How FP&A Professionals Use AI to Move from Reporting to Strategy with Carolina Lago
FP&A & AI Software Showcase: Explore Leading Tools
FP&A leaders are embracing AI, and we’re here to help. On May 21st, I’ll host the FP&A & AI Software Showcase, featuring FP&A planning and AI tools like Concourse, Sapien, Drivetrain, and Una- AI. Join us for live demos, insights, and no sales pressure.
Register today: www.thefpandaguy.com/fpa-software-showcase
Onto this weeks episode
In this episode of FP&A Unlocked, Paul Barnhurst speaks with Carolina Lago, founder of Tactic Financial, about the impact of AI on financial planning and analysis (FP&A). Carolina shares her extensive experience in the industry, discusses the power of AI in transforming finance workflows, and explains how AI is amplifying the work of finance professionals rather than replacing them. She delves into her use of AI agents, like Claude, as thinking partners in her work, and gives practical advice for finance teams on leveraging AI for strategic decision-making.
Carolina Lago is the founder of Tactic Financial, where she helps FP&A professionals integrate AI into their workflows through hands-on training and practical tools. With over 20 years of experience in financial planning, analysis, and modeling across multiple industries and continents, she brings a unique perspective on how technology can amplify (not replace) the work of finance teams.
Expect to Learn:
The evolving role of FP&A as a bridge between strategy and operations
How AI can reduce manual tasks and enhance strategic thinking
The importance of understanding workflows and processes in FP&A
Carolina's approach to using AI as a "thinking partner" to amplify human skills
Here are a few relevant quotes from the episode:
"AI is not going to replace humans; it’s going to improve how we work and give us more time for strategic decision-making." - Carolina Lago
"The differentiator for companies will be how they use AI to complement human intelligence, not replace it." - Carolina Lago
Carolina explains that AI is not here to replace humans but to enhance their capabilities and free up time for strategic decision-making. She also emphasizes that understanding workflows and the human factor will be key in effectively integrating AI into finance roles.
Follow Carolina:
This week's Guest Carolina has launched a hands-on program for FP&A professionals who want to move from manual processes to AI-powered workflows. Learn to use Claude for FP&A:
Course: https://hub.tacticfinancial.com/accelerator?utm_source=fpaguy&utm_medium=referral&utm_campaign=accelerator
LinkedIn: https://www.linkedin.com/in/s-carolinalago/
Earn Your CPE Credit:
For CPE credit, please go to earmarkcpe.com, listen to the episode, download the app, answer a few questions, and earn your CPE certification. To earn education credits for the FPAC Certificate, take the quiz on earmark and contact Paul Barnhurst for further details.
In Today’s Episode
[00:00] – Trailer
[02:50] – FP&A : Strategy & Operations
[06:43] – AI in Reducing Manual Tasks
[10:52] – Transforming Finance with AI
[14:26] – Learning AI & Human Judgment
[19:37] – AI in Finance: Real Applications
[25:57] – The Future of AI in FP&A
[33:12] – Elevating FP&A with AI
[39:45] – Adopting AI in FP&A
[45:30] – AI for Better Decision-Making
[50:15] – Getting Started with AI
[54:05] – Final Thoughts on AI & Collaboration
Full Show Transcript
Host: Paul Barnhurst (00:00):
Are you tired of being seen as just a spreadsheet person while others get a seat at the table? Well then welcome to FP&A Unlocked where finance meets strategy. I'm your host, Paul Barnhurst, AKA, aka bearder wonder, I mean the FP&A guy. Each week we bring you conversations and practical advice from thought leaders, industry experts and practitioners who are helping shape the role of FP&A in today's business world. Together we'll uncover the strategies and experiences that separate good FP&A professionals from great ones helping you elevate your career and drive strategic impact. Today's guest is someone who's been on the show several times, Carolina, Lago, Carolina, welcome to the show.
Guest: Carolina lago (00:49):
Thank you for having me
Host: Paul Barnhurst (00:50):
And everybody will laugh. I've had her on the show and how many times have we chatted now? Probably 10,
Guest: Carolina lago (00:55):
Maybe. Maybe a little more
Host: Paul Barnhurst (00:57):
Unofficially and had your name wrong for two years and you didn't correct me. How did this happen?
Guest: Carolina lago (01:03):
I never noticed, but I have to say that Floyd also gets wrong. So I tried to correct because whenever he says my name, he says Carolina and I tried to correct and then he started writing Carolina with double E. So I gave up. I give up. It's all wrong. There's a lot of people that get it wrong, but it's Carolina actually.
Host: Paul Barnhurst (01:27):
Well, there you go. Secrets out of the bag out of wrong for two years. What I really took away from that is she said, Claude made a mistake and she made sure to mention that Claude was a man. It's him. Did anyone notice that? I did.
Guest: Carolina lago (01:40):
He has the name of a man, but I have to say when I'm talking, I have one agent. Let me say that I have one agent that is my thinking partner. This is the one that I talk about strategy and I put up strategy together and it has access to my calendar, to my Gmail, to my tasks, my notion, my second brain, which for me it's very important. I am a second brain adopter from Tiago Forte, if you've ever heard about it, it's the second brain methodology and I docked it like seven years ago, and so I have a whole second brain on the ocean. So I gave access to this Claude agent. That one is a woman I call her. Whenever I talk about her, I say, I'm going to ask her. I'm going to ask my Claude. I'm going to ask her if I can do this today or maybe it's better next week or if she manages my schedule. All of that. That one, that specific one is a woman. I don't know why, don't ask me why. Maybe because of an extinction from my brain. I don't know.
Host: Paul Barnhurst (02:50):
So there you have it. We'd love to hear from you, Claude. Man, woman, it pet. What are you? Treat it as.
Guest: Carolina lago (02:59):
I try to do it. Sometimes it comes to him, but this one, this specific agent is a one for sure. This one is an extension of my brain, my thinking partner, which I highly recommend.
Host: Paul Barnhurst (03:14):
Speaking of partner and man, we interviewed somebody the other day for future finance, one of my other shows. So I'm going to cross promote for a minute. You'll want to listen to this episode. The guest, his hot take, was one day in the future, agents will have the right to vote.
Guest: Carolina lago (03:31):
Wow, never thought of this.
Host: Paul Barnhurst (03:34):
Yeah, I hadn't thought about it either. I'm like, I don't see that. But it was very interesting and he had his reasoning and logic behind it. So there you go.
Guest: Carolina lago (03:42):
I never thought about this. I've seen people
Host: Paul Barnhurst (03:44):
Baby. So we vote as a man or a woman if they can vote.
Guest: Carolina lago (03:48):
True.
Host: Paul Barnhurst (03:49):
Alright. Probably not what people came to the show for, but hopefully they got a little enjoyment of our fun little conversation there. So Carolina, give me a little bit of background and then we'll jump into some questions here. And she's the founder of Tactic Financial where she helps FP&A professionals integrate AI into their workflows through hands-on training and practical tools. She has over 20 years experience in FP&A and modelling across multiple industries and continents. What we got, South America, North America and Europe. Do we have Asia in there yet? No, it hasn't done Antarctica, but it's on our list. She brings a unique perspective on how technology can amplify not replace the work of finance teams. She partners with organisations like the Financial Modelling Institute, FMI highly recommend them and speaks regularly at global events, including the FP&A guys podcasts. She talks on AI agents, financial modelling and the future of fp and a. She also runs the tactical room, a newsletter and community for finance professionals navigating the intersection of AI and finance and her brain or AI brain is a woman. Did I cover it?
Guest: Carolina lago (05:07):
Yeah, you did. Exactly.
Host: Paul Barnhurst (05:13):
Alrighty, so first question I ask every guest before we jump into ai, and I know we're going to go deep there and try to give some hands-on practical advice today. What does great FP&A look like when you think of what great FP&A is? How would you define it? What comes to mind?
Guest: Carolina lago (05:30):
I think FP&A is the through line from strategy to operations. It's the one thing that connects the strategy and operations and that's financial planning and analysis that didn't change. It's going to continue to be like this and it's just going to get stronger now because now we have more tools to make that happen. Does it make sense that probably before we were already entitled to be that through line to be that connection between strategy and operations, but to be honest, we were overloaded. It's so much analysis that we had to do that it was almost impossible to cover everything. What AI brings to the story now is that now it's possible. Now we can amplify what we can do and I'll get to that later, but it's actually the essence of what AI is bringing to the table right now.
Host: Paul Barnhurst (06:28):
Yeah, so the big thing you talk about is FP&A really should be that strategic connector from a financial perspective between operations and finance for the broader business, but historically we've struggled. There's so many tasks that have taken up our time,
Guest: Carolina lago (06:43):
Too much analysis, too many possibilities, too many decisions taking on the bit of the way, and most of the time what fp a was doing is reporting because once we finish reporting, it's already time for another cycle.
Host: Paul Barnhurst (06:56):
See, I think I spent most of my career cleaning data, but probably just from one
Guest: Carolina lago (06:59):
Nightmare. Cleaning data. Yeah, cleaning data and then reporting and then explaining and when you see it, there is another cycle coming up and you have no time to influence. You had no time to make decisions. So having a seat on the table, having your voice heard most of the time, finance professionals, FP&A teams are small and we just don't have time to cover everything that we should be covering and that's where I get so interested.
Host: Paul Barnhurst (07:31):
You've leaned very heavily into AI in a particular Claude. What's motivated that? What's forced you to spend so much time learning it? I know you've spent hundreds of hours there.
Guest: Carolina lago (07:43):
I've been a heavy user of AI since I came up. The thing is, between what AI could do back then and what AI can do right now, there was a huge difference, a huge gap. So now I see more possibilities ever since CLA became the system cloud code, cloud code work ever since they started to lean heavily on finance as well, I've seen a lot more possibilities of using AI in the real world. So it's not only a chat bot that you're just talking to and getting instructions on how to do things. It's actually doing things for you the way you could do. You would instruct AI and use the intelligence of AI to do things for you. So now there is actually a good use for it. So I see many possibilities.
Host: Paul Barnhurst (08:37):
When do you think that kind of changed? Is it really with the latest release of Claude that you felt like there was that big leap forward or
Guest: Carolina lago (08:43):
I don't think it's about the model itself because if you notice, Tropic will launch a model and then OpenAI will launch a stronger one and then the other one will launch. Gemini launched a very good one. Everybody's talking about, it's not about the model, but the way the topic is set up as a system in the way that you can actually implement this in real life and it's not just talking to a model. It becomes much possible to do that on workflows. When you set up ai, I was working before with NAN, which was basically a workflow and I was using many types of models, sometimes different models in the same workflow. I had a tool, I think it was an open router, the name of the tool that would choose, I could choose the model that I wanted so I would pay for open water and then I would choose the model that I wanted to use.
(09:38):
So it was good. Gemini was good for images, perplexity is good for reasoning or for research and then Claude would be better for finance and things like calculations and whatever. So I would choose the model, that would be something. But the workflow was there on the N end. Tropic with Claude was the first one to bring that, to bring the possibility of having the workflow of using that as a workflow, as a system. So I think that was the big game changing. It was not about the motto itself, their motto, for me it's the best one. It's the one that hallucinations less and it does a better job. You can tell that it does a better job, but it's not only about that because the motto, it's a race, so now Opus is the best one. Tomorrow it might be one from OpenAI the day after Gemini or Perplexity or whatever. It's not about the OOM model, it's about the system. Their on profit created and you can see that by the cps. That was the first thing that they did. They created the cps, now all of them are adopting the cps.
Host: Paul Barnhurst (10:52):
Sure, they've done a good job with their ecosystem and I agree with you, the models are going to be a commodity. Everybody's going to take different approaches to how they implement them. Google, it's really right now about the consumer and their existing products similar to Microsoft, a lot of what they're doing, but they're letting others build the hard part of the models and putting their own layer on top of it, right? Claude is very much an enterprise. I think chat GPT has leaned harder on the consumer side even though they want the enterprise as well. And so there's just different ways you build and you put those tools around it. But the models themselves, yes, there are certain things all models are better at, but to a certain extent, the models are going to become a commodity. How they're all trained on a tonne of data, they're all using very similar probabilistic things on the backend. There'll be breakthroughs from time to time where one will be a lot better, but I don't think that's seeing the differentiator in this AI race, which obviously isn't the goal of the podcast, but I don't see that as the differentiator.
Guest: Carolina lago (11:57):
So what is the differentiator when you look at the whole thing is the human factor, the differentiator for the companies and that it's already happening. All companies will have access to some type of LLM and it becomes a commodity. The differentiator is how humans are inserted in the workflows on a way that they can improve the model, the model inside, not the model, but the model that is being implemented in the company and the model can improve the human factor as well. So with that exchange, with that interaction of the human and model, that becomes a better workflow. So that's going to be the differentiator. I think about the future. Competitiveness is going to come all from the human factor.
Host: Paul Barnhurst (12:52):
Interesting. It'll be interesting to watch. I get what you're saying and I think there's some technicality as well, but yes, the human factor is huge, no question. So what's been the most surprising part of your learning journey as you've leaned into AI over these last few years?
Guest: Carolina lago (13:07):
The most surprising thing is exactly what I just said. If you think that AI is not going to replace humans and just think that if AI replaces all humans, everybody's going to have the same system, the same look and the same opinion and the same model to take from. It's just like if you have the same employee working in all companies, if you don't have a human guiding ai, if you don't have a human interaction with ai, you're just going to have the same model, the same responses that everybody, all your competitors are having. So that was my biggest surprise. There are a lot of people thinking, I don't want to rely so much on AI because if I teach it how to do my job, it's going to replace me and then I'm going to be dispensable. I think this is the biggest mistake, this is the biggest mistake that people are making is thinking that you cannot imprint your knowledge in your thought into the air and you cannot take more from that by imprinting that knowledge.
Host: Paul Barnhurst (14:14):
There are a lot of people scared of it and worried about the job. I also saw something today, there's a study they recently did, which shows people who rely too much on ai, they're actually seeing impacts to their brain.
Guest: Carolina lago (14:26):
There's a way to interact. So that's the thing.
Host: Paul Barnhurst (14:30):
A hundred percent agree. And they were specifically talking about,
Guest: Carolina lago (14:34):
I have been doing some tests, I have been doing some tests and doing different skills on a way that is just technical and just telling AI what to do and AI will do it. Okay. And then I have the same skill doing the same task, but in the middle of the AM asking for human judgement . So it's stopping in the middle of the way and it's pushing me to think more about certain topics for example. And that changes the way he's going to respond after
(15:07):
The results from the second one are incredibly higher, incredibly better than the first one. That is just, okay, here's the instructions, do it. And it's the same. The technical part is exactly the same. It's just the interaction in the middle of the process, in the middle of the workflow. That's the change, that's the, that's what changes from one to the other. So I have been developing this how to create skills that will improve the employee as the employee improves the skill because this is also something that I am implementing on every skill that I'm creating and every workflow especially is on every mistake I will have a session of rewriting the skill, not only every mistake, but every now and then rewriting the skill and improving the skill so it doesn't get over not updated. You have to keep it up to date all the time. So that's the best way. As you are interacting with the skill, it gets better and you get better as well because it's pushing you to think a little bit more about everything.
Host: Paul Barnhurst (16:13):
Sure. The more you use it as a thought partner and the more it challenges you to think the better, but that's not how a lot of people are using it. And so it, it's interesting to watch. What's the coolest thing you've automated so far?
Guest: Carolina lago (16:26):
The coolest thing I've automated so far, well, I just did this week. I thought it was really cool, a really cool experiment I just did this week. I'm very bad at visuals. I am not good at all. I am good with numbers. I can't analyse the numbers. I can extract what is the message that I need to bring. But then when it comes to the storytelling, it's not even the visuals, it's the storytelling and translating that from the report to the board deck, that's where I get stuck and I don't get stuck, but it takes so much of my time because I keep thinking about what is the story that I want to tell and what is the so what now what and all of that, those frameworks. So I created a full workflow that goes from the report all the way to the deck, but then I was struggling with the visuals and then I tried something different.
(17:24):
Instead of just doing a PowerPoint, I did HTML. So at the end there comes a presentation in HTML and some people were already suggesting instead that a presentation would be clickable and so you can deliver the results in an interactive way. And I just discovered the whole world of HTML and then I have to challenge what you just said, that people get dumber. The more they use ai, I don't think so because this is something that I didn't know was possible. And you start to find out if you have an AI, that you interact the right way. You start to find possibilities and once you know the possibilities, you don't have to know everything. Coding for example, you don't have to know how to code, but you have to know what is possible with coding. Once you know what's possible, then you know how to get that from ai. So that's the big secret. So you are always learning something if you are interacting the right way.
Host: Paul Barnhurst (18:27):
I mean who's to define the right way. What I will say is I agree you can use AI to improve learning and you can learn with it. A hundred percent agree. All I'm saying is the studies are showing, and these are scientific studies that the way many people are using AI is changing their brain. That fish is scientific. I'm not saying that's how people should use ai. There's a difference and I get that, hey, depending on how you're using it, it can help you learn just like how you use the internet, how you use a computer. There's obviously different ways, but there is a real concern and I think it will, I personally believe the average person will be dumber because of AI 20 years from now. Not saying that people can't be smarter, but I think the average amount that people will delegate to AI will make us dumber as a society in the long run.
Guest: Carolina lago (19:21):
There's a huge difference between working with AI and delegating to AI.
Host: Paul Barnhurst (19:26):
A hundred percent agree. I'm not disagreeing at all. I'm just saying what I think will happen. I know how most people, the average person works in my opinion, but we shall see. So I get what you're saying. Yeah, but
Guest: Carolina lago (19:37):
That's what I'm developing this method of workflow because it keeps pushing people up. So people keep saying, oh, it's going to replace the job. It's not going to replace my job. Maybe it's going to push employees or professionals to go to a higher level.
Host: Paul Barnhurst (19:54):
It'll be a mix. It will replace new jobs. The internet replaces jobs, the typewriter replaces jobs, every technology changes and shapes jobs. Do I think we'll get rid of most jobs? No. If we go to a GI all bets are off when we hit that, but then I will worry about that when it happens. So I agree with you. For the most part it will change jobs, but it will definitely displace a lot of people. We've already started to see it. There's no question that people will lose their jobs because of AI, but that shouldn't change the fear because on the whole the economy will adjust. It's not like everybody's going to be out of work. Yeah,
Guest: Carolina lago (20:31):
Exactly. Yeah, I don't believe that either. So I think people should take advantage of it instead of just waiting because they're scared.
Host: Paul Barnhurst (20:41):
A hundred percent agree. People have to lean in. What I say is let's assume it could even do 50% of people's jobs. Let's just say it can. And they displace 50% of people who're going to buy the AI tools with money, nobody has a job. The money all goes to a few wealthy people and you're much broader discussion than for fp and a, but you're going to have some huge problems throughout civilization if that happens to deal with. We shall see. We're all guessing to a certain extent. Nobody knows how this all ends. And if you do, let's go to Vegas. So what would you say are the top FP&A use cases
Guest: Carolina lago (21:18):
Everywhere? I see now on every position everywhere, I see edge advantages. So it's really hard to say. I think it would be easier to say what is not. Maybe not even that. I see some use cases, a good use case on everything. Everything from FP&A to all of, because the way I'm seeing ai, it's not replacing, it's not just doing manual tasks, it's doing the manual tasks, but it's also working as a thinking partner. And when you do that, when you put intelligence, it's like you have the workflow that does the job and you already have the people doing that and then you put intelligence x extra intelligence on top of that. So a lot of people are thinking " let's give that to ai. So we don't need one person there, but maybe the one person they're going to do something that they didn't have time to do before. So it's going to be another level of thinking of processing the information.
Host: Paul Barnhurst (22:20):
Okay. So where do you think AI is not good? Let's start there.
Guest: Carolina lago (22:24):
Probably on different levels. I would say medicine probably does help, but used to need a lot of human judgement , more human judgement than artificial intelligence. I would probably say that.
Host: Paul Barnhurst (22:41):
Are there any fp a tasks you think it shouldn't be used for or that it would struggle with
Guest: Carolina lago (22:45):
Fp a in fp a I yet to see something that AI cannot be used at. Let's go back to this. I am not saying replacing, I'm saying using. So I have seen for now, up to now, I have seen a good use for AI in almost everything we do in fp and a. And I have to remind you that I'm, I dunno if you know any and you tell me if you know any FP&A team that is capable of doing everything they should do, I haven't seen yet. I haven't worked in a company that would say, okay, we are done. We can go home. All our tasks are done. I haven't seen that.
Host: Paul Barnhurst (23:29):
I've worked for a year where I had a company where I did 40 hours a week. But that's different than saying everything is done. But I mean you could say that of any role, there's always more to be done. That's life. There's more to be done in your personal life. I mean there's always more you can do and AI doesn't change that. No matter how much you're using ai, you can always find more opportunities.
Guest: Carolina lago (23:50):
It's going to bring you up, it's going to bring everything up. You can do more with less time. So once you can do more and better quality with less time, you can, when I say more, it's not more just in quantity but understanding more of the business, being able to analyse more of the business, being able to take action on more things of the business. Once you have that possibility, you just go higher. You just go up and the companies are going to be more and more and more efficient with the use of ai. And it's not like a rainbow, everything is wonderful. It's not like that. It's not paradise. Of course there are problems and needs to be solved and there are going to be issues anyway. But the way it's going to work, it's just going to take people to another level and it doesn't matter what level of the company they are.
(24:46):
Let's say a junior analyst that used to be only filling up cleaning data, let's say now he doesn't need to clean data anymore. He's going to get the data cleaned in 50 minutes, 30 minutes, and then now he's going to be able to put together reports. But then with ai, maybe he's going to have an agent that does a report for them. Now he's going to go up and he's going to start analysing the report. So he's going to work with AI on the first draught of the analysis. Then that analysis will go to a senior analyst and then it's already taken care. A little bit of it's not like before he was doing the cleaning and then, I dunno what medium analyst would do the reporting and then a senior analyst would do the analysis. It's not like that. It's going to be requested early on that they come in and they understand the business and they analyse as well. And that's with the use of ai that comes with the interaction of ai. It's going to be supervising AI more than just doing the manual things.
Host: Paul Barnhurst (25:57):
FP&A Guy here. AI analyst agents are one of the biggest shifts in finance tech in a decade, and most FP&A leaders are trying to figure it out. On May 21st, I'm hosting the FP&A & AI Software Showcase with two of the leading AI agents on the market, Concourse and Sapien, plus planning tools Drivetrain and Una AI. One sitting. Real demos. No sales pressure. Register at www.thefpandaguy.com/fpa-software-showcase and see you on the 21st.
(27:00):
I think there's a lot of issues that still need to be resolved. I'm not going to disagree that it can be used in a lot of different places. I think the data cleaning on an individual level, on a foundational level, AI is not there to clean your data, which is going to impact results. If you're dealing with large multidimensional data, AI breaks apart context, windows, and a bunch of other things, there's still a lot of challenges. Doesn't mean you can't use it throughout. There's energy challenges. There's a lot of things that have to be resolved before we see this idea mass adoption at the level. A lot of people think that doesn't mean you can't use it, everybody should benefit from it. I'm with you, but I think it's going to be a much slower adoption than most people think.
Guest: Carolina lago (27:44):
Yeah, but like you're saying for clean data, the contacts and everything, that's the way you structure everything. If you know how you
Host: Paul Barnhurst (27:51):
Tructure, it's more than that. I was talking to a guy who's an engineer who's ran multiple, one of the smartest people in the world dealing with very large data sets. AI is bad for very large data sets. It's not just structure. You're dealing with tens of billions and hundreds of billions of calculations. You should not be using gen ai. It does not work well, but his customers all want it, but he has to tell them every single one asked for it and he's like, it's not good. There's other things I should be doing. So what I'm saying is there are limitations that are beyond how you do things. They are legit. Exactly.
Guest: Carolina lago (28:27):
Exactly. But AI will help you create the tools to do that. So it's Python, what's going to happen. You
Host: Paul Barnhurst (28:34):
Could use it. I'm not going to disagree there. I agree. There are things you can use it to do.
Guest: Carolina lago (28:39):
Yeah, I'm not talking about a large data junior analyst, he's going to clean data for his report. He's not going to clean a large set of data, he's not going to clean
Host: Paul Barnhurst (28:48):
Agree for junior analysts. I'm not going to disagree at all.
Guest: Carolina lago (28:52):
What I'm saying is now he can use tools with the help of AI like Python, SQL and all of that to just bring the data that he needs without having the need to call it and ask for a new project and wait for, I don't know how we're
Host: Paul Barnhurst (29:08):
Not speaking from experience there. You have never had to do that, right?
Guest: Carolina lago (29:12):
No. Until I became the FP&A of it and then it was much easier because now they needed me.
Host: Paul Barnhurst (29:21):
See, I worked in a business analyst role where I wrote it. I was technically called a finance analyst and I switched into FP&A and they never took away my access. So my boss loved it. I could pull all the reports from the database for it and put it together. So it was very valuable in my career. So here's the question. It was a thought tool and I'm curious to get your take. I think this is a fascinating discussion. I love the differing opinions in the back and forth and then we'll get into some use cases. People are like, all right, stop talking about all this theoretical stuff. What I was going to say, what's your take on how well people need to learn the technical skills with ai? My view is AI as a magnifier. If you don't know, well, it's just a matter of time until it magnifies that If you know them well, it'll magnify that. Your thoughts.
Guest: Carolina lago (30:07):
Yeah, I totally agree with you, but I have a take on this the same way I'm doing HTML now, and of course I'm very basic on this, but I learned something. So what I'm saying is if you do a proper use of ai, you can magnify your errors. If you don't know it, you can get a very good process if you know that field already because you can amplify that. But we can also learn. So
Host: Paul Barnhurst (30:35):
A hundred percent
Guest: Carolina lago (30:36):
My take is you start to understand that you don't know. Understand like I'm doing HTML presentations now, websites and whatever. I'm not a web designer, I'm not a coder, but I am using AI to help me with that. But the way I am approaching this, I am getting it step by step and piece by piece. AI is doing for me, but I'm understanding where it's changing, what it's doing, what it's capable of doing, and that is the first step of learning something when you know what the capabilities of that tool is for you. So once you learn that, then you start exploring those capabilities and that's the basics of learning. That's how we learned Excel, that's how we learned everything. So I am not an expert in web designing. No, I'm not of course, but I have an expert by my side and if I'm curious enough, I can just watch what he's doing and then when I ask for changes, I can see where did he change what he changed and how it's going to be much faster for me to ask to get what I wanted to do the next time if I know where he's going to change and what he did wrong.
Host: Paul Barnhurst (31:50):
A hundred percent. The more you can use the tool to help you learn by asking questions, following along versus using it to just augment a task, the more you're going to learn it. Could I have Claude teach me Excel, teach me Power query, a hundred percent HTML, Python, whatever or any of the LLMs, right? They're all capable of putting together a plan. I've had it put together a running plan for me. Now I'm a runner so I could review it and I made changes. But even if you don't know a topic well, it can get you started and it's all about how you learn. Just like you can learn from a book, you can learn from a video, but you have an expert that can go much deeper now you have to know when because they are going to hallucinate, they're going to make mistakes. That's where human judgement and a lot of judgement comes in and that's where you have to have the judgement to know whatever you're doing, let's talk fp and a. You have to have the judgement to know when it makes sense to use the LLM to return the answer versus using it to code something that returns the answer. I think that's enough. People don't realise, well, it's no good. Well yes, but this should be done. Deterministic AI can help code the deterministic piece
Guest: Carolina lago (33:03):
And it's good on everything that is deterministic. Everything that you need to get a code or to get formulas or to get all of that. And that's when pythons and data structures like none comes in place because that's what AI understands better. Then just asking them to calculate something to give me the chart, probably give me the formula like we were talking before, give me the formula, put it there. What is the chart that you're going to use, what formula where it's going to get the information? Then it becomes deterministic, then just having it create something
Host: Paul Barnhurst (33:43):
A hundred percent. And so let's step back. We've had a great discussion here, but I want to leave people with some very practical advice. So first, I think many people are still struggling. Where to start we see on LinkedIn is a whole different world, but the reality is numbers show less than 10% of companies in finance have really productionalized what they're doing with FP&A and most are testing. So what's your advice there? Where should, what should people be doing right now to get better at ai? Give me two or three practical steps of what you'd recommend they do.
Guest: Carolina lago (34:19):
I would say diving, that would be my first thing to say, but it takes time.
Host: Paul Barnhurst (34:27):
What does that look like when you say dive in, elaborate, give an example.
Guest: Carolina lago (34:31):
Install, install and start testing. But it takes time. So that's what my second recommendation would be. Find somebody that already got the way, already found the way and that knows what they're doing, especially in your field, in your niche. And I have to say, I'm about to put a programme that is very complete. It's coming up at the end of May, but I won't say just wait for this. It's coming up in a month, in a month or so. But I won't say, just wait for this. Just get the first steps, get the subscription. I think everybody, and that's the first resistance a lot of people get is I don't want to pay for it. My company doesn't pay for it. My company has a copilot. Oh my god, I'm so sorry for the people that have a copilot because they feel stuck. They cannot experiment with other things.
(35:25):
It's getting a lot better, but it's way behind in the race. So for those people especially I'd recommend try on your own. Don't wait for your company because when your company gets everything figured out in terms of security, in terms of governance policies, all of that backend that the companies have to figure out, and that's why they stuck on Compile because it's safer because it's Microsoft. But when they get a figure out, when philanthropic co-work gets to co-pilot finally and it's for everybody, it's going to be the professionals that already got this are going to be more valuable.
Host: Paul Barnhurst (36:08):
FP&A Guy here, and I often hear finance leaders say, 'I know I need FP&A software, but I don't even know where to start.' That's why I started the FP&A Software Showcase, so you could see top tools in action. This year, we're featuring Drivetrain and Una AI on the planning side. And for the first time ever, we're adding two leading AI analyst agents to the market , Concourse and Sapien. These tools are already being deployed at public companies. You will get to see demos, without the pressure of a formal sales pitch. Ask your questions and compare tools, all from the comfort of your chair. Join me on May 21st for the showcase. Register at www.thefpandaguy.com/fpa-software-showcase.
I want to follow up on this. I think it's important. Let's just say my company uses a copilot. Are you recommending they go out and get a different tool and learn that tool instead of trying to learn copilot deep? Is that the recommendation here? I
Guest: Carolina lago (37:43):
I would recommend that you go out and learn on yourself by yourself. Don't worry about the company if your company is not willing to do that yet. Because at some point I think all of the tools are going to be following the same structure that Claude is doing now. I don't have
Host: Paul Barnhurst (38:00):
So you think everybody should learn Claude right now?
Guest: Carolina lago (38:03):
That's my opinion.
Host: Paul Barnhurst (38:05):
Okay. I would disagree, but I can understand the logic and that's why I wanted to ask because I think it's fun to see. It's just like everything, right? Modelling. When I ask people should you do circular references, shouldn't you, what I want to draw out of this for people and then we'll keep going. I love the perspective. Look, there's going to be a lot of different opinions here. You all have to jump in and decide what makes sense for you. I think we can both agree. Number one is learning, so thank you. I just wanted to make sure that was your position. That's what it sounded like.
Guest: Carolina lago (38:37):
I think Claude makes it easier for you to actually have a workflow that you can see from beginning to end. It just makes it much easier. So when you learn Claude, you can apply that to other tools because you can get a skill and put on a project in chat GPT, it's not going to work so well, but you can do that. But when you are learning on CHA GPT, you cannot transfer that. It's just limited for now and I think it's going to get there. All of them, I think all of them are going to get there. Just like the cops. The CPS started with them tropic and now it's open source and everybody's using and all of them are using just like the cps and we already see skills for Codex. We already see skills for other tools. So it's the same. It is the same methodology.
Host: Paul Barnhurst (39:30):
Yeah. What I'm hearing is your opinion that if you're going to learn, the most complete tool to learn on is Quad six months from now, a year from now, they may all be the same. So you're going to get benefits. They'll help you in the long run. Is that a fair statement? Exactly.
Guest: Carolina lago (39:45):
Yeah, exactly.
Host: Paul Barnhurst (39:46):
I come from, my angle is you can go really and learn a tonne of stuff on chat, DPT or copilot or whatever your work does. I think you'll be better served doing that and you can pick up the rest later. I don't think there's one necessarily right methodology. I haven't gone as deep as you though, but that's my initial thinking. So interesting.
Guest: Carolina lago (40:06):
Yeah, I just think we can do much more and that's why I was quiet about AI before I experienced cloud because I could check for hours on CT PT and they just told me what to do and even Codex was not that good. It was like just coding. It was not like, okay, this vibe coding, there's a lot of prejudice around this, but I think with Claude and the way it's set up, the way the ecosystem is set up, it's much easier to make something that actually you can use. It's not only talking to AI and giving prompts. I just think people sharing prompts and copying and pasting prompts and I'm like, this is such a waste of time. But that's what works for the other tools. There's no other way. There's no skews on the other tools, there's no plugs on the other tools. Maybe there is a barrier beginning
Host: Paul Barnhurst (41:02):
Of it. They've added quite a bit with the latest rollout of chat GPT. I mean you can create agents, you can do instructions. I think Claude's the most complete tool. I use it 80% of the time. So I'm challenging you mostly because I think it'll help people with the conversation and to dry out those differences because I think it's really helpful. Course, of course. I get what you're saying and it's just fun to disagree. Ask my wife, I'm just kidding. I'm going to get myself in trouble if she listens to this episode. She does listen from time to time.
Guest: Carolina lago (41:32):
That's always the thing I do.
Host: Paul Barnhurst (41:34):
We might have to put that in the trailer.
Guest: Carolina lago (41:38):
Yeah, he's also his company's copilot and he watches me doing stuff and he sometimes sees my YouTube videos and then he goes like, how do I do that? Come here, come here. Show me. I cannot show, it's not a copilot in scl. You have to do it on a personal computer. Just go on a personal computer. I'll pay for your subscription. And he's like, no, I need to learn this. I don't need my personal computer. I just watch movies on there. I work here on this computer. But I strongly advise everybody to go on your own and try yourself even for your personal life, even if you don't want to do it work stuff, just do for your personal life, you like to cook, just create an agent to go find recipes and create your menu and things like that. Organise your life, organise your computer or I don't know if you have a hobby maybe used for that like you did your running stuff tool for your personal life because this is the skill.
(42:34):
It's when you, this skill, what I'm saying, not the skill, not the AI skill, but your skill is to learn how this ecosystem works. The sooner you learn that, you start to see the patterns and start to see how it applies to real life and to work and to how your job is going to be done. Once you do that, that's a big shift. So when the companies all bring this ecosystem because it's inevitable. If they want to automate stuff and they will automate stuff, even if all of the other tools become something similar, you already know the concept. So start, it's much quicker for you to pick up the knowledge and pick up the way things work.
Host: Paul Barnhurst (43:19):
I want to get into something specific here in a minute, but having a conversation with a guy who's an AI expert, he studied at, I think Harvard, he's worked deep in AI and it was really interesting. We were talking about how much of it should be clawed versus companies that provide tools because to go really deep, to get really good takes up a substantial amount of time. I think we can both agree to that. When you start really building agents and the more and more you try to automate things. So I think there'll be an interesting balance. There'll always be those who do it yourselfers to go really deep and I agree, everybody should be using at a minimum as a thought partner and there's workflow, but there's going to be this interesting balance and I think most companies will settle on. We want someone to come in and build a lot of the workflow and different things and people will be using it as well as a thought partner. But I don't think most people will go deep into coding and co-work in their jobs. I think it's too much to ask for every FBNA professional to be able to do that on top of everything else. So I think we'll see the ecosystem pick that up more by developing the agents and building a lot of that stuff for us that doesn't say people shouldn't learn. I think there's two different sides to that coin.
Guest: Carolina lago (44:31):
For FP&A specifically, that's a discussion that I have way before AI is how much FP&A needs to know tools like SQL or Python or whatever other tool we have, databases, power bi, whatever. Instead of being users, they get a little bit of development, get a little bit of the side of developers. How much of this is actually the function of fp NA and I have my opinion is that the FP&A is they know a little bit of every tool, a little bit of something on every tool. They're more capable because they know what the tool is capable of doing. It's much more likely that an FP&A that knows a little bit of SQL. They'll be able to extract better the data from a bigger database and they'll be able to go straight to the point. If they know a little bit of Python, they will know what is the stuff that they will be able to do in Python that Excel cannot do. By the way, in a month I'm going to be in London,
Host: Paul Barnhurst (45:38):
Lemme do the commercial. I'll
Guest: Carolina lago (45:40):
Go ahead, do
Host: Paul Barnhurst (45:40):
Your commercial. You got 30 seconds Python
Guest: Carolina lago (45:42):
In Excel, in the global Excel summit. I'll be in London in six weeks teaching Python Excel for six weeks. So I don't know when these episodes are going to go live, but in a few weeks I'll be there teaching Python, Excel, which is something cool to learn too. Other than ai. AI can help you learn that too. But it's something cool to have Python inside the grids, which is something that we all love in Excel. So yeah, so back to the tool. I think everybody should learn the capabilities of each tool and that will improve your job gigantically. It's just so much improvement when you know what a tool is capable of doing. That comes to coding as well. When you design the workflow, you understand the workflow. You're not a player anymore. Now you are the orchestrator of all of this air. I think the good FP$&A, the higher level of fp and a, the ones that are able to analyse and have an opinion and have a good say on everything, they'll be the ones that are able to orchestrate AI and for that they each understand how to create workflows.
Host: Paul Barnhurst (46:56):
I would challenge that. I think for some I wouldn't go that far. I'm more in the middle on that, but interesting. But I want to keep going beyond that. So I love the back and forth here. We've talked about what people should do in the sense of, hey, everybody should get accounts, start learning things. So in your opinion, learning, obviously prompting is the start. You need to really figure out how to manage it with workflows. You think everybody should learn skills. Everybody, if I'm hearing here, should figure out how to build an agent. Is that the minimum? What would you say are the kind of the minimum things that you think everybody in FP&A should learn when it comes to ai? What's your minimum level that people need?
Guest: Carolina lago (47:38):
I think learning workflows and learning how the system works. It's not just for ai, it's something FP&A should know already because if you're a business partnering, you should know how the workflows throughout the company, you should know how the money flows throughout the company. Otherwise how you can advise your business partners. But
Host: Paul Barnhurst (48:01):
There's a limit to how many workflows you need to learn. You need to know the operations of the business and you need to understand the idea behind workflows and be able to look at something and figure out the flow. A hundred percent agree, but you're not going to know every single workflow in a business.
Guest: Carolina lago (48:14):
But if you are going to talk about one workflow, if you're going to analyse something that touches a workflow, you shouldn't understand where this comes from and where it's going.
Host: Paul Barnhurst (48:24):
Sure. If you're working on a workflow, you have to have some understanding if you're analysing the workflow, if you're just yes, agree, you need to understand the process in the business. No question.
Guest: Carolina lago (48:32):
Yeah, that's a great knowledge for FP&A and once you know processes, you can learn the workflows. It's not that technical. It's not as technical as it seems to be. It's basically a natural language.
Host: Paul Barnhurst (48:45):
So it says the person who knows Python and SQL and has a much higher technical level than the average. I
Guest: Carolina lago (48:51):
I know on the curious level, I know on the curious level, I am not a coder. I kept saying that I'm not a coder, I refuse to be a coder.
Host: Paul Barnhurst (48:59):
Python, you're a coder. Sorry. You're going to lose that battle. I just know SQL, so I call myself not a coder, but
Guest: Carolina lago (49:07):
SQR is code as well. No
Host: Paul Barnhurst (49:08):
Script. It's a scripting language. You're not going to write a programme with it if it's something you're not writing a programme with. I would not call it coding now it's close, but I would, that's where I draw the line for me
Guest: Carolina lago (49:20):
In a sense, Excel coding in a sense.
Host: Paul Barnhurst (49:23):
Sure. I mean depending on how you want to define it.
Guest: Carolina lago (49:26):
Do you know
Host: Paul Barnhurst (49:26):
VBAI almost knows VBA, but I agree VBA is coding. That's why I don't consider myself a coder
Guest: Carolina lago (49:33):
Lambda as well.
Host: Paul Barnhurst (49:34):
I all do. I think I, outside of a couple testing, I've written one Lambda. I don't write Lambdas. I don't think someone who's doing a lot of VBA extensive lambdas, I will call them a coder. I would agree with that. But I think someone who's doing SQL DAX Power Query, I don't consider 'em a coder. I say a coder needs to be able to write a programme.
Guest: Carolina lago (49:57):
It's true. I don't use Python to write programmes. I use Python to write charts and to use for analysis and to do forecasts. So I'm not a code,
Host: Paul Barnhurst (50:06):
But you could write a programme anyway. We better stop. People are like, where are these guys going?
Guest: Carolina lago (50:10):
I can with Claude, I can do anything. It's my superpower now.
Host: Paul Barnhurst (50:15):
I would challenge anything. I could find some things you couldn't do with Claude, but we'll leave that alone. So how do you see the FP&A role changing over the next few years and beyond just learning Claude or LLMs, what are the other things people should be doing to prepare themselves for the future? What would you tell someone starting their career? What's the advice you're going to give 'em just in general?
Guest: Carolina lago (50:39):
I think the FP&A profession are going to be able to do so much more and on a higher level. So it's finally going to be able to do what it was supposed to do in the beginning because we never have time. So I think it's going to be another level of understanding of the business. We're going to be able to do that and it's going to be requested that we do that because there's no excuses anymore that we don't have time. I mean, there's two ways, but in the future I think it's very near. There won't be excuses anymore. Anything is possible. So it's just a matter of orchestrating all of the resources and AI is one of the resources. People are another one. So it's just a matter of orchestrating everything together and knowing what to take the best of each one, the best of ai, the best of the intelligence of AI, and the best of the judgement from people and trying to combine them together to elevate the profession, elevate the teams to another level and getting a better understanding of the business. And now it's going to be possible. Now we're not going to be burning out doing the close and pulling numbers. AI can help with all of that.
Host: Paul Barnhurst (51:55):
Got it. Alright, so we're going to wrap up here in the next couple of minutes. I'm going to move into our FP&A section. I'm going to limit it. I think I'm going to ask two questions here. We've talked a lot on the technical. So what's the number one soft scale fp a professionals should master
Guest: Carolina lago (52:09):
Communication? I think it was back then even more because now we still have to communicate with AI as well and your communication skills are going to be requested more than ever.
Host: Paul Barnhurst (52:24):
Alright. If Excel removed one feature tomorrow, which one would cause you the most panic?
Guest: Carolina lago (52:30):
I think if Excel removes the grid, that's going to be a huge problem.
Host: Paul Barnhurst (52:35):
Everybody's screwed If Excel removes the grid,
Guest: Carolina lago (52:38):
That's the big advantage of Excel. There's no other, I would be using Python for everything if it wasn't for the grid, and that's why I love Python and Excel so much.
Host: Paul Barnhurst (52:49):
Yeah, the grid form factor, whether it's Excel, whether it's Google Sheets, whether it's one of 20 other spreadsheet products, is incredibly valuable to be able to see that and work with it. It's to order a spreadsheet. So if it goes away, I would argue you don't have a spreadsheet, so I'm not sure if that's a feature. That might be the core functionality, but I'll give it to you. I like it. I haven't had that one before. All right, we're going to move into our get to know you section. I'm going to ask two questions. We already covered whether Claude's a male or female, so we don't have to do that one. If you could have any one superpower in the world, what superpower are you going to have?
Guest: Carolina lago (53:26):
I think managing time, extending, going back,
Host: Paul Barnhurst (53:32):
Do you want to be able to time travel or just do whatever you want with Tanya?
Guest: Carolina lago (53:36):
I'll put some rules on that, but I could do whatever I want with my time. It will be awesome. You're
Host: Paul Barnhurst (53:40):
Going to be really rich if you can do whatever you want. Go back and make a few, place a few bets.
Guest: Carolina lago (53:45):
That's why I'm saying I would put some rules to it so I don't abuse it. But yeah, I think I would love to
Host: Paul Barnhurst (53:51):
Somehow the abuse would be fun. No, I get it. I like that answer. Alright, last one. If you could have any job in the world for one week, other than mine, of course, what job would you have and why?
Guest: Carolina lago (54:06):
Tasting wine. Can you imagine being paid to be tasting wine?
Host: Paul Barnhurst (54:11):
I'm not a wine drinker, so I can't.
Guest: Carolina lago (54:14):
Okay. My husband says he would be a judge because he didn't know when he was going to college. There was a profession called judge and he would be paid to judge people because now he does that for free.
Host: Paul Barnhurst (54:30):
Says he's judging everybody anyway, he might as well get paid to do it.
Guest: Carolina lago (54:33):
But I would love to test wine, like tasting wine and I don't understand anything about wine, but I lived in France, so that's enough. And I love wine. So
Host: Paul Barnhurst (54:43):
10: 30 for you right now on a Friday. And you're talking to me instead of tasting wine. So when we hang up here, I'm not a drinker, but go have a glass for me
Guest: Carolina lago (54:52):
Before we hang up. I have a joke and I have to tell you because I've been saving this one for you.
Host: Paul Barnhurst (54:58):
Alright, let's hear it.
Guest: Carolina lago (54:59):
Okay. What is the difference between an accountant, an FP&A and AI
Host: Paul Barnhurst (55:07):
Accountant, FP&A and an ai? Well, I know accountant and fp and a, so this has to be different. I don't know. You're going to have to tell me this one. I haven't heard this one.
Guest: Carolina lago (55:15):
Okay. An accountant gets creative, it goes to jail. You know that,
Host: Paul Barnhurst (55:20):
Right? All right. You're using my same one. All right.
Guest: Carolina lago (55:24):
Fin A gets creative, it gets promoted, and then AI gets creative. It goes to the psych ward. For how? Isation. It's a good one.
Host: Paul Barnhurst (55:37):
I'll have to use that one. That one's bad. I love it though, because all my humor's bad.
Guest: Carolina lago (55:42):
It was bad, but it was not as bad as yours, and it came from Claude. So that's what's amazing. It came from Claude himself.
Host: Paul Barnhurst (55:52):
I'm not surprised that it came from Claude.
Guest: Carolina lago (55:55):
Can you see now that I have a superpower, I can do whatever I want now. Does that make sense?
Host: Paul Barnhurst (56:01):
Let's not go too far.
Guest: Carolina lago (56:03):
I can tell jokes. Paul, come on. I can tell jokes and I can tell jokes. How can I say tailor?
Host: Paul Barnhurst (56:10):
I'll say the audience say jokes from AI are pretty bad when it creates 'em. That one was better than a lot. I've seen it create.
Guest: Carolina lago (56:16):
It's, I think it's very clever. I think it's very clever because it's tailored to the audience.
Host: Paul Barnhurst (56:23):
Well done. Alright, before we wrap up, I do have to wrap up here. If people want to learn more about you, what's the best way?
Guest: Carolina lago (56:31):
LinkedIn, YouTube, and my website. If you want to reach out to me, collaborate somehow. I have a course on AI coming up. I still have my financial modelling course, which I still think it's the best technical thing to know about a company is financial modelling. So I didn't give up because of ai. But anyway, so my course on ai, it's coming off. It's very complete. It's going to be comprehensive. You're going to learn from zero to Hero and you're going to know everything about ai, especially on cloud, and you're going to be able to apply to other ones. So just look up to me and try to find in a few weeks it's coming up.
Host: Paul Barnhurst (57:14):
Thank you so much for joining me. I enjoyed the conversation, love the back and forth. Appreciate being able to have some different opinions and the audience can tell we're very comfortable because we've chatted enough that we get talking. They're probably like, where are these two going? That's probably what they were thinking. But thank you so much for joining me. I love the work you're doing in AI and keep it up.
Guest: Carolina lago (57:35):
Thank you so much for having me, Paul.
Host: Paul Barnhurst (57:38):
That's it for today's episode of FP&A Unlocked. If you enjoy FP&A unlocked, please take a moment to leave a five-star rating and review. It's the best way to support the FP&A guy and help more FP&A professionals discover the show. Remember, you can earn CPE credit for this episode by visiting earmarkcpe.com. Downloading the app and completing the quiz. If you need continuing education credits for the FPAC certification, complete the quiz and reach out to me directly. Thanks for listening. I'm Paul Barnhurst, the FP&A guy, and I'll see you next time.