'm not bringing a spend report, I'm not bringing a category playbook. Right. I am, um, I am part of the business and I'm thinking about the P and L of the business that I'm in and I'm going to use AI to help me be a productive member of that team.

Host: Maybe we connect this to the next level up. Right. So we talked about the people side of it. If we think about the technology side of it. You had uh, the famous spider map, um, that you retired.

Dr. Eloise Epson: Retired, yeah.

Host: Curious of why by the way, but yes. Yeah. So definitely, definitely. Um, but then maybe it's related to. But in many ways there's also this question of if it's a black box or even if it's on its way to be a black box, some categories within technology categories, software categories within those departments that have been raining for a while. Do they even have a place in this world? Uh, eventually and immediately.

Dr. Eloise Epson: Right, but we can make that argument about finance or HR IT as well. Like what is HR's job? It's to sort of collect and manage data that sits in workday. But imagine, I mean take all the like and workday is just the gigantic fancy HRIs. HR, ah, information systems database. Well it's not telling me anything about my workforce and it's certainly not telling me anything about my agentic workforce. So it's built for the wrong time and place. I was picking on them. Normally I pick on SAP, but we'll pick on workforce work day to day. And so if I'm going to build logic in terms of how to build an optimize workforce, uh, I'm going to do that in an agentic layer. And if I have that agentic layer and all of that, or in AI layer, I should say with agents and everything else, that's pretty much a black box. I don't need to know how it does, the AI does its work. I just need to tell it what I want it to do and that's the difference. And so in your example a lot of these jobs won't exist. I don't need go down finance, like look at almost every ah, position finance. AI does that better than humans.

Host: You know part of this is also the responsibility aspect of it. So if we do take the, again the canary in the coal mine if you will with R and D, it's very clear that the job of a uh, developer or a software engineer is already completely different. This is not a futuristic thing. It's like most, you know, I think, I don't know, Microsoft said it, Google had their own thing, everyone are talking about it. There's multiple now at this point where like 75, 80% of every new code in the line of code in the world is already uh, written by AI. And it makes a ton of sense. And you know, I'm coming from that background of product and engineering and we have had this saying for like I don't know, 30 years where it's like there's a developer and there's an engineer or there's a coder and there's an uh, engineer. It's not the same like an engineer can code, but engineering is including the architecture, the converting the business needs or the use Case or the user story, whatever you want to call it into. How do I build in a way that scale and stuff and the world is definitely moved there where you don't write the code, you review the code. Right? Uh, uh, in many ways. Now, again, using that as an example, I think what's interesting to me is that I have an attorney's counsel. I don't need them to review the NDA. I can do a. Have AI do it. I don't need them to review other type of contracts as well. But for some of them, I do send it to them, even though I already ran it through ChatGPT for that matter. Because I need them to be responsible for it. I need them to say maybe they run it with GPT2. I don't care at this point. All I care is that they say, yeah, that looks good to sign. Now what happens there is the transaction became less about the work and more about the responsibility. And the same happens again with the coding. Right. It's like who committed it, uh, who approved that code to get into production? Becomes the question less about who wrote it because everyone know they're not writing it anymore. Right. How do you think about that? Because to me that's an un. Um, we don't speak enough about the fact that we're just walking people to people. Like this entire thing is all fake, you know, like all this world that we live in does not exist if other people don't exist in it. So that black box at one point becomes irrelevant if there isn't a person from the other side that cares about the outlook.

Dr. Eloise Epson: Well, yeah. And I'll make two comments. One is uh, or two different. Two vastly different responses. So we are humans living in a physical world. So at the end of the day, business is between relationships. Now the question is if you are critical supplier. If Tonkin's like the thing that I've built my business around, that's an important relation. Like our relationship is important. But if I'm going to buy Microsoft Office, that's not ah, like I don't care about uh, like it's just like give me the price, execute it. And the black box goes and does that and, and that's. So when I talk about the black box that allows or I need pencil or office supplies, like nobody is sourcing office supplies anymore because we buy so few of them. I just want the black box to do it. We need banker boxes and Sharpies and whatever and I don't know, just send them to the off like that's it. Like that's. And so the black box should be able to take care of that. But if you and I are talking about the health of my business, then yes, and. And by the way, I thought you were going to go with. Because of all artificial construct and money. Artificial construct poor.

Host: It's part of it. Yeah, right.

Dr. Eloise Epson: Yeah, exactly. But I know that's all. But the other, going back to the coding, think about your coder and your engineer and everybody else, the QA and all that. Uh, I now, because I'm none of those things, I can build my own AI apps just by prompting it. And I do this in my masterclasses. If I can do it, it's like, oh my God, it opens up so much for me because I know the AI is m rudimentary and how it codes and builds these apps, but what happens in six months or a year when I can build anything I want? It's just my imagination. And that, I think is a fundamental change in all of this.

Host: This is kind of. My point is like what happened when you stretch it for a second and it's more than just saying like your point about work there or SAP or a lot of the massive companies that came by just wrapping around a database right now, that is a big question. But also if you can build your own anything. Let's take a CRM as an example. Oh, I want to build a CRM. Papa be papa. You do it very quickly. Now here's the thing though, that was true before too, because you can just run your CRM on a spreadsheet. You can use other tools, you know, spreadsheet like, um, that, uh, spreadsheet I call them, you know, that, that allows you to do some of those things. Still, most people pay a lot of money for Salesforce, so there's something in the culture, there's something in human behavior. I think part of it is the trust or the blame on the other way of it. That is interesting because even if you buy yourself, if you buy yourself an AI app that is used just by you, that's one thing. But if you need to build it and now other people need to use it, there's a little bit of a dense there of what happens when there's need a change or does it need a problem? Yeah, of course you can go and prompt the AI to make the change, but who's going to test the change and then communicate the change and then make sure that people are not confused about the changes, so on and so forth. Now this might be looking at the world in with the old glasses to your point.

Dr. Eloise Epson: That's exactly what I. Yeah, that's going to be my response.

Host: Yeah. So, yeah, let's talk about that.

Dr. Eloise Epson: I think when we. I think that's all true, but also at the end of the day, actually, I built my own CRM vibe coding because what one critical diff. Because I have, I have a, uh, uh, spreadsheet plus plus for my CRM and I use Airtable. Um, but it doesn't capture at a spreadsheet, does not capture what I need. Because I will note that we recorded this podcast. So every event I do, I. I have a who is the sponsor? Or, you know, not, not sponsor, but I mean, who, uh, who did I do this for? And so I have Tonkin and who was the other person? And it's a ma. It's a podcast, not a masterclass, not a keynote, not a whatever. Why wouldn't I want to collect? Actually, why wouldn't I want to point my CRM record for this podcast to the actual podcast and pull out the common themes? And that's where the power of, like, I need a CRM that is not constrained by a spreadsheet because I can't do that in spreadsheet construct. So this is about making it useful to me, an audience of one. And so to your question, this is not like the old way of thinking is I'm going to build a tool for everybody. Where I'm looking at it is I'm building a tool for me. And it may not seem efficient by today's standards, but I actually think there's power in it because, like Carney, Carney rolling out Salesforce, it's a disaster. And I can't use it because it's not geared to me. It's geared towards different type profiles. They're never going to get to my needs. And so I'm better off building my own CRM, um, that stores everything essentially in memory. I'm using that term loosely, but not in a structured column or database. So that I think is the big difference, um, in terms of old mindset and new mindset.

Host: Yeah, it's fascinating. I think it's going to be. I think there's a term for the coin, like, not specifically the way you're doing it. So, like, what you're describing is really, really is a personal application. There's also the concept of burner apps, which is literally, you know, not even like an application for your kind of personal use for a specific area, uh, of your life that you are doing, but literally I need something right now for this, like the next 10 minutes and I need like you know, a quick application to do that.

Dr. Eloise Epson: But that's my example of the screenshot. Like I need to go find a screenshot from June or heaven forbid January and like if I just took 10 minutes I could just build that thing and then I will get, I can find the screenshot I need and then oh, uh, by the way, now it's just running in the background. But I also want to make a point because I think in this old mindset versus new mindset, the way I look at it is you can't have a free for all, not in the enterprise. So what I like to talk about is there's a, has um, to be a governance layer and governance is a way of operating so many corporate legacy. Corporate thinks of governance as a document that sits in a SharePoint folder that nobody looks at. Whereas I'm talking about like there are certain things that we have to tell AI. The AI or the procurement brain is like call it, you can do this, anybody can do this, but only a couple people can do that. And one thing might be the classifications. We're going to have one taxonomy classific for classification. You, anybody can access it, but only a couple people can update it. But if we, as long as we use that, we use this, whatever, I don't know, date of heat and whatever else, then you can build whatever app you want, whenever you want it. And that's the. I think we often lose the thread of, of governance because yes, it would be a complete free for all. And by the way, in that governance, let's say you work for me in whatever and we work for some big enterprise, you leave, you built all these apps and agents and then you leave. I should be able to tell the AI to go in, pull, interrogate what you did, pull all your agents over, pull whatever apps you have, summarize it, store it off and then if I need it or ask it to ask the AI to reference it if I need to. And that's like this is a whole, this is that whole new mindset and I, I don't even know everything but my, But I want to direct people to start thinking like that.

Host: Yeah, it's really interesting, um, it's really interesting to think about what would stay and what would not. Um, I think your kind of last point is probably the biggest point in my opinion. More than just the governance, even the people living because I think personal productivity, you kind of go back as we kind of come to the, to to kind of like our time. But as you look back into uh, where we started, everyone, there is no excuse to anyone to not be using AI for their personal productivity. Like hands down, just unacceptable at this point. And then you take it from there and you realize that when it does get to a point where you are part of a group and a team and if you want, uh, for your own thing, do your own thing. When you kind of like need to work inside of that organization, large organization, even if it's a, you know, small organization, definitely big organization. This is where I think to me there's not a clear definition yet where the compliance lines comes in. When is the thing that you've been doing an amazing job at because you were really good at kind of building an army of agents around you to do great things, but you only thought about yourself to do your job well and then you move. Forget about, you're never going to get fired, you're doing an amazing job. But, but you just decide to move. Now this, this team is holding a bag. Uh, they don't know how you've done it. It's a black box and, and they're stuck. There's interesting situations going to happen.

Dr. Eloise Epson: Oh, I know, but, but ultimately it's a black box to me too. So like, however I did, let's just take my screenshot example. Like let's say that was something that other people could make use of at, at a certain point, like I don't know how the AI is doing it. I just said I have a full turbo screenshots and I need you to redo the file name. And I want, and then so the instructions I. So when I leave the AI knows that it built that thing for me because you and I are using the same AI. And so when I leave you can just say what apps did Eloise build? And it's going to say it has this screenshot thing. And by the way, when it gets to governance, it's sort of like macro level governance. Part of it is if I make the request to build, like I say, here's my prompt, I have a uh, folder full of screenshots. I want you to retitle it. And then you come in the next day and you put the same prompt in, it says, oh, I built one for loes. Would you like the same one? And because it's again learning and it's keeping all of this in memory, this is the mindset shift that uh, uh, certainly those of us who are 40 are going to struggle with because this is reimagining how systems work. And I think ultimately, to get back to your earlier point, this is going to reshape how businesses work. And I'm actually excited for that because it is going to be. The next 10 years is going to be wild in terms of, like, mind sort of expanding. So.

Host: Yeah, absolutely. Well, that's a perfect, perfect spot to end in. And maybe we, maybe we meet again, uh, doing this in two years and see what we got right and what we got wrong.

Dr. Eloise Epson: Exactly.

Host: Awesome.

Dr. Eloise Epson: Sounds great. Thank you for having me.

Host: Thank you for. Thank you for joining me. And, um, if anyone wants to continue nerd with you, I think LinkedIn is where they find you. Is that correct?

Dr. Eloise Epson: Yes, I'm a little slow, but as I, uh, do respond also, uh, you can find me on drelloise.come L O U I S E awesome.

Host: Eloise, thank you so much again and we'll talk again soon. Thanks.

Dr. Eloise Epson: Cheers.

Speaker C: I hope you enjoyed this episode of Modern Business Operations. You can see the show notes and all of the resources mentioned on today's episode at tonkin. Com mbopod. Thank you for listening and be sure to subscribe for updates on future episodes. And if you're interested in staying up to date on the latest in business operations technology, head over to Tonkin.