Ashok Manthena has supported FP&A teams at some of the biggest and most well-known companies on the planet including Google, GAP, and Ingersoll Rand. This included a period at Google which was striking because of “the amount of resources available for finance” particularly when it comes to killing manual finance processes. He says there is an in-built DNA to “automate manual processes”. He says: “This bubbles up naturally when they find there is an, there is a manual process and everyone comes together and thinks about how to automate that process.” In a second career stage Ashok has been a leader in AI finance carrying out practical research, giving keynote speeches and providing practical advice on transforming businesses through the use of AI.
- Rapid fire questions: Why I want to meet Gandhi – to doing Doing sensitivity analysis in ChatGPT
- Why we should all be embarrassed by the stat that 70% of FP&A is spent on getting the data (vs only 30% on insights)
- The power of meetups meetups and the return of face to face meetings for finance learning
- What my career FP&A career at Google taught me
- The near future of automation and AI including running daily reports to the CFO
- Why data priority needs to be the #1 priority alongiside faster surfacing of data issues to leadership
- How I became a leading AI finance speaker and writer and started a tech startup
- Two ways AI is going to revolutionize FP&A Departments forever
- How ChatGPT will be a new interface for finance – practical examples
- The playbook for smaller finance teams and businesses to thrive in the AI age
- Advice to get started with AI in FP&A
- The role of data science in the future of FP&A
FP&A Today is brought to you by Datarails,the AI-powered financial planning and analysis platform. Keep your own Excel financial models and spreadsheets and benefit from AI for data consolidation, reporting and planning.
Follow Paul Barnhurst on LinkedIn
Paul Barnhurst:
Hello everyone. Welcome to FP&A today. I’m your host, Paul Barnhurst, aka the FP&A Guy. FP&A Today is brought to you by Data Rails, the financial planning and analysis platform for Excel users. Every week we welcome a leader from the world of financial planning and analysis. Today we are delighted to be joined by Ashok Manthena. Ashuk, welcome to the show.
Ashok Manthena:
Thanks Paul. Thanks for having me here.
Paul Barnhurst:
Really excited to have you. So, lemme just give a little bit about his background. He comes to us from the San F from San Francisco, the Bay Area in California. He’s currently the ceo and founder of Chat Fin, a conversational AI tool for finance, which we’ll talk more about later. He previously worked for companies such as Google and Gap. He earned his M B A from the University of Akron. And so that’s a little bit about his background. What we’re gonna start with is we have a section where we like to get to know our guests a little bit, and we’re gonna try something different today and have it at the beginning. So these questions, you get 30 seconds to answer each of them. All right. They’re kind of a rapid fire type thing. There’ll be four questions here we’ll ask you, and then we’ll give you an opportunity to tell us a little bit more about yourself and dig into the interview. So, the first question here is, what is something interesting about you? Not many people know.
Ashok Manthena:
I just got into woodwork. You can see my art in the background. But I’m really bad at crafting and that’s what all the people have been telling me. I can’t really cut the word in a straight line yet, but at some day I’ll perfect it and become better at it.
Paul Barnhurst:
I think it looks pretty good back there. You’re having fun.
Ashok Manthena:
Yes.
Paul Barnhurst:
That’s the main thing. Woodworking, you know, my advice is measure twice, cut once. Ah, okay. That’s about as far as I go on woodworking.
Ashok Manthena:
That’s
Paul Barnhurst:
Right, I agree. Versus, you know, measure once and cut twice. It’s not
Ashok Manthena:
Fun. That’s right. That’s right. <Laugh>.
Paul Barnhurst:
So, next question. If you could meet one person in the world, alive or dead, who would you meet and why?
Ashok Manthena:
I think the one person I want to meet is Mahatma Gandhi. I, I read a lot about him, his thought process and thinking about when you have, when you can achieve something in short time using violence, this person opted for totally out of box thinking and said, we can achieve the same results using peaceful protests. And I, I just want to understand the thought process or the state of mind to think something like that. So I would love to meet him in person and just understand his thought process.
Paul Barnhurst:
That would be an amazing person to meet. I would love to sit and have a chat with him and learn how he came up with his approach, how he thought about it, why he felt that was the best way. It would be a definitely be a fascinating conversation without a doubt.
Ashok Manthena:
We should do a podcast with him then.
Paul Barnhurst:
Alright, well if you can get him on the show, I’ll go ahead and interview him. Alright. So this is kind of a fun one, a new one we’ve added, and I’m gonna change it a little bit, but what is the last thing you Googled looked up on, you know, YouTube or ChatGPT while I’ve added in there about how to do, how to do finance fp and a Excel, something on one of those subjects?
Ashok Manthena:
I think it’s more chat ChatGPT these days than Google, right? <Laugh>
Paul Barnhurst:
<Laugh>. I think it’s,
Ashok Manthena:
Yeah. One thing I’ve seen is sensitivity analysis, how to do sensitivity analysis in Excel because I’ve been using machine learning to do sensitivity analysis, but I also want to compare it with the manual ways of doing it. I’ve done this before, but I was like, okay, lemme check if there are any other ways to do analysis in Excel. So that’s one thing. I looked it up, did some research, and then made a comparative video of how you can do that using manual ways of Excel and how you can use machine learning for it.
Paul Barnhurst:
And so I’m curious, when you say Excel sensitivity now, so you’re talking like a Monte Carlo type simulation, or are you just talking, you know, being able to look at various scenarios or what kind of sensitivity?
Ashok Manthena:
So let, let’s say if we have one target variable and four drivers.When we are starting with it, how do we even think about it and how do we collect the data and just the basic methods to do it in Excel. Right? Then we’ll go slowly with Monte Carlo, then we can use machine learning. So I’m looking for that basic method of doing sensitivity analysis.
Paul Barnhurst:
Makes sense. Got it. Okay. Speaking of Excel, what’s your favorite formula feature thing about Excel?
Ashok Manthena:
I think the one thing is about the ability for us to run code within Excel, right? We have run VB scripts and, and all macros, right? Now we can build plugins in Excel, so we can bring in AI into it. We can bring the chat GPT kind of stuff into it. So that’s what I’m very excited about. The Excel feature. Excel is not going anywhere.
Paul Barnhurst:
I, I totally agree with you. I saw someone comment on LinkedIn that AI would kill Excel. I’m like, if anything, it’s gonna enhance Excel and spreadsheets, right? Not kill them. Right? you and I are on on the same page,
Ashok Manthena:
Right? It could be me saying it, but I won’t say Excel will be killed. But most of the manual process will be moved out. Excel, right? Excel as an interface will still exist.
Paul Barnhurst:
Sure. A lot of things will be automated. That’s, and we’ll use it, we’ll use it differently, but I think the form function of the spreadsheet Yes. Is just, it is so easy to use. So perfect for so many situations is that is never going away. Whether Excel is always the dominant spreadsheet or whatever the tool may be called, the spreadsheet is just too valuable to not be used.
Ashok Manthena:
I I mean, you brought up a very good point. We are so used to reading text, right? Or a single data point, but Excel or the tabular way of looking at data mm-hmm. <Affirmative>, I think human brains evolved into looking at tabular data now. So think about having these hundred data points and still our brain can see those hundred data points and make sense outta it. I think, I think human brains evolved in Excel, helped human brains evolve.
Paul Barnhurst:
That’s interesting. I never thought of it that way, but I agree. And, you know, Excel is the definitely the dominant spreadsheet for the foreseeable future. It’s amazing what they’ve done. But why don’t we kind of move on here and just give us a little bit about yourself and your background. You know, kind of where’d you start and how did you end up where you’re at today?
Ashok Manthena:
Sure. I, I have been working supporting finance teams, controller teams, tax teams for more than a decade now. I worked at companies like ingersol, ran Google Gap, basically supporting all these teams. But my, my most interesting parts is solving the problems for, for the finance teams, whether it’s a tax problem, whether it’s FP&A problem, whether why a lot of teams spend time on reconciling data. I still feel like I think we should be able to solve this problem. Do we really have to spend time reconciling data day in and day out? So that’s what drives me every day. How can I automate all this work? How can I make it easy for finance teams and increase the efficiency in finance teams?
Paul Barnhurst:
You know, and there’s definitely a lot of opportunity if you look at FP&A, you know, studies show I’ve seen anywhere from 40 to 75% of our time is spent on non-value add activities. Exactly. And the biggest of those non-value add activities is dealing with data. Sure. Whether it be data cleaning, data prep, whatever it it might be. And those are the type of things we obviously wanna automate so we can focus on supporting the business. So I’m glad we have people out there working on that. ’cause I appreciate that for sure.
Ashok Manthena:
I I have done this survey as well, Paul, and what I’ve seen is, in FP&A, 70% of the time is spent on business execution. Right? Only 30% on insights. And when I, when I say execution, it’s more of getting the data, cleaning up the data, talking to people. See this mm-hmm. <Affirmative>, this takes 70% of time and 30%, and less than 30% is on insights. I think it should, it should be reversed. We should spend more time on the insights part than FP&A is in a driving seat, giving the rest of all the business teams insights. What, this is what we can do, this is what we shouldn’t do. But that will only happen when we automate this 70%
Paul Barnhurst:
Agree. I, I’m with you. We definitely all should be spending more time on the partnering, the insights, the strategic, the driving value versus the, well, why is this file delta and this file over here is Delta Airlines and how do I get it all matched up?
Ashok Manthena:
That’s right. <Laugh>.
Paul Barnhurst:
Right. We’ve all, we’ve all dealt with something along those lines, so I That’s
Ashok Manthena:
Right.
Paul Barnhurst:
For sure. So you and I had the opportunity to meet this summer at Operators Guild at the summit. Maybe can you tell her a little, little bit to our audience about OG and what you learned from the summit and why you’re involved in Operators Guilds? Maybe just give us a little bit about that.
Ashok Manthena:
Right. I really like working from home, but I don’t think there’s nothing that beats the face-to-face interactions. OG Summit gave that chance, of course, COVID for two couple of years. We, we didn’t move from the house. We, we all stayed home. We did the Zoom call. Zoom calls are good. I’ve, I’ve been following FP&A Guy and your Podcast and your videos for a while. I feel like I’m connected to you, right? But only when I met you at OG Summit, you know, we just hit it off right away. We had a conversationthe empathy flows and we’re like, okay, whatcha doing? How can I help? And that kind of things that only happen at summits like this, it’s just because people are much more relaxed and we’re much more willing to hear what other people is saying, what their problems are. So that’s what I liked about OG Summit. And it’s also, and the one thing, peculiar thing that I’ve seen is everyone there is celebrating other people’s achievements. And I, I really liked it, right? I mean, it’s not like there’s like, okay, you did this. It’s amazing. What can I help? How can I help for you to go to the next step? And I think that’s a, that’s a very rare thing that happens. So I’m looking forward for the next summit as well.
Paul Barnhurst:
Yeah, no, it was a lot of fun and it was great meeting you and others. And there is a difference when you get to meet somebody in person versus talking to somebody on the phone. When you can hang out with them and have those conversations informally and see their face and their body language, there’s definitely benefit to that. It’s great to be able to be out. We all went through those couple years of covid and there were some real challenges feeling locked up for a long period of time.
Paul Barnhurst:
[Datarails Ad]
Paul Barnhurst:
Now, obviously you’ve worked with the finance tech stack quite a bit, the FP&A tech stack across the, across the whole spectrum there. So how do you see that evolving in particular with kind of AI coming and everything? Where do you see it going in the next maybe 18 months, kind of three years to five years? What’s kind of that outlook?
Ashok Manthena:
Yeah. Wow. That’s a, that’s a really big question. I think when we start with finance, right? I’m, I’m sure you, you remember all those Hyperion and pre Hyperion days, right? Where we used a lot of stuff on Excel. So particularly with finance, we started with accounting software. Each company used to build their own accounting software. And then slowly we said, oh, there’s packaged software. Now, these CDs that used to come in, we used to load it into our laptop or computers, then, install it, then start using it. Then slowly it became big. We started installing on servers and then everyone started using it. But then we realized, okay, there are problems with servers. There’s a lot of maintenance. So there is a cloud software that came in. I remember this conversation during 2014,15 big corporations, right? They said, we’ll never go onto the cloud, particularly for finance software.
Ashok Manthena:
And they, they have a very strong reason. They said, this is finance software. We are just huge company. We don’t want our finance data to go outside. It’s like, okay, makes sense. Couple of years, they’re all on cloud, all the software is installed on cloud. And then they said, we are not going to use any of the SaaS softwares because we want, we want this software to be installed on our, our cloud instance. So it it, it’s with us, right? Two years down the line. Now, you know, everything is on SaaS softwares. It’s, it’s also good. So SaaS is gonna stay when I say SaaS, right? For, for our audience, it’s software as a service where you don’t have to install anything. You just get a U R L and your credentials. You log in and you start using it. You don’t have to install it.
Ashok Manthena:
The biggest advantage that happened with with SaaS software is there is no maintenance. You don’t need a lot of database people or your SREs to maintain the software. It’s all maintained by the vendors, right? So all you have to do is use it. Of course, there will be integration. You still need some engineering team that manages integration. But we have taken all that work away from IT. And now finance can do a lot of stuff. Finance can do the configuration of the software itself. Some, some finance folks are really good with it. And they start using it. So this is, this is like the on-prem to SaaS evolution that happened and it’s gonna stay mm-hmm. <Affirmative>, right? That’s one part. Now the ai, the big thing that is happening with it predictive analytics is a huge thing. And it is coming after finance, right?
Ashok Manthena:
It’s finance has been a little late in, in adopting AI for predictive analytics for various reasons. The, and then there is generative ai, the chat g PT part. Right now, when we are thinking about how can we use even this chat G P T for finance, a lot of people have been started using like Google, right? Gimme some formulas. But that’s a very basic use case once you start going into the details where you can use finance, ah, and it’s gonna be amazing what you can do. I just heard one of this use case from one of my colleagues is when you’re a CFO or a VP of finance, you get a lot of questions from your team. Your team. I mean, if you, if you let them, your team always wants to pick your brain on something, right? Can you document that?
Ashok Manthena:
Can you document your thoughts as a CO or a VP of finance and give it to a a, a chat GPT or a large language model so that it answers questions for all your employees. Think about that. Think about how it changes, right? Of course, let’s say if it can’t answer, you can always find the, the, the real CFO. But if, if you actually create that content and train the L l m, it’s gonna be a huge advantage for your team. What, what C F O wants to think about a specific aspect of it. So that’s a very unique use case. I haven’t heard about it before, but now I’m like oh, this, this makes sense. But as we discuss more about ai, let’s talk about more use cases about predictive analytics and also generative AI.
Paul Barnhurst:
Yeah. We’ll get into that some more. And I appreciate kind of the history and you know, definitely we saw the cloud, now we’re starting to see how AI comes into it. You know, we’ve seen the last few years, native integrations, APIs become so much easier, right? Yeah. It’s nowhere near as difficult as it used to be to have systems talk to each other. There’s a lot more web hooks and ways now we’re seeing much better integration. And now we’re gonna see that next step of how does machine learning, generative ai, other things move us forward to be more productive. Yeah. I heard someone say that, and we’ll talk more about this later, but they called AI instead of the fourth industrial revolution, the first productivity revolution
Ashok Manthena:
For sure.
Paul Barnhurst:
And I thought that was an interesting way to look at it. So, you know, going back to a little bit of your work experience, I know you spent roughly five years at Google Yeah. And you did mostly financial reporting and financial systems. So can you talk a little bit about that, that experience, kind of what you did and what you learned from your time there?
Ashok Manthena:
One thing that is striking at Google is the amount of resources that are available for finance, right? I haven’t seen that in any other company. Maybe it’s, it’s companies of that scale have that that resources available, right? What it means is that every module or every team are have specific people that takes care of that process. For example, if you take, just as an example, fixed assets, there is a set of team that takes care of fixed assets and, and it’s, it’s a really big team, and there is mm-hmm. <Affirmative> teams for every module. So people do a lot of specialized activity for that processes. But in other companies, I’ve seen, there’s one person managing everything. And, and that’s, I think it’s a, there’s an advantage and disadvantage as well of doing that. It’s good for an employee to actually learn a lot of things.
Ashok Manthena:
But also now we are, you know, it’s, it’s a lot of work for you to, to manage everything, and you won’t be specialized in anything special. The second thing is the, the mindset. How, how the automation is done in the company. For example, in finance teams, this is not driven by of course some of them are driven by leadership, but this is not, this is, this primarily comes from the DNA of the teams itself is is the willingness to automate manual processes. So they bubble up naturally when they find there is an, there is a manual process that is going on, and now everyone comes together and think about how to automate that processes. And they’re okay to invest money to automate that process. And I’ve seen amazing automation that was done there. And probably still, still doing it, but I don’t, I haven’t seen that kind of the, the ability to automate it within the teams and also surface it so that everyone knows this has to be done.
Paul Barnhurst:
Interesting. So you’re saying just kind of from a resource standpoint of any of the places you’ve been, they did the best job of really helping surface and automate the various tasks within
Ashok Manthena:
Finance? For sure. That’s right. Yeah.
Paul Barnhurst:
That’s where I should have been working instead of some of the companies I was in. Oh, wait, did I say that out loud? Skip. Definitely have done my share of the manual stuff, but that, that’s great. That’s what you want. That’s
Ashok Manthena:
So
Paul Barnhurst:
How, you know, if you were involved in that kind of environment, how did you work with FP&A what advice would you offer to FP&A of working with someone who’s kind of liaison between IT and finance? Like how can they work better together?
Ashok Manthena:
I’ve, I’ve done that before, right? Like being a liaison between finance and IT making tech people understand the finance language and the other way as well. Mm-Hmm. <affirmative>. But I still think, do we really need that job? Do we really need a person to make two teams talk to each other? Right? But again, if we see the evolution of the finance technology, we are going more towards the user experience of finance users. So now more and more thanks. I mean, you, you Paul, you have seen the new FP&A software. Mm-Hmm. <affirmative>, they look really nice, right? When I said the, they look nice. It’s not about the colors, it’s also about the user experience. If you have to find something, you can go easily find a lot of things there. So that kind of user experience gives ability to finance users to perform all their, most of their duties, right?
Ashok Manthena:
They don’t need someone, an IT expert or a functional expert, technical functional expert to come and help them. So that, that is helping. So I think this job of creating, you know, that, that, the conversation, enabling that conversation between IT and finance will probably shrink more and more with I, with AI coming into picture. And finance should be able to talk to tech directly. And also, as I said, the tech impact and the tech work has been reducing over the years. And hopefully it’ll reduce more and more in the next few years. So finance people can drive a lot of this stuff. Like, for example, if you want to build an automation within finance, how do we do it right now? We, we have to have an RPA tool and ask someone to do it for us. You need an RPA expert to come and help us. Can we do it with AI? Can we just say a simple chat message and say, for example I have to run this report every day. Why do I have to do it manually? Can I run, can you run this report every day at 8:00 AM and send it to CFO? Automation should be that easy. And it should be a power for finance users and think about what happens. Then most of our process will be automated just because they have the ability to do it.
Paul Barnhurst:
Yeah. Great point there. And definitely as we automate more of that, it does reduce that need for the liaison. And as tools are more no code, low code self-service, you know, I’ve filled a little bit of that role quite a bit in my career. I, you know, I did an MBA in finance, but I also did a master of science and information management. And I was a re report writer for a year, done some BI and so often I kind of filled that bridge in different roles between, you know, the data side, a little bit of the IT and the finance. And so I can relate to that. We’re trying to make sure both sides understand. And I do agree with you. I think you’ll see a reduction in the need for that role as we continue to, you know, bring easier and easier tools forward to use like generative AI and other things. You know, one other question around data. I’m sure you’ve spent a lot of time around data integrity and worked that a lot with the systems. So when it comes to data integrity and maintaining that, any advice you’d offer to FP&A, you know, any best practices that we should be thinking about as we’re, you know, working with data, working with systems, you know, kind of building things out?
Ashok Manthena:
So data integrity should be your highest priority. There’s no other way. It’s, you have to fix it. If you find the problem, you have to fix it. Don’t go with workarounds or your manual processes to it. Do it. Stop everything. Get your investment and focus on building that data integrity. It’s just because it’s not even a long-term investment for you. It’s, it’s even, you get the written short. And I’ve seen this in one of the company where there’s a lot of planning happening. The planning data is scattered across various systems. And when they bring into one place, data doesn’t match. Right? And they lived it, its five, six years without even matching it. You know how they build another workaround mm-hmm. <Affirmative> to match the data and, and they lived with it. But instead, the company should have done this long ago to, to normalize the data to make it right. So one thing is data. Take data integrity very seriously. If you find it, solve it, and also automate the processes, manual loads into any of these systems is not a good practice there. There’s lot of room for error that happens when you start doing this manual loads. So automation is the best part where you can ensure everything isn’t right, and also it’s consistent over months.
Paul Barnhurst:
I love the part where you said basically, if you see a problem, don’t just build a workaround, but figure out how you can solve it. I mean, ideally with data integrity, whenever possible, you should always strive to solve it at the source. That’s right. Anything other than that is a bandaid, you know, as you were saying. And so that’s, it’s great advice. But so often we don’t worry about the source and say, well, here’s the big workaround. That’s true. And sometimes that’s because, well, we tried to get it fixed and they didn’t give us investment. So sometimes you’re stuck. But you should at least have that conversation and always try to go back to the source. My you know, interestingly talk of data integrity, it’s kind of a passion of mine. My internship out of grad school, I actually worked in a data integrity team for American Express.
Paul Barnhurst:
So I spent all summer looking around credit card codes and trying to find how accurate the coding was to what we had expected to be and what were concerns around it. ’cause They just rolled out a new new card number system. They had just updated all the card system and gone to a whole new database. All the processors had to be updated. And that was a really interesting experience to just see how important that data integrity is. Yeah. Realizing that, oh, if that card’s not coded right, somebody’s not getting their rewards because it’s not coded as a proper gas station or a grocery store, or whatever it might be. ’cause Right. There’s all kinds of different rates with credit cards. Right. So it was a really good learning experience early in my career to just see how huge the impact can be of simple data mistakes that sometimes people think, oh, it’s not a big deal, I’ll just fix it here. And don’t realize how big the impact can really be.
Ashok Manthena:
Right. And Paul, let me ask you this. When, when, as a finance user, when you go through this process, right, like let’s say forecasting process or a planning process, and you see these data integrity issues, what needs to be done to surface it to leadership, right? Of course. You can’t really stop all your forecasting process and say, I can’t do this until I fix this issue. You have, you still have to continue. And that’s what Yeah, of course all user does. Right? What do you think is the best way that it can be surfaced to the leadership?
Paul Barnhurst:
Well, I mean, obviously I think there’s a few different ways, but I think the first thing is, whenever you come across it, make sure you keep track, keep a running record of it. Second, have that conversation with your boss and ask him. ’cause He’s often gonna know, or she, you know, your boss is often gonna know what’s on the horizon, right? Like, maybe they’re bringing in a new ERP or they’re bringing in a new billing platform, right? And so maybe they’ll say, Hey, we need to do a bandaid till this is here. But I think it starts with having the conversation with your boss. It also can, there’s ways you can bring it into your BI with your business partners and not in a way of, Hey, here’s the problem. But as you go through and you explain data saying, look, here’s all the assumptions we had to make because of data quality issues, right?
Paul Barnhurst:
And so we have less uncertainty that this is the right path forward. You know, we have some recommendations of how we think would improve that. And we’d love to have those conversations with you about maybe how we fund it if you’ve already talked to your manager. So I think there’s a few different approaches depending on what business you support and where you’re at in the business, right? If you’re working in corporate, you may have more access, especially with a large company to, you know, C F O or high leaders where they’re gonna easily see it, you know, in other roles you might be more distant and you just kind of have to run it up through the chain. So I don’t know if there’s one right way, but I think it, one, it starts with recording it. Two, you ought to be able to show the impact, right?
Paul Barnhurst:
Because if the data’s wrong, but it doesn’t impact anyone that doesn’t really need to be fixed if it’s very low impact. So you also need to have some kind of prioritization. And then, you know, three, whenever you’re bringing it up, when able come with a solution, don’t just tell the boss, our data’s a mess. Here’s the problem. Fix it for me. That’s right. Because they’re gonna be like, all right, well, you just gave me one more problem. Why do I pay you to give me problems? That’s right. And so I think those are things you need to think about is tracking it, kind of ranking importance of it, trying to think of those solutions and ha then having those conversations, don’t just bandaid it.
Ashok Manthena:
That’s right. I think there should be problem forums in companies, right? In finance teams, just talking about problems. You said you should come up with solutions. Sometimes it’s, it’s totally out of your, out of your scope to even think about the solution, right? Sure. Probably it needs, it needs a data warehouse. And as a finance user, I might not be even able to think, sure, oh, I need a data warehouse, but I know the problems that what, what is happening with me? So just discussing about problems. And I think a lot of IT, and people supporting it, like product managers will be really happy to hear problems from business users because that will flow into their roadmaps as well. So just talk to your business partners, as you mentioned. Talk to IT and tell them, even if you don’t know the solution, tell them what, what’s, what’s what’s the problem and how it is impacting you.
Paul Barnhurst:
No, that’s a good, a good point. There are definitely times when you don’t have a solution. You just need to say, Hey, here’s the issues I’m seeing with the data. Can you help me with the solution? What I mean is just be careful to not come across as always complaining where possible, bring a solution or bring a problem with a, here’s the issues, here’s what I’m doing. Help me find a long-term solution.
Ashok Manthena:
That’s right. So,
Paul Barnhurst:
Yeah, I, I agree with you. There are definitely times we may not, you know, someone may not even have an idea where to start. All right. Well, now that we’ve covered the exciting stuff, I’m sure we board the audience about data integrity. We’ll move on to what everybody wants to know about AI these days. Right? So for the last three years, you have been working as a speaker and author about AI for finance. So can you tell us a little bit about that experience and how it came about?
Ashok Manthena:
Sure. so this is almost like four or five years ago, I started thinking about how machine learning models can be, can be used in finance, right? It’s just a thought because I’ve seen marketing using it, supply chain, using it, but not finance. And it’s like, okay, in in finance we have a lot of data and can we use this in anyways? We are using the data, for example, in FP&A, we use our historical data, we use our future assumptions of, of our span or, of the markets. And then we may, we forecast, we predict can we use much more data-driven way to forecast this? And that’s how I got started. I started talking to various finance teams just to get an idea how, how how things can get started. And this is, there’s a lot of trial and error method because at that time, there’s not many companies that are working in applying machine learning models for finance.
Ashok Manthena:
That’s how my learnings, I started publishing a lot of articles and speaking at conferences and, and you know, the advantage of publishing articles and conferences is people come to you and also they tell you about their problems. And that’s how you are, you’re like, oh, so this is a problem that exists, right? How can we solve this problem now using the technology that is available? So that’s how I got started, and it’s going great. So far. It’s a very, a very fulfilling task of, it’s not just about me learning, but also sharing it to all the people. And again, all the audience, they provide information back. So it’s, it’s a cycle.
Paul Barnhurst:
Totally agree with you. It’s a cycle. You know, I’ve spoke at conferences and written articles and people coming, Hey, have you thought about this? Or What are you doing about that? You know, one of my favorite sessions at a conference last year was around machine learning. So kind of what we’re talking about here and using it in finance. So it was at the Association for Finance Professionals in Philadelphia last year. Microsoft came in and talked about their algorithms. They, they developed fin, which is open source, and how it had changed the way they were forecasting and had increased accuracy. And it was just fascinating to hear their journey. ’cause They brought in one of the guys who I think was one of their very first, if not their first finance data science hires about seven years ago. They helped implement the whole process.
Paul Barnhurst:
And so they had him and another, you know, leader from the finance department and talked about how they got to the point where, you know, they did analysis. And most of the times the machine learning was more accurate than the human forecast, but they still had that human involvement. It’s like, okay, well tell us why. If you’re gonna override it, give justification, and why is it different? And really starting to try to integrate the two. So, as I like to think of about it, the machine learning could help inform the human so that they make a better forecast.
Ashok Manthena:
That’s right.
Paul Barnhurst:
So, you know, kind of next question here. As we talk about AI , you know, particularly in finance, how do you see that changing the way kind of FP&A departments work?
Ashok Manthena:
Right. Alright. Now there are two ways, right? One is predictive analytics, as I, as I mentioned earlier, is take your historical data, take your future assumptions, feed it to a machine learning model and get predictions out of it. This is a high level, what, what is done in predictive analytics. Sure. Right? And you get a forecast and you or, or your prediction, you can start using it in any of your other processes. The second is Generative AI. Now, I think it has been six months so far that ChatGPT came into picture. And now our imagination of how we can use this new technology, and particularly with Generative AI, I think this is an iPhone moment where when, if you remember Paul in two, 2007, 2008, all the developers started building apps, right? These apps already existed in web, but then they brought those apps and they built it in, in, in mobile apps.
Ashok Manthena:
Of course, there are new kind of apps like Uber that came into picture because of the location. Sure. So this is that moment with generative AI. What is gonna happen is we are all gonna build systems or technology that already existed and also new applications using this. And even all our finance software will have these features in the future. How can we achieve things using, using just text as an input. So let’s start with predictive analytics, right? How, how predictive analytics can be used. Let’s say if we take an income statement, we do income statement forecasting every month, right? Most of the companies we do every month and then every quarter, every year. And we do budgeting as well. And then we do long range planning. And most of this process in, in big companies, what I’ve seen in midsize companies is there’s a line item owner, right?
Ashok Manthena:
Each line item, there’s an owner who predicts it, but does anyone know what this person does to predict that, right? Did anyone go into the process to find out how he actually predicts it? Of course, he’s accountable for the forecast and probably for the forecast accuracy. But we never ask him, what is your process? And is this process consistent with all the other line items? So this is where machine learning can really help. The best way to is to start with one line item. Take the historical data, take the future assumptions, and then automate it. Same thing with revenue forecasting. Revenue forecasting is a huge thing. Every, every company wants it to be much more accurate. Can we take revenue line item in an income statement and automate that forecasting process? Now, if you take revenue forecasting, how is, what are the drivers that impact revenue?
Ashok Manthena:
It could be your marketing expense. Usually there’s a direct correlation between marketing expense up to a point, right? To your, to your revenue. It could be an inventory if, if you are a retail company it could be market conditions. It’s macroeconomic conditions like interest rates and all these factors. Right? Now, how do we take these factors? How do we take this data and give it to the machine learning model? What machine learning model does here is the sensitivity analysis. It’s nothing. It’s just the sensitivity analysis. It does. What it means is that it changes different factors of it and see how it impacts your target variable, which is your revenue, right? And then it comes up with an equation with, with weights around it, with weights for each driver. So in the future when you say, okay, I’m, I’m gonna spend 20% more on my marketing, keeping all my other parameters the same, it’ll give you an output, it’ll give you a revenue prediction for the next year and 20% it’s more.
Ashok Manthena:
But you can always add other factors into it. What if interest rates reduce by percentage point? How will it impact? So you can always give that input as well and see how the modeling changes and how you get an output. So this is basically what is done in predictive analytics and coming to generative AI. So I know a lot of finance teams started already using generative AI like ChatGPT or writing emails, for finding, we look up formulas for Excel formulas. But the biggest advantage biggest benefit that comes from ChatGPT or more of large language models in finance is chat as an interface. Till now, Excel is our interface. Word documents are interface. Some of the ERP softwares are our interfaces. If we want look at a report, we go into ERP, run the report and we see that report, right?
Ashok Manthena:
But now chat as an interface is disrupt all this. So you can say it’s just how will impact my, my, my operations. I’ll give you an example where if I want to run a report to get my data for cash accounts, right? I want balances for my cash accounts for this month. How do I do it at this point of time? Either I go to Excel, run my Excel plugin, right? Which connects to the software and gives me the data. But for me to do this, I have to know all the URLs, how to connect, right? And then I need to know what are my cash accounts. Sometimes you remember what are cash accounts, but most of the times you don’t. When you have 500 different accounts in your income statement, you won’t remember what are all your cash accounts, right? But now you, you have to use your brain power and memory to come up.
Ashok Manthena:
If you don’t remember, go check your chart of accounts, find out all your cash accounts, put that in the Excel. So this process itself is gonna take 10, 12, 15 minutes for you, even if you’re an expert in it right Now, if you’d say in a chat feature, if you can say gimme balances of cash accounts for this month, the system will understand what are cash accounts, because it, it has that information already. It knows what are cash accounts, it goes and matches. What are my cash accounts? Okay, these are my cash accounts. And it says okay balances. Now it should ask, do you need a translated balances or do you need it in your local currency? Those are the questions it’ll be able to ask so that these are taken out of my mind as a finance user. So this is where the biggest value edition of generative AI happens is chat as an interface because we are adding all the intelligence to it, which makes our life much more easy as a finance user.
Paul Barnhurst:
So I have two follow up questions on this. The first is, when we talk about predictive analytics and, you know, writing algorithms, machine learning, whatever, trying to forecast revenue and things, what, what, one thing I’ve always heard is having enough data to get a level of accuracy. I don’t see that going away anytime soon. So what would you say to companies, I mean, there’s a lot of small companies that may not or may not have much data or new companies. So is predictive analytics, do you see it as mostly something for big companies? Or how do you see those kind of two things playing the fact that you need a lot of data and then there’s a lot of companies that don’t have a lot of data
Ashok Manthena:
Right? So the requirement of having data, whether it’s the size or the quality, it is kind of non-negotiable with machine learning. Right? You need to have good data. Let’s say if you don’t have data, what do you do? You still, I mean, if you don’t have a data as a human, what do we do? We go for guesswork, right?When I really don’t know what’s gonna happen, I’ll say, okay, let’s have this a hundred tons of inventory and see if it sells, right? So that, that guesswork will be done when there is not enough data. And it’s the same thing will happen with machine. When you just give very little bit of data and let it guess, the accuracy really suffers. So as a company, let’s say, if you’re a very small company, if you don’t have data, you can’t really do much with predictive analytics, but what you can do is start saving data for the future, right?
Ashok Manthena:
Not just with predictive analytics. Even for your, your chat, ChatGPT and, and for your generative AI is once you have good data in next 3, 4, 5 years, you, that data is very valuable for you to see how things will change, how you can automate this processes. That’s what that’s about, the size of the data, right? So if you have good data, three, four years of data, you’re good to start with. You can start experimenting with it, see if you get good results with predictive analytics. But if you have less than that, start saving it. Don’t never delete any of the finance data. . Usually when we, when we move from ERP to the new system, just to cut costs, you know that, right? We’ll, we’ll say, okay, I I don’t need this much data. I’ll take two years of data. My things changed a lot in, in the past few years, so I don’t really need that data. Lemme purchase that data, don’t do it. Even if it’s a little bit of investment, bring that into the new ERP, put it there so that you have all this historical data saved, which will be valuable for you in the future.
Paul Barnhurst:
Got it. No, that’s kinda what I figured. That’s what I wanted to get at, is, you know, predictive analytics, you need enough data, right? That’s critical. We all talk about it, and so many people, we should be using it to forecast everything. And it’s like within reason, you have to understand, you gotta have quality, and you gotta have enough of it that it can be valuable. And that’s where that human judgment comes in. So, second question around generative AI. You know, we hear, obviously everybody’s talked about it, all the amazing things it can do, but there’s two things that I wanna ask about. One bias in the data, you know, how do we overcome that? And second is accuracy. We’ve all seen it where it gives wrong answers, but it, right, it’s conversational, at least what we’re seeing with Chat GPT and a lot of these things. And so it’s gonna come across as if it’s right, even if it’s wrong. So what advice do you offer to finance professionals to be careful of those two things? To make sure, you know, the accuracy and the bias doesn’t become a major issue.
Ashok Manthena:
Right? so one thing we have seen, particularly for finance teams is don’t put any of your personal data or any of your company’s data on ChatGPT or any of these public things, right? Yeah. If there are products that come up with this application where they will separate your data, so there will be okay to do it, right? So don’t, I mean, if you want to just write an email, very generic email, yeah, not a problem. Use it. If you wanna ask a raise, yeah, you can, you can use ity to do it, but not with the finance data that you have. That’s one, right? Yeah. Second hallucination is a real problem with with all these large language models, but there are ways to actually curtail it, right? To make it much more accurate. For example, we are building ChatFin, which is, which is large language models for, for finance teams to go query your ERPs to run reports or to, to get all your data into one place.
Ashok Manthena:
Everything you can do it using large language models, using generative ai. There is a way, there are technological ways where how we can stop these systems to hallucinate and make it much more specific about the tasks. There is something called knowledge graphs that, that we build. What it means is that now if we want to, if we wanted to run a report and give us data, it’s not going to go and create that report. We’ll say these, these are the existing reports you have to go select, which is the report that is matching to the user input. Now that’s how we are, we we’re gonna regulate large language model not to hallucinate and give you a data, which is, which doesn’t even exist in the company, but be very specific about what, what are the ranges of things that are there. So there are ways how we can do it.
Ashok Manthena:
And that is very important for anyone who’s building a product with large language models for finance teams, make sure it’s not hallucinating, make sure the data is right. And there is also few user interface ways, how we can make sure to build trust with a finance user, for example you’re, you’re, you’re asking to trigger a workflow in an e R p, right? You’re saying, okay, I have to do this. Can you trigger this? Then it, the, the AI can ask you back the question and say, Hey, this is what I’m running, is this what you want? Right? Or this is, these are the legal entities that I’m going to run this job on. Is this what you want? So you can always build that iterative process to build a trust within the system. I think it’s gonna be exciting future
Paul Barnhurst:
De it definitely is. I think those are some great points you made there around one it coming back and explaining what it’s gonna do, kind of having that conversation so you know that it’s under understanding the question. There’s definitely ways you can prompt it, different things you can do. But yeah, one thing I like to say, especially right now is I like to think of it as kind of a junior, junior analyst or an assistant in that I would never take a junior analyst work and just give it to the C F O without looking at it. That’s right. Right. Same idea applies here is you need to make sure we’re reviewing it and not just blindly trusting that everything is right, or at least I would say do it at your own risk, right? That’s right. So, so can you talk just a little bit about what you’re building at ChapFin a little bit about how you think about, you know, generative AI and the future and how you’re playing a role there.
Ashok Manthena:
So our vision is to build the first general finance artificial intelligence. What it means is that it should understand anything finance, right? You ask it, you, you ask to perform a task or you know, you ask a question, it should be able to understand everything. Finance, right now we have a lot of large language models, but you go very specific about some data within your company, it don’t even work, right? So that’s what we are building. How do we take data within the company, like, you know, the existing data and also combine this intelligence so that it understands all the finance language, finance inputs, finance data. So that, that’s what FIN is about. It’s, it’s a digital assistant, it’s a platform agnostic. So it can connect to any of the finance software that you have, but it can, it’ll be like an assistant for all the finance teams. You ask a question, you ask, you can automate tasks. Imagine everything, right? Imagine querying data without, let’s say you are, you’re an Oracle person, now you have to work on SAP, you know how difficult it is to do it, right?
Paul Barnhurst:
I’ve worked on both. Both, yeah.
Ashok Manthena:
It takes a year and a half or two years to even actually get into that system and know everything, how things are happening. So take that the, the technology, the, the domain, the tech, the specific software knowledge out of it. Now you can have a finance analyst who’s a rockstar without that very specific technology knowledge with him, right? You can hire a lot of people with SAP background to your article or Anaplan or whatever softwares that you have. So that’s what we’re building, a platform agnostic, which can connect. All you need is the finance knowledge. And now your analyst will be a rockstar, or your manager will be a rockstar. Your, your CFO VPs. They don’t even have to rely on the teams to get some of the data. They can do it by themselves.
Paul Barnhurst:
Maybe I should go back to being an FP&A professional instead of this business gig. If it doesn’t work out, I’ll come find you.
Ashok Manthena:
That’s right. <Laugh>, if you want to be a rock star. Yes.
Paul Barnhurst:
Well, I, I do prefer being a rock star. I really don’t like being a low performer. It’s not what I strive to be. So. Well, you know, this has been a great interview. I could probably talk to you all day about, you know, ChatGPT, generative ai, how the world is Changing Before Our Eyes, but we’re coming up on the end of our time here. So I just have two more questions for you. So the first is, if we have someone listening in our audience that wants to start learning about AI and take advantages of tools available to them, particularly someone working in FP&A, what advice would you offer ’em?
Ashok Manthena:
So particularly FP&A, having basics of data science will give you an edge in the job you’re doing, irrespective of the tools that are available right now which, which can automate most of the data science tests, right? If you have time, I always suggest this to FP&A people learn some basics of data science, right? How data science works, what are the techniques, what are the statistical ways we can, we can use our data or, or even automating it. Have getting some Python knowledge is, is good. So you can use a lot of other tools to do it. Now with chat, with the getting your Python code, Ready’s not that difficult. So learn data science basics some coding language so that it’s easy for you to automate tasks. That is how you can get started. But with generative AI right now, maybe it’s too early for me to say it, I don’t think you have to learn anything specific about generative AI because the whole purpose of generative AI is to make things easy for you to use English right? To get your things done. And that’s how all the, all the softwares that are built is enabling finance user just to use their finance knowledge in English or any language basically to get things done. But data science is something that you have to learn though.
Paul Barnhurst:
Okay. Got it. Do you have a, a book or anything you’d recommend or maybe something we could put in the show notes for people that are looking to learn a little bit more?
Ashok Manthena:
My book, supercharging is coming in the next few months. So that would be a good start point where you can talk about, we talk about some of the use cases and how generative AI is gonna impact, but there are few courses, free courses on Coursera. Basically the data science. They’re not very finance specific, but you’ll learn there is for now, there’s no specific finance data science course that is available. Maybe, maybe I should start building it, right? I’ve been talking about it for a while, but I’ve never done it. <Laugh>, <laugh>. But there are a few courses on Coursera that you can start with. Data science basics of data science with Python. And that’s a good start.
Paul Barnhurst:
Okay. Thank you. Appreciate that. And we’ll put those in the show notes. So last question. If someone wants to get ahold of you, what would be the best way for them to contact you?
Ashok Manthena:
Linkedin connect me on LinkedIn. Send me a message what you want to learn. I’m really happy to answer any of your questions and if you want to know more about ChatFin ping me.
Paul Barnhurst:
Perfect. Sounds good. We’ll definitely put your LinkedIn profile in the show notes so people can reach out to you and appreciate your time today. It was a pleasure chatting with you. Thank you for joining us and hope you have a great rest of your day.
Ashok Manthena:
Thanks, Paul. Thanks. It’s an honor to be here on this platform. Thank you.