
Loading summary
Juan Orlandini
Foreign.
Podcast Host
Welcome everyone to the Emerge AI in Business podcast. Today's guest is Juan Orlandini, CTO of North America at Insight. Juan joins us to discuss how finance leaders can move beyond manual data mazes by using AI for compliance heavy lifting. Our conversation also explores optimizing workflows through data engineering to clear data swarms and leverage leveraging existing SaaS tools to ensure mathematical accuracy and long term ROI. Today's episode is sponsored by K1X. Please note that the views expressed by Juan Orlandini do not necessarily reflect the official position of Insight or its leadership Position your brand alongside the Fortune 500 leaders defining the enterprise AI roadmap with the opportunity to showcase your solution to the executives currently funding and scaling global initiatives. Partner with Emerge to reach the decision makers holding the strategic mandate. Secure your partnership@go.emerge.com partner that's G O.E M-E-R-.com B A R T N E R let's dive into the conversation with Juan.
Interviewer (Sam)
Juan, thank you so much for joining me on Emerge's AI in Business podcast today.
Juan Orlandini
Hey, thank you Juan. Thanks for having me.
Interviewer (Sam)
It's great for us to have you as well. So our discussion today is going to be very relevant to the finance guys and I think you all people would agree that in a highly regulated area like tax and financial reporting, the data lives in chaotic states. We've got PDFs and we've got spreadsheets and we've got legacy portals that should have been phased out years ago already. And I know that you've written about mistakes that businesses make when deploying AI enterprise apps, one of which is trying to build on AI to a mess. So we're starting with the mess and we're just putting the AI on top of the mess. How should a CFO look at their architecture so that they aren't just throwing more people at at a manual data problem, but actually building a system where AI handles the heavy lifting of compliance.
Juan Orlandini
So that's a really good question. And so you know, one of the things that I always caution CFOs and any anybody that's in the financial side of any organization is that these generative AI models, which is what most people talk about today when they talk about AI, they're really not good at math. They're terrible at it. In fact, if you ask them to do some basic math stuff, they will very much sound convincingly true and giving you an answer, but they're truly giving you a statistical response. They're not giving you math. All Right. And what we live in in finance is math. All right? So you need to be aware that when you're using these tools, you have to put them in the right place where the superpowers that they have operating the constraints of math and the things that you as an organization have built to make sure that the math works out, at the end of the day, you have a balance sheet, you have, you have all the things that have been built in financial operations forever. Use those, all right? Don't, don't throw them out. Because these things tells you convincingly that it's got the right answer when it matter, might not. You don't know. That's what they call hallucination. It's very deliberate use of the tool in the right place.
Interviewer (Sam)
Makes sense. So I know that for the cfo, automation always comes, or AI always comes with this promise of oh, your life is going to be easier and everything's going to be automated. But manual steps do creep back in. Why, from your experience, have you seen where the workload often result in adding more manual input or more people instead of just fixing the underlying architecture?
Juan Orlandini
Well, yes and no. Right. So absolutely it can, it can happen. All right, because of the verification that's inherent in the operations of any financial organization. And if you have a tool that comes in and it is maybe collecting PDFs from one source and CSV files from another and you're trying to ask it to reconcile this stuff and all of a sudden there's an error and that reconciliation, all of a sudden you might be asking your people to do not just double the work because they have to do that reconciliation, but then figure out how this thing went wrong in the first place, that they might not even know how it works. All right? So absolutely that can happen. And, and that's why I, I caution is make sure that if you're building an automation that it's in the places where these tools have matured. All right? And there's some places where they're amazing right now. Absolutely mind blowingly good. I'll give you a personal example and this will give you an idea of the crazy superpowers that you can leverage these tools. I'm a frequent flyer, I travel all over the country. I'm on a plane pretty much every week and I have been doing this for 30 some odd years. So I'm multi million miler with Delta and all that other stuff. But me being a nerd, I love to see the places that I've been and Delta used to have this really cool App where it would show you all the maps and locations of where you live. And. But they took it away from their website and I was like, oh, well, but I'll build one. I vibe coded this thing, all right, by having it export a PDF of all of my transactions with Delta over the last 18 months. They give you this report without telling it much. I slapped it into one of these generative tools. I said, hey, parse this PDF document. Draw me map. And it did. First pass with no additional input, and all of a sudden came up this beautiful map with all the arrows and gold graphs and all that other stuff. Superpower. Amazing, right? But then I started looking, I started looking at that PDF file and I wanted to verify it because, again, I'm a super nerd, right? And it missed a couple of them. And I had to tell it, hey, you missed a couple. And they said, oh, my gosh, I did. But the first pass at it looked perfect. It was beautiful. It was amazing, right? So that's the kind of Superpower that you have to be kind of careful of, right? Is that it will do this amazing thing because it took that PDF file that I didn't even describe. I just said, hey, take this PDF file. So report from Delta shows me all the places that flown. Draw me a map. That's all. I gave it as instructions, and it worked all right, but it wasn't 100% right. So this is. We talked in another episode. Well, AI tools really should do two things, right? And maybe a third. All right, the two things that should do is either save you money or make you money. And then the third one is, particularly for finance leaders, is keep you from going to jail. And if you misstate, misstate your statements, right, There is a penalty there that might actually land you in the wrong place, right? So I would tell you, I'm not saying no, don't be scared of these things. Don't, don't not use them because they are a superpower. Just use them correctly and use the decades and actually centuries of experience that we have in accounting and finances to make sure that these things are leveraged in the right place for the right tools and all that.
Interviewer (Sam)
And obviously introducing these new tools and these new AI programs or capabilities, in the beginning, the process will start off slow. It's going to start off slow. It will have, like you just mentioned, teething problems, basically. Is there a way? It's. It's inevitable that's going to happen. But is there certain things that we can do to minimize the teething problems or Minimize the slow in the beginning and just certain things that we can look out for flags or notes to take.
Juan Orlandini
Yeah. So I'll, I'll tell you one of the things that I would tell Finance in particular is that you probably don't want to write your own, your own tool. All right? Part of the value that we get from working with our SaaS providers or our vendors that give us the tools that we use is that they have institutional knowledge that's baked into their tools that do the safety and controls and checks and all that other stuff. Right. And that gives you a certain amount of assurance and that if they give you an AI power tool inside of their thing. So think about like a workday or SAP, SAP or one of those kinds of things, those are operating under controls that are already inherent in the system. And that would be a great way for you to start seeing how these tools superpower your, your operations. Now, there are going to be things that are very intrinsic and native and unique and differentiated to your organization. We all have more. We're all slightly different. We're all snowflakes, right? We look at them and we're slightly different. Right. And you are going to have to, at some point create or adjust or modify these things to your specific way of operating. Learn first from something that is going to get you up the learning curve so that you know how this is going to affect your people and your processes, and then start adding things that are actually going to be very customized to your very specific needs. Things. It sounds simple, but, you know, we get so, so overwhelmed by these promises from all the vendors and all the noise and all the hype that's happening in the industry that sometimes you have to take a step back and recognize, hey, these things are going to change how I operate and what I do and, and the ways my people operate and my business operates. So let's be good stewards of that change. Not, not prevent it, but steward it into, into proper alignment to what we want to do.
Interviewer (Sam)
So I love that you said it sounds simple, but it's not that simple. And it actually does sound simple. But then let's dive into that part. What does a truly scalable AI ready operating layer actually looks like for a tech department? So if, if, if it's not that, what should it look like? If it's not that simple, how would we know if it's scalable and if it's ready?
Juan Orlandini
You know, we've actually done this for a long, long time already. And if you look back when computers first got introduced into finance operations. What they allowed us to do was to actually very efficiently process thousands, then millions and billions of records and add and subtract and all that other stuff. That, by the way, did not require fancy AI, right? It just required raw computational. Right. What we're introducing, and that hasn't changed, and you should continue to do that. Obviously you can't not operate that way anymore. Right. But what AI is building to you is ways of doing reporting that's different, or ingesting data that's different, or ingesting different kinds of data that give you better visibility and maybe reporting that was very difficult to create before. So in the past, creating a dashboard with specialized views of how data was organized and all that other stuff required very specific skills from developers, users, user experience people, finance people that knew what those numbers actually meant and whether they're wrong or right. Now you can actually create these dashboards by literally saying to these tools like, build me a dashboard that does these things and the tool will go and build that stuff. All right? So that's some very basics. Now, AI, generative AI, is built on one big premise, and that is that the data is there in a way that makes sense. All right? And most organizations live with a bunch of data lakes that are loosely connected into what I call a data swamp. All right? And it is not a very good place to be because you start to try, say you went through an M A kind of a thing, and also you have a data lake from your original company, a data lake from that other company. They might have data lakes from other companies that they bought and they never fully merged them. And that's how you end up with these data swap. All right? And now you're trying to do the reconciliation and all that other stuff is difficult. These AI tools can potentially help you with that. But recognize that what you have at the very core of this thing is a data problem, not an AI problem. At Insight, if I could look at my data scientists and my data engineers, the people that focus on working about worrying about those data swamps, I could probably keep more data engineers busy than I could keep data scientists busy. Because the data engineers are the ones that build the models. They do the cool science and experiments around around AI. The data engineers are the ones that actually make sure that those models are built correctly and then can report correctly, right? So make sure that as you're looking at this exercise, you're focusing your energies and on your data problem, that you have the right kind of data engineering people to support you that actually understand the Financial operations side of the house and work in your business unit that way. It's a cooperative effort.
Interviewer (Sam)
That also triggered something else that I thought of, is that okay, if I now commit to a vendor. We're starting out, we, we introducing AI into our finance department and we start off with a specific vendor. How do we make sure that we just. How do we make that decision? How do we make sure that we connect with the vendor that will not trap us into only evolve with them? Is there a way to make sure to, to just kind of insert this information or the platform or whatever it is, whatever this vendor is giving us will be able to connect to different architecture in the future? Is there, Are you trapped and that's it?
Juan Orlandini
No, you're never trapped. Yeah, it's a really good question and I'll tell you, it's actually it and it in general has a cycle and people tend to get excited at the very beginning of the cycle where there's this excitement about the capabilities of this company, this product, this tooling, whatever it is. And then that thing actually does bring value to your organization. So you jump in with both feet and you buy into it. But as part of that tooling, you typically change your operations to fit what that tooling brings to your organization. And then that technology matures and then it ultimately, eventually, always docks, becomes technical debt for your organization. Right. People forget that all, all of it is going to have that happen to it, right? There was a point in time when you could not be fired for buying IBM, right? Well, IBM is still around, but it's not that anymore. They're amazing. Nothing against IBM. There was a time when you couldn't get fired for buying Deck. And I'm showing my age, all right. And Deck is no longer even around, right? There was a time when you couldn't get fired for buying Sun. Sun is no longer around, right. And, and same thing for databases, same thing for financial applications, Same thing for any of software. They all are going to have this, this cycle that happens to them. What you have to recognize is that it just behaves that way. Make sure that the systems that you build for yourself are, are based on the assumption that you're going to take advantage of the value that they bring to you today. But don't get so married to them that when their tail end happens that you're going to go down with them, that you have the systems in place so that you can actually migrate, adapt, consume or whatever it is into a new technology. And this will happen with these AI things. It's just going to happen faster and faster because it's maturing so fast right now.
Interviewer (Sam)
And have you seen certain red flags that we can look out for? See these red flags with certain vendors that suggest a solution? And the flags are then if you had to say to your best friend, listen, this and this and this, I would look out for this. I wouldn't do that. Any, any red flags?
Juan Orlandini
Yeah, it's a really good question. And there's, there's a few red flags that always pop in my mind if the vendor is promising you to fix all of your problems because they have an amazing tool that is a danger signal. No tool fixes all problems. You have to actually know how to use that tool, make sure that it's the right tool and those things. And, and I've seen that over all parts of a business, whether it's in finance and operations and it doesn't matter, it happens across all of them. So if it's a tool first messaging that you're getting, absolutely, that's a red flag. All right. Now the other one that, that's a red flag is that because this generative AI space is maturing so rapidly, is there you're seeing a whole slew of startups that are popping up left and right. All right? And I've seen this happen over and over again. You, you buy into the vision of the startup and then that startup, they're doing startup things, they're trying to find more funding for the next round of funding or they're trying to figure out how to do an exit for their investors and then they do their exit and it gets consumed by somebody else and all of a sudden that thing just gets eaten up into some other thing that you don't want anything, any part of it. Right. So I always look at any AI startup to understand whether or not, hey, they truly do have a transformative technology, that the leadership team is actually mature and will actually be able to steward this organization into some sort of growth and that they do have a viable long term strategy that is beyond, hey, I'm gonna get acquired. Okay? Because at that point you don't know what's going to happen. Right. And if they do showcase at least some of that, work with your IT teammates, work with your, whoever does that vetting of vendors team, it could be your vendor management. Make sure that they've done some of those very basic steps. Look at that organization as a whole, not as a technology.
Interviewer (Sam)
I like that. Let's close with this question. So if you had to give finance leaders A tip or some piece of advice today. What should finance leaders do today to ensure that they can scale their data workflows without having to rebuild everything? How do they maintain long term flexibility today?
Juan Orlandini
So this is gonna also sound basic again, right? Invest in your people as much as you invest in technology, all right? Because our people are really our superpower. The tools enable our people to do the things that we do. All right? So make sure that the, the teams that you're building, that you're educating, enabling with these technologies are also adapting and recognizing that this is a new way of operating and these are new tools that help them do their job. Not doing their job has helped them do their job. The nature of their job might change. Right? But it is helping them do their job. And if you invest in your people as much as you do in your tools, you will be able to adapt because people are amazing. And that's the way I would look at it.
Interviewer (Sam)
Oh, I like that. So from our discussion today, what I've gathered is, first of all, don't end up in jail. So know your regulations.
Juan Orlandini
That's right.
Interviewer (Sam)
That's very important. And then you said start slow, start with clear, usable data. Make sure that everyone understands the change, embraces the change, accepts that it will continue to change. And then I think also just make sure that you align with a vendor that understands your vision and your challenges.
Juan Orlandini
That's right. That's right. And invest in your people. Don't forget to invest in your people. Super important, right?
Interviewer (Sam)
That's 100% message. Thank you so much, Juan, for sharing your insights with us today. I cannot wait for our audience to just get into this and I hope to speak to you soon.
Juan Orlandini
I appreciate it. Thank you.
Podcast Host
Wrapping up today's episode, the three key takeaways for finance and technology leaders from our conversation with Juan. First, recognize that generative AI is statistical rather than mathematical, meaning it must be constrained by existing financial controls to prevent hallucinations and ensure accurate reporting. Second, prioritize data engineering over data science to transform data swamps into clean, usable environments that allow AI to function effectively. And finally, avoid vendor lock in by viewing all IT as a temporary cycle and investing in your people as the primary superpower for long term operational flexibility. Position your brand alongside the Fortune 500 leaders defining the enterprise AI roadmap for the opportunity to showcase your solution to the executives currently funding and scaling global initiatives. Partner with Emerged to reach the decision makers holding the strategic mandate. Secure your partnership@go emerge.com partner. That's go emj.com p a r t N E R For further executive level analysis and to join our network of leaders delivering workflow impact with AI, visit emerge.com on behalf of the team at Emerge. We'll see you on the next episode.
Interviewer (Sam)
Sam.
The AI in Business Podcast: Scaling Regulated Data Workflows Without Lock‑In
Episode Date: April 17, 2026 | Guest: Juan Orlandini (CTO, North America at Insight) | Host: Daniel Faggella
This episode focuses on how finance leaders can scale data-rich, highly regulated workflows using AI—without falling into vendor lock-in or operational chaos. Juan Orlandini shares practical strategies for leveraging AI and data engineering to boost compliance, ensure accuracy, and future-proof organizations. The discussion centers on real-world pitfalls in AI adoption, the importance of data architecture, tips for evaluating vendors, and the central role of people in driving successful AI transformation.
AI’s Limits for Finance:
Juan emphasizes that most enterprise AI—especially generative AI—relies on statistical predictions, not true math, which creates risks when used for functions that demand strict accuracy, like financial reporting.
“These generative AI models...are really not good at math...they will very much sound convincingly true...but they're truly giving you a statistical response...and what we live in in finance is math.” (Juan Orlandini, 02:08)
Proper Placement of AI Tools:
AI tools must reinforce, not replace, established financial controls and processes:
“Use the decades and actually centuries of experience that we have in accounting and finances to make sure that these things are leveraged in the right place.” (03:29)
Manual Steps:
Generative AI “Superpowers”—and Pitfalls:
“I slapped [the PDF] into one of these generative tools. I said, hey, parse this...Draw me a map. And it did....But it missed a couple of them...The first pass looked perfect...That's the kind of Superpower you have to be careful of.” (Juan Orlandini, 04:53–06:18)
Three Key AI Outcomes in Finance:
“AI tools should do two things, right? Save you money or make you money. And for finance leaders: keep you from going to jail.” (06:50)
Minimizing Early “Teething Problems”:
“Learn first from something that is going to get you up the learning curve...and then start adding things that are actually going to be very customized to your specific needs.” (Juan Orlandini, 08:43)
Lessons from Legacy Systems:
Data Lakes vs. Data Swamps:
“Most organizations live with a bunch of data lakes...loosely connected into what I call a data swamp...What you have at the very core...is a data problem, not an AI problem.” (Juan Orlandini, 11:14)
Data Engineering Over Data Science:
The IT Cycle and Technical Debt:
“Make sure the systems that you build...are based on the assumption that you're going to take advantage of the value...today. But don't get so married to them...that when their tail end happens...you can actually migrate, adapt, consume...a new technology.” (Juan Orlandini, 15:04)
Red Flags in Vendor Selection (16:12–18:25):
Over-promising Tools:
“If the vendor is promising you to fix all of your problems...danger signal. No tool fixes all problems.” (Juan Orlandini, 16:34)
Startup Instability:
Assess whether a vendor/startup is focused on building lasting value or just seeking an exit:
“I've seen this happen...you buy into the vision...then that startup...does an exit and gets consumed...you don't want any part of it. Look at that organization as a whole, not just as a technology.” (17:51)
People as the Superpower:
“Invest in your people as much as you invest in technology...The tools enable our people to do the things that we do...If you invest in your people...you will be able to adapt because people are amazing.” (Juan Orlandini, 18:46)
Cultural Change:
(For more insights or to connect with leaders in AI adoption, visit emerge.com)