AI must look for work your finance team hates to do – Hyoun Park
48:45 Share

If you would like to earn CPE credit for listening to the show, visit earmarkcpe.com slash FPA. Download the app, take a short quiz, and get your CPE certificate. Finally, if you enjoy listening to FP&A Today, please go to your podcast platform of choice, click the subscribe button, and leave a rating and review of the show. And now, on to the show. From Data Rails, this is FP&A Today. ♪

Welcome to FP&A Today. I'm your host, Glenn Hopper. Today, we have an incredible guest joining us, Hyun Park. Hyun is the CEO and Principal Analyst at Amalgam Insights, where he helps companies optimize IT costs, build financial business cases, and navigate the evolving landscape of AI and finance.

With over 20 years of experience in IT expense management, FP&A tools, and business analytics, Hyun brings a wealth of knowledge to our discussion. Today, we'll dive into the latest trends in finance technology, AI-driven decision-making, and the future of FP&A. Hyun, welcome to the show.

Oh, thanks so much, Glenn. Pleasure to be here. Yeah. You know, ever since you and I had our first meeting with, we did a webinar with Brian Lapidus, who many of our listeners will know from Association for Finance Professionals. But even when we were talking about what we were going to cover in that webinar, I thought, I got to get you on the show. So here we are, you know, weeks and weeks later, and we finally have you on the show. So thank you again for joining.

Oh, yeah. I've been looking forward to it. Well, let's dive in because I think, you know, maybe you're a little bit different than most of our guests. I'm either talking to a FP&A seasoned veteran or a VP of finance or a CFO. And you have this kind of IT background, but I know with the finance crossover and everything. So I know you've done IT expense management, AI, finance, all that for the past 20 years.

including your industry analyst role for FBA tools, which I think our listeners will be very interested in hearing about that. But kind of walk me through your career and how you got where you are. Yeah. So I'll be honest. A lot of my early career was, I guess, a stepwise mistakes to get to where I'm going. When I graduated from

college, I had a degree in women's and gender studies, as well as I had taken all the pre-med classes. So I had a bunch of science and math under my belt, even though it doesn't look like it from my degree. So based on all this, I had to figure out what am I going to actually do in the real world

I found out that chemistry jobs at the bachelor's level are pretty boring and women's and gender studies jobs were not in full supply at that point. So it was the dot-com era, the late 90s, early 2000s, and there was a ton of jobs in tech.

So I got into a tech company and finally learned there's this thing called CRM, you know, customer relationship management. And from that, I learned that there was a database behind that CRM. And then I learned that databases are used for other applications. And then I started following the money and realized that there were applications that

deal with payments and deal with budgeting and deal with accounting. So over about six year period, I shifted from doing basically CRM administration to going into databases and then getting into payments. During that time, I realized,

wow, there's this whole bigger picture of how technology deals with the way that we do transactions in the business world. Like I know there's also consumer payments, but I was really interested in the business side of dealing with

invoicing and dealing with the customer relationships associated with the business, and then how that fit into planning, budgeting, and forecasting. It's just where I was pushed into over that initial five or six year period. And I would say that everything after that period has been trying to figure out how these markets work to a greater extent.

And so when you were in that six-year period and you were understanding all this, was your role CRM administrator or were you more working with the database or a little bit of both? What kind of roles were you in? Yeah, I would say the great thing about that period of my life, about the six years of working in startups, is that you're...

Job can change rapidly depending on your interests. So I started as a CRM administrator within my first six months of being in the workforce because nobody else understood CRM.

what this customer relationship software was supposed to be. And then I was off to the races from a self-learning perspective that I had to learn the database behind this. And then I started teaching myself SQL. And then I started working with the actual database itself. And then I got my next job at another startup where I got to work more deeply with databases in general.

And then from that point, I started working with the payment and billing systems of the telecom company I was working with. So then I started learning about OSS and BSS based systems, like all the operational and billing systems that telecoms use. And from that point, I started learning more about the accounting aspects of what was happening. And this all happens fast because in a startup, you don't have time.

time to learn stuff. You either get it quickly or you don't get to do the job because that job gets filled. So each of these sprints were like three to six months to get up to speed and then to start doing the work. And it ended up being almost like a graduate program, just

to figure out tech. I shouldn't say graduate because it's not like my initial understanding was that deep, maybe more like a new bachelor's or like a certificate program, but they were paying me to learn. Yeah. Yeah. And that's the amazing thing about startups is you are exposed to so much and you there's not anyone you can turn to. So it's, you know, the pedagogy of the startup is, you know, it's not going to the professor's office hours and and getting help. It's there is no professor

We need to figure this out right now. So, I mean, you're forced, you know, trial by fire, forced to learn. But I've talked to so many people and I share the same experience where that startup environment lets you wear so many hats and be exposed to so much. So it really is like a continued education experience.

Yeah, I will say I had a couple of great bosses who were able to sometimes help me with subject matter expertise, sometimes just able to hand the next book into my hands to teach so I could read about what's happening and say and just give timelines like you have to learn this in two months or else I've got to hire somebody to do it. Yeah.

So I think it's interesting that with this experience and background, it would have made sense for you to go on and become a DBA or even moving into data and analytics. But with Amalgam Insights, or even like going into IT, but with Amalgam Insights, it's such a laser-focused thing.

service that you provide. So it's not just doing IT, it's helping companies cut the IT costs and optimize their financial strategy. So tell me about what Amalgam Insights does and kind of how you got to this as a focus. Sure. I started Amalgam Insights in 2017 as a bespoke industry analyst for

firm. And we have two main areas of focus. One is IT cost management, where we look at telecom network software and cloud spend and solutions to be able to cut those costs. The second is around as the initials of Amalgam Insights intimate AI. In 2017, I knew that AI was going to be

a big topic in IT. I just didn't know when. We were already starting to see efforts with data science and with people learning more about these data science and machine learning platforms. I was even hearing from finance people who were already trying to get master's or certificate

in machine learning because they knew they were going to have to improve their chops from a forecasting perspective and from a strategy perspective. It wasn't everybody, you know, it was probably one in a hundred, but it was just enough to see that there was an emerging trend in this area. So I wanted to be on top of that

trend and see where AI ended up going. And then of course, when generative AI took off three years ago, that it was off to the races on the AI front. But I think it's important to also remember that there were many other efforts of AI over the past

60, 70, 80 years. Everything from statistical forecasting to algorithms to machine learning, all these things that came before the last couple of years of generative AI. So I've been working on providing these solution guidance to end users

on the end user side. And then I also help vendors with product messaging because often they create products, but they don't know how to align what they're doing to actual use cases. So I help them both with the language and kind of the roadmap for some of those efforts as well.

And I got started with doing that back in 2008. I became an industry analyst at a company called the Aberdeen Group, where I was introduced to this business. I initially thought I would be there a year. I would talk to 100 cool companies, and then I would probably choose one of those to go to to do my next startup work.

And I did meet 100 cool companies, but what I found is that I actually loved talking to all these companies and figuring out what was happening next and seeing what was next on the roadmap and being able to compare between all these different solutions and start mapping out what the future might be able to look like based on all of these

confidential briefings that I was getting and I just got hooked on the job. Yeah. And you know, you are, you're in a unique position, one, because of the, you know, the focus and the research that you're doing, but two, because you are getting two-way communication with the people in these companies. So I get asked a lot of questions and I know the little slice of the world that I do business with and where, you know, what problems they're having and how they're, where they are and sort of, you know,

their data maturity, AI use, systems, technology, and all that. But it's a very small group, whereas a big part of what you do is talking to all these people. So I guess, obviously, I want to lean way into AI. But before we do that, just thinking about technology in general and with the kinds of customers that you're working with, and because of the focus on cutting IT costs and optimizing their strategies,

Maybe it's still anecdotal because I'm asking you to tell me this now, but you're talking to so many. Your sample size is much bigger. So I'm wondering, what are the biggest inefficiencies that you see in enterprise IT and finance today? I would say that from an IT cost perspective, there's a lot of focus on cloud right now. A lot of companies started using cloud because they thought it would be cheaper and it is cheaper to start.

Almost always, it's easier to start with cloud, but those costs can quickly get out of hand if you're using something like Snowflake from a data warehousing perspective, or if you have created AWS, Amazon Web Services,

It's very easy to duplicate a lot of services and then lose track of what is live and what is not live. Especially if you're an engineer and your job is just to get something working quickly, you might quickly put up five, six, 10 different instances of basically the same job because you're doing comparisons and trying to figure out what works and what doesn't and doing all these experiments. But if you don't shut off those 10 instances afterwards, you're paying obviously 10 times as much to do the work.

So stuff like that happens all the time in the cloud. And then I would say on going back to the finance side, I think the biggest challenges right now are, I'm not going to say this is anything new, but understanding marketing and sales spend, as well as being able to adjust budgets more quickly. We obviously got tested with this really hard in the COVID shutdown era.

But I feel like a lot of that experimental, faster cycles of finance have started to go away again. And people have gotten, I don't want to say lazy, but back to the traditional cycles. And it's not necessarily a time to do that when business is changing so much. And frankly, we have a pretty volatile geopolitical and business environment right now.

Yeah. And I don't want to go too far off on a tangent, but I'm actually personally curious about this. I worked at several companies that had pretty significant cloud spend every month. And I, you know, me as finance guy, even someone who considers myself a fairly technical finance guy, I don't know how much S3 storage is,

too much or what, you know, or what servers we spun up that can be spun back down. And like, so trying to understand all that as a CFO was, was stressful. And it, well, you know, even like on the marketing spin, yeah, I can go look at it and say, okay, we did, you know, this many pay per click ads and the price was up because we were going, you know, I can understand all that, but it's hard. And then even trying to extract that information from engineers, but the technical debt is a, is a real problem. So if you're coming in and advising companies that,

I mean, you know, you could do a one-time assessment. What's the guidance for companies like ongoing to monitor cloud spend and how they could understand that from the office of the CFO more than just from the engineers? So in aligning finance to cloud costs, it can be really challenging to understand

simply match up what finance is looking for from a cross-charging and planning and budgeting perspective with what the engineer is providing. And that mapping has definitely been a challenge in managing cloud costs. I feel that there's also a challenge that on the finance side in general, when financial professionals are looking at

IT in general, there's not quite the right amount of information for cloud costs to be aligned to the business the same way that marketing, finance, and other operational costs are often mapped to either a product or a business unit.

Yeah, I'm laughing because our audience is probably thinking I've lost my mind because I had just a couple of weeks ago, Damon Fletcher on who was the CFO at Tableau and at DataRobot, but he now has a company called Caliper, I think, that they track cloud costs. And so we spent a lot of time talking about cloud costs there. Now I've got you on and we're talking about it again. And it's been several years since this was an issue to me, but it was such...

an issue every month. I mean, just seeing how much we were spending on the cloud and I felt powerless because I didn't know what was going on. And it's funny, you mentioned the sales and marketing spend and it's that John Wanamaker quote, I know half the money I spend on advertising is wasted. The problem is I don't know which half. And so that sort of just makes

sense. And it's always hard to figure out the ROI on a marketing budget. But then with IT, you don't even know the widgets. What are we even counting? What am I paying for? What are we doing? So apologies to the listeners for dwelling on cloud costs, but you can tell I'm having PTSD about it. There's a lot of people who get obsessed about this stuff. When I'm looking at the cloud spin-off space, I currently cover over 80 solutions. And a lot of these founder stories are very similar.

I was either an IT director or a finance manager, and I got ambushed by this $20 million cloud bill, and it's haunted me ever since to the point where I had to create a software company just to get these nightmares out of my head. It seems to be a pretty common obsession out there. Yeah, that's funny. So, all right. Well, finally, I'm going to let that go, but I can't say I'm not going to bring it up again next week. Yeah.

So now the other thing that I dwell on is AI. And because this is the panel you and I were on, and also because you are talking to so many companies. So I don't remember if we talked about it in this webinar that we did together, but a question I get all the time is, what are practical applications of AI?

generative AI in finance right now today. And I know, I mean, I know some companies that are out there. I know some companies that are doing, you know, bespoke stuff where they've got some cool stuff that they've built out in-house. I know a lot of companies that have given their employees access to chat GPT and put guardrails in and letting them use generative AI. But so the question I would ask you is,

Do you, are any of the companies you talk to, would you say, wow, this company, and you don't have to say the company name, but just an indication of what they're doing. Wow, this company is really ahead of the curve. They're actually have integrated generative AI in some kind of meaningful way in their workflow.

Yeah, so I'm not going to say that the AI use cases I've seen are, say, world-shaking use cases. But I do think there's been some really interesting work done with generative AI to...

parse invoices and to better understand contracts because every corporate contract is a mess. There's always 100 pages of cover your butt language in every single contract that you have. So how do you enforce these things and how do you figure out which terms are potentially most challenging?

Of course, you can read through it yourself and you can eventually get to them. But generative AI provides a decent shortcut for helping procurement and finance and accounting to work together to enforce contract terms.

And then also simply from billing perspective, especially when you've got some of these more complex spend categories where you can have a lot of vendors in a specific marketing area. For instance, in SaaS, the average enterprise at this point has like a large Fortune 500 enterprise has over a thousand apps.

that are in their market. And a lot of these invoices are not standardized in any particular way. They do not follow a set data format. So often it can be useful to use something like generative AI to automate some of the parsing of these invoices and find some of this information rather than have to manually check every single one. So that's another area where I'm starting to see generative AI

being used to find things that would not be easy to do humanly. I would say that going, taking a step back, not just from generative AI, but from AI in general, there's a lot of interesting work with transaction matching as well, which is not necessarily new, but it's definitely something to keep in mind if you're still have an old school mindset of manually checking everything. There's a point that it's just not practical

to look at these thousands and millions of transactions and always try to audit and match by hand. Yeah, yeah. Reconciliations are an area where it's begging for even more automation than what we've seen now. It's funny that you mentioned the contracts because that was one of my first really good practical uses. I was doing some work for a public company that was headed into audit. They had made an acquisition. It was like a 170-page contract.

asset purchase agreement. And there were just so many details. And the auditors, as auditors do, had a million questions. And nobody could get a handle on treatment of deferred revenue. And it was mentioned in multiple places. But I think about using the viewer, a PDF viewer, and trying to search...

for keywords or whatever, and just how bad before generative AI it was. And this might have still been in the 3.5 era of chat GPT. So you couldn't upload. Maybe there was a cap on how big a document you could, but I found something called like chat PDF that has now gone away because I assume they're gone because you can do it all in the frontier models. And it was a lifesaver. I can remember it was like 8.30 at night. And I thought, if I have to read...

flip through this entire 170-page APA. I'm never going to find it and I'm going to lose my mind. You plug it in there and you just ask in context and getting the relation and then, you know, tying it to accounting rules. And I mean, it was just...

It's weird, like trying to find those sort of enterprise-wide where we just inject generative AI in our workflow across the company. It's hard to get to those use cases right now, but really the power of generative AI, I think is when you put it in the employee's hands directly and they're able to automate

or get better at their specific job. They have their own use cases. You can't dictate the use case, but they'll find ways to use this new superpower. Is that kind of what you're saying as well? Yeah. The phrase that I'm using most often, I find myself using most often right now is,

to look for the work that your workers hate to do. If you just hate doing work, you know, that's one issue. But most of us like doing certain types of work and hate doing other types of work. And it's the kind of work that you hate doing that often is well-suited to some sort of automation because you hate doing it, not because it is

bad work, but because you feel like your time could be better spent doing something else. So actively look for those types, that use case. Talk to your coworkers and your employees about the work that feels like it is just an extreme waste of time or really difficult to do over and over and over and over again, but has to be done for compliance or simply doing a good job.

Yeah. You know, it's when I think about, because, you know, obviously people ask me all the time and of how generative AI can be applied in finance. Like we're talking a lot about sales and marketing in this episode, but it's easy to find use cases there because it's a large language model. It's the language and it's, you know, that's the purview that it works in and makes the most sense. But when you're doing, whether it's accounting entries or financial statement analysis or whatever it is, you have to, you know, find ways to step in there. But I think

that there are applications to it. And to me, the biggest application is, we talked about democratization of data for years. In a way, generative AI feels like

democratization of data science. And what I mean by that is you don't have the barrier of, you mentioned, you know, learning SQL and or learning Python. And the way that you used to have to access data science was through these computer languages, that that's a huge barrier to entry. But now if you can just in plain language interact with your data, that's a potential breakthrough.

Yes. Part of me thinks about all this time I spent learning how to build reports, learn SQL, at least have an idea of what's going on with Python and R. And now I don't actually need any of that because I can tell ChatGPT.

to do it or to write me a quick script. And of course, I wouldn't have ChatGPT write an entire app for me, but it can write specific scripts or specific applets that have some sort of specific functionality.

and or translate the code into or back into English for me. And that actually saves me a lot of time as well, because then I can have this little functionality that I can use from a data science perspective. But also, if you're just trying to do applied data science, now you just ask

the model to do a thing and it just does the thing. And it is completely impossible for the end user to tell that that actually took a lot of work to figure out. It would have taken a couple of weeks to do through a human and to open up a ticket and to get this app written and to actually get the result that you're getting.

Yeah. You know, the funny thing, if I'm an IT guy right now, I'm freaking out about citizen development because it's making it so much easier. And if my IT department, if we're not shipping what our internal clients are asking for, they're going to just go out and start blowing up collab projects and just...

having these systems build code for them and doing stuff on their own, which, you know, as someone who throughout my career was very guilty of citizen development, you know, thinking of this from an IT perspective, that's pretty scary to think about what's going on with your data and what are they uploading this in and how is it being used? Because it's just so easy now. Yeah.

Yeah, it's potentially a gigantic compliance and security problem because the problem isn't that these apps won't work at all or that the prompts aren't being asked correctly, but that the people who are doing all this stuff in AI, line of business people, the capitalists,

citizen developers don't necessarily know what issues are out there from a security and compliance perspective. So that's the part I worry about most. And I think that there's going to be all these agents and scripts a year from now that are just abandoned and create a new type of technical debt for us that

I don't even know what it's going to look like yet. But I do know that when technology gets abandoned, that we always have some sort of problems that come along afterwards. Yeah. If your controller is out there posting stuff on GitHub, not being a developer,

And just, you know, not having the control. I mean, who knows? Or just even well-meaning, hey, how do I do this thing? And even in asking the question, you've just accidentally uncovered some backdoor for your own corporate data. Yep, yep. You know, you mentioned earlier people going back to school for data science and for machine learning, engineering and all that. So the interesting thing in trying to figure out if I'm, you know, if I weren't the geek that I am and I were just starting out in my FP&A career, I think...

obviously domain expertise in finance is important, but also having that, you don't have to learn to code a lot, just know enough to sort of recognize it and understand what's going on. But having domain expertise in data science, no matter what field you're in, I think is going to be important because now if you can access and you can be a data scientist who doesn't have to be a great coder, if you know the right questions to ask,

your data-driven decision-making and your ability to generate reports and find correlations and do things with the data, you know, you have a superpower now, but you have to know sort of the rules of big data. And, you know, does your data even classify as big data? And if so, and if not, what that means and how to do the analysis. Yeah, and I think that there's even a further sense that if you know what questions to ask of the data, you can then tell,

all of your other financial stakeholders, including your line of business stakeholders, here's the type of questions you can ask your data and then bring that feedback back to me because I know what to do with that. And if you just ask your data how to summarize your costs this way or all of your expense reports in this format, then you don't have to do the additional work of formatting. And I get exactly what I'm looking for because I have just shaped this

prompt or this quote or this query exactly how I need it to work with the rest of my budgeting challenges. I think that's one of the big advantages to knowing how the AI works. You can then start creating these even just citizen prompts that are better aligned to your financial department and your own governance needs.

Yeah. And, you know, thinking about that, if you we are at a place now because we're early in the technology where I hate the idea of having to have a prompt library. But truthfully, like there are certain ways that you need to interact with the data. So there's information that the GPT needs to have.

before it can respond to you. So you kind of do have to save like, okay, we're using gap accounting here. We need this rule. We need the response like this. So these prompts get long just by setting up to make sure that you're not getting a hallucinated answer and getting it in a format that you want. And so it's clunky now. It can still be efficient, but we're at the bleeding edge of this if you're using this now. I mean, how do you see, like, and to me, that's an adoption hindrance

is when you have to come out of your ERP or your CRM and put this data into an external system to manipulate it and then come back and then have the data come back to you in a format. How do you see, what's it going to take to overcome that hurdle where the generative AI is just in our SaaS tools that we're using today? So we're starting to see something called agentic AI show up. And

Two things, agentic AI and computer use AI, where AI is starting to figure out how to click around on your computer and go through websites and be able to get to a response. So basic consumer version of this might be you use computer use AI and you ask it to

order Cheetos on Amazon for you. Not a fantastic example, but just saying that based on this, if your computer use AI is set up correctly, it can go through, it can read, it can type in Cheetos in the search button

bucket and then click on the Cheetos, click one, click order and have it come to your house. Still pretty rudimentary, but having an AI that can do that level of work on your computer or on your website is a starting point because then you can figure out which fields on the website

or in your system need to be aligned to whatever target system or your financial system perhaps is in place. So I think you're gonna need some of these autonomous agents to do some of the backend work of mapping

one piece of data to another piece of data or one set of fields to another set of fields and do some semantic mapping in the middle. We're talking about some pretty sophisticated stuff, honestly, to get to that point of making that and to make your scenario a more realistic, automated, AI-powered process at this point.

FP&A Today is brought to you by DataRails, the world's number one FP&A solution. DataRails is the artificial intelligence-powered financial planning and analysis platform built for Excel users. That's right, you can stay in Excel. But instead of facing hell for every budget, month-end close, or forecast, you can enjoy a paradise of data consolidation, advanced visualization, reporting, and AI capabilities. Plus, you can stay in Excel.

plus game-changing insights giving you instant answers and your story created in seconds. Find out why more than a thousand finance teams use Data Rails to uncover their company's real story. Don't replace Excel, embrace Excel. Learn more at datarails.com.

So I got to tell you, Operator was, that's OpenAI's computer use agent for our listeners, was the final nudge I needed to upgrade to the $200 a month OpenAI. And here's why. I'm not using this for any kind of client data. It's still too early. I'm

experimenting like crazy with it to see how much I can push it. But for those, for our listeners, Operator is the agent that Young was talking about. If you've ever seen RPA at work, where the mouse is magically moving across the screen and doing things on your behalf, that's what Operator looks like. And I will say, this is pretty interesting to me from an automation standpoint going forward. Because if you've ever worked with any of these RPA companies, it was

cool, but rule-based. And the implementation was very expensive, took a very long time, very, very picky, could slide off the rails. You'd go and correct it and it would learn and come back. But it was very difficult to do RPA. Now the operator is...

individual RPA. So think about if every employee had a tool that was, you know, that was more reliable. We're in alpha testing phase, but if every employee had their own RPA agent or agents, plural, the potential there is pretty interesting.

Yeah. And so I'm looking forward to that and then the agentic part of having these AIs being able to actually finish off specific tasks. Of course, you're going to have to test that out before you make any AI-based agent autonomous.

but because that has to be done based on a variety of scenarios and a variety of language prompts before you really let it loose on its own. But I think there's some potential there.

It's interesting to me, it seems counterintuitive, but when there's APIs and connectivity that the computer use model and where the AI is moving the mouse around and interacting the way a human would, it seems counterintuitive. But when you think about it, it's the same reason that robots are being built in a

human form factor. It's the world that we operate in is designed for humans. So 10 fingers, two arms, you know, six feet roughly and you know, whatever weight, I mean, yeah. Yeah.

So to navigate the world, then it makes sense that if we're going to build a robot to do that, it should have that shape. And the digital world, yes, a lot of sites have APIs. And if there are APIs, that's great. But think of all the data that's out there and what's on the web, you're not going to have APIs and backdoor access to all these systems. So if it operates

the way a human does. It's going to be interesting to see how the web itself evolves if there's another, the regular web, the dark web, and then the agent web where they're just going through and maybe something is optimized there. But, you know, the amount of traffic when the agents and computer use really catches on, it's going to be interesting to see what it does to the web in general. Yeah, but funnily enough,

But fundamentally, I think one of the big differences now is that for most of our work lifetimes, we've had to learn how to think more like computers. We have to put our data in the right spreadsheet cell. We have to access an API where we learn some schema that is based on a piece of software. So we have to think like the software to make things easier.

And now we finally kind of flipped that on its head a little bit and said, okay, now computer, you got to meet me halfway. You got to start learning English and doing this the way that I would phrase it or at least be able to go through the website the way that I would go through it. And if you're failing, this shows.

that you're not ready to do work yet and kind of holding the computer to that standard of basic human competence from a user experience perspective. I think that's an interesting flip that has happened, that we are starting to hold the computer responsible to understand a human UX rather than forcing the human to be in the computer UX. Yeah, yeah, very, very good point. I hadn't thought about it that way, but that's absolutely right.

So you and I, this is every time we talk, I feel like we should just be in a think tank, just, you know, some academic ivory tower. We're just brainstorming the future. But for a lot of our listeners, you know, they're looking for more practical applications. And I always, I struggle with this question. So I'm going to,

I'm going to throw the ball in your court and ask, you know, you and I have just talked about what we see for the future. And I think we're probably right. Who knows what the timeline is on it. But, you know, right now, CFOs and people are getting pressure to, you need to implement AI. You need to use this. I mean, what is, where should finance teams be investing in AI right now? And what should they expect in an AI implementation? We talked about, you know, the

generative AI kind of where it is. We haven't really talked a lot about sort of the classical machine learning and what you can kind of the deterministic and these algorithms that have been, we've been using for 15 plus years. If I'm given a directive and a budget to implement AI for the finance team, what should I be thinking about doing today?

Yeah, I think definitely consolidation is a starting point. I do think that we need to go back to zero-based budgeting and planning, budgeting, and forecasting. Part of the reason that those have been challenging is that it is hard to get to this constituent starting points and really figure out how to build that all from scratch without a whole lot of manual work.

I think AI can be a way to fill in some of those gaps a little bit more easily and also to build more of a portfolio of planning of plans that could be dependent on what if situations that

are a little bit more holistic than simply changing a single cell or a single unit. You can ask broader questions with AI than you can with simply doing basic what-if analysis.

For instance, you can ask generative AI something like if the business is more aggressive, how will that affect the budget? And it will give you a number of suggestions of key spend areas or key revenue areas where you might

see significant changes based on what typically happens in the world or what typically happens to a business. So it's not going to be perfectly aligned to your business, but it will at least provide a starting point for you to say, oh, well, I hadn't really considered that this business unit is unusually volatile and dependent

on having a strict business execution. And that might be a problem if we tried to get more aggressive or if we cut off funding a little bit because we have to push money from one place to another. Or maybe you have one business unit that is like groceries where you have to deliver every two weeks or your stuff goes rotten. Just little things like that that can be pointed out quickly through generative AI that

may not pass the test of simply manually looking through each category of what you're looking for, especially if you're a large conglomerate or a multinational, or if you have experimental areas, or if you are using interesting names for your business units, all sorts of reasons that you might not necessarily pick up on something that generative AI would pick up on.

I think there's some strategic value to be able to use generative AI like that as well at a time when the CFO is being asked to be a strategic officer in more and more companies. And this one, and this is what you and I, the webinar we did together was about, and I'm guilty of this, but I think with reason, but we still need to try to figure it out. But the ROI for an AI investment right now, and how do you account for that? I'm not seeing...

wholesale swaths of jobs being wiped out by AI. So if you're trying to use efficiency and replacing employees, I mean, yes, I think automation will, there will be maybe if you used to have 10 people in your accounts payable department, maybe you can do the same job with three in the future. I could see that happening, but you can't, if that's your only ROI, it's not today and it's a way out and it's speculative.

But there's also, you know, the kind of insights you were just talking about. There's a value to those. It's hard to put what that value is, but the whole idea of, we've been talking about digital transformation for 30 years, but the whole idea of being in a transformed company that is making data-driven decisions is,

there is a value to be put on that, that it's hard to factor into just an ROI calculation. But what, I mean, and because you help companies with their budgets around this kind of stuff, what's your approach to find, to ROI for AI investments? Yeah, so I know a lot of the,

vendors try to dance around labor replacements and say that their AI is about helping employees to be more productive, I'm going to be honest, there are a bunch of line level jobs that are going to be taken out by AI. And the people who are going to lose their jobs are the people who only know how to do the job itself, but don't know why they're doing the job.

Whereas if you're an AP or AR and you know why the payments are being processed, and you have an idea of how that fits into cash flow, or you have an idea of how this fits into people being able to operationally buy what they need to do their jobs. If you have that next step of logic, you're going to be able to use the AIs and

do whatever the next version of your job is. If you think your job is simply to scan the invoice into the machine and click the button and enter the number into the software, yes, you are going to lose your job.

So that is one point of value. But I think of production, that productivity as kind of a red herring because it's a zero sum game. You don't actually make more revenue by being more productive. You just change your margin a little bit, but it's not really a growth exercise. I think it's much more interesting to think about the potential growth associated with bringing AI into place, what you're able to pursue, what you're able to do differently as a business

hopefully a ways to be able to cut areas out of your production or your supply chain or manufacturing efforts to be able to move forward. And I know not all of this is directly related to the finance function, but I think it is something that finance people should keep in mind because finance will be asked to be a stakeholder or a gatekeeper for budget.

for AI projects. So you need to start identifying where the value does really exist and to be able to push back on AI projects where people can't articulate where AI actually provides growth across the business. I think that's probably going to be the more important part for finance, looking at AI in the near future. It's not necessarily how AI directly affects the finance department, but knowing that finance will be one of the departments that

decides whether an AI project will move forward by looking at how legitimate the ROI is. And I think it's really important from that perspective to think about the growth and the revenue at the revenue chain and the value chain that is associated with the business actually making money and where AI fits into that. And that is usually not cutting five seconds off of a specific job, unless it is a job that is done

a hundred times a day by every employee in the company. Usually the improvements will come from things like supply chain risk or manufacturing improvement or things that are more tangible or logistics. Things that will actually affect how much gas you use, how much power you use, the SKUs that you have in place, those nuts and bolts things that actually run your company.

Very well said and agree with all that. So, all right. So as always, I think you and I could just keep going all day together, but I need to start bringing this home. I'm going to switch gears a little bit. And the first question I'm going to ask you, it's because I get asked this all the time and I'm always wrong. I'm a terrible futurist. I love talking about the future, but if you ask me to put a date or a guess on anything, I'm no Ray Kurzweil. Yeah.

So I guess I'm going to ask you, how about a bold AI prediction? Oh, I got the bold book right here. Yeah, I still have it behind you. So a bold AI prediction for this year, for 2025.

I think at least, I know this doesn't change the market as a whole, but I do think there's going to be at least one startup that shows this year that is successful. That is basically just one person and thousands of agents. And people are going to wonder what the hell is this because there are no employees. It's just one person and it just drives 50 million or a hundred million in revenue. And it's going to come out of nowhere. And it's

It's going to end up being this case study for people to look at for decades. People or agents to look at for decades, maybe. So, okay. Well, so now I'm going to bring it home with our two questions that we ask all of our guests. So the first one is, what is something that most people don't know about you?

I'll say that although I pointed out that I was a women's studies major in a college, one thing I actually did as part of that was I spent my junior year at a women's college. I was the only male student there. This has nothing to do with, I guess, the politics or transness or anything. I'm just a normal cis male dude who was at a women's college with 2,200 women. So that was a really interesting and challenging year of my life. I bet. Yeah.

If you've ever lived with a group of women, multiply that by 500. Everything you dealt with there. And that's what I dealt with.

Wow. Yeah, that had to be amazing. So that is very interesting. All right. And the last question that we ask all of our guests is, what is your favorite Excel function and why? You know, I'm always going to have a little piece of my heart dedicated to the B lookup. I'm with you, brother. I'm right there with you.

VLOOKUP was the first time I felt like I was an Excel pro. It's like I have the power to find things in a big array and I'm not just dealing with B2 plus C2 equals D2 anymore. It's like I now control this entire

tire spreadsheet. I know what's going on in it. And, you know, I've got these V lookups and H lookups and I can figure out what is conditional, what is not. Like, I can do all this stuff now. Like, I have the power. All started with the V lookup. V lookups and then pivot tables right after that. Oh, yeah. Like, whoa. Yeah.

Love it. Love it. All right. So finally, how can our listeners connect with you and learn more about Amalgam Insights and follow you? What's the best way for them to do that? Yes. So I both have a website, amalgaminsights.com, as well as I do a weekly podcast called This Week in Enterprise Tech, along with Charlie Araujo, who is the Chief Strategy Officer at Symphony AI. And we talk about the biggest AI

AI news topics of the week and how that is affecting the CIO and CFO offices. Awesome. We'll be sure and put links to that in the show note. And now I've got another podcast to add to my already busy podcast listening schedule. So I'll put you up there with Scott Galloway and Kara Swisher and the Hard Fork guys and all. Yeah. Hyun, thank you so much for coming on the show. Really appreciate your time. Yeah, it's been a pleasure.

♪