Transcript

Sebastian Taylor (00:13)
Welcome to another episode of the CFI Podcast. I’m Seb Taylor, subject matter expert for our BIDA program. That’s our Business Intelligence and Data Analysis certification.

And today, I’m excited to have Ashok Manthena with us. Ashok has many years of experience in finance and analytics, mostly in leadership roles. He’s an author and a speaker. And in 2023, he founded Chatfin.ai .

Thanks so much, Ashok, for being with us.

Ashok Manthena (00:41)
Glad to be here, Seb. It’s amazing to be talking about what we do every day, and I also want to learn from your experience as well.

Sebastian Taylor (00:49)
Absolutely. Before we dive into AI itself, I was wondering if you could just give us a brief introduction of what Chatfin.ai does.

Ashok Manthena (00:59)
Right now, I work for Chatfin.ai. Chatfin.ai is a generative AI tool for finance and controllers and tax and treasury. Basically, using the power of AI, can we do our finance processes? It could be reconciliation. It could be power flux analysis, variance analysis, data query, all this using the power of generative AI. Yeah, that’s what Chatfin.ai does.

Sebastian Taylor (01:05)
Yeah.

Awesome. And so to get a little more specific about that, I know you mentioned in a recent article of yours that for FP&A in particular, there were sort of three categories of benefits that you see for these types of tools. You mentioned data collection tasks, forecasting and simulation, and then real-time analysis. I was wondering if you might just be able to give a quick real example of a task that involves those things and how these AI tools can help us.

Ashok Manthena (01:56)
Right. So FP&A team or FP&A role itself is a very collaborative tool, right? Collaborative work where they talk to the business users, they get all the information. They also get all this different data from different systems. It could be your general ledger, it could be your CRM. They get all this data, they synthesize the data, and they process it and put it in a format that management really understands so that they can make a decision.

So this requires a lot of manual tasks at this point of time. Of course, ERPs, integrations, the planning cubes, all this really helped FP&A users, but still there’s a lot of manual tasks, just getting the data into the right place, getting the data in the right format. This is where AI can really help. How do we make that easy access to data within a company? If the data is already there in a company, it shouldn’t be difficult for a person to access, whether it’s finance or marketing.

I mean, as a finance person, you should be able to, if you have access, you should be able to access marketing data as easy as you’re accessing your general data. That way, as an FP&A user, you can bring all this data into your modeling and get your model accuracy much better than doing it in a limited scale. So data access, the data querying, which makes things a lot easier easier FP FP&A users, that’s one. And then getting the data and modeling is another part, right?

How do we model it right now? We model all our data. Even I’ve seen companies of a billion-dollar size still doing their modeling in Excel. There’s a reason why we model in Excel because it’s easy to access and also it’s easy to understand. You can easily understand the cause and effect of some data points, right? For increasing marketing, how is it impacting your revenue? Is it by .2, .4?

You’re just doing a linear equation and making it easy. But that also has some cons to it, right? So now, if you make your model very simplistic, the accuracy is really softened, and you’re not making decisions in a really data-driven way. You can use statistical modelings, you can use machine learning models to make your models much better, your accuracy is much better because these models can read all the patterns in the data, the existing or historical data and then…

predict it much better than we doing it in Excel. So those are the things that can really help. And real time analysis is a combination of this, right? Combination of easy access to data, this modeling. Now you can ask a question, it should be as easy as someone says, can you forecast my next quarter using my assumption in marketing? Let’s say I’m spending 20 % more on marketing. Can you predict my next quarter revenue? It should be as easy as that. Someone asking that question to you and

they’re getting the results, right? You’re not going to take that as it is. Of course, you’re going to double-check it. You’re going to double-check your assumptions. But having that ability to do that scenario analysis quicker, that’s going to be a game changer for FP&A.

Sebastian Taylor (05:01)
Right, okay. Okay, so you see the real -time analysis as like a combination of the other two categories in a sense. So you’re obviously working with companies day in and day out who are operating these kind of processes. What’s your sense of the extent to which they have adopted AI to help with these tasks at the moment?

Ashok Manthena (05:26)
There are a few companies, companies, again, I’m talking about large companies, medium and large companies who already started in this journey, right, adapting AI. The process usually is someone is a finance champion within the company becomes an AI champion, right? They are like, okay, I think we should bring in AI into our processes, see how it’s going to benefit. It starts from there. And then they take one process and they say, what if we apply AI? It could be your  FP&A, it could be your controllers,

reconciliation or a month in process. Taking that process and say, how can we start somewhere here, see how it impacts and then slowly we can expand it to other. So most of the companies are at this level right now, where they’re taking AI tools, applying it to one of the specific process, see how it’s going to benefit, how their teams are reacting to this new tool. Are they willing to share their process with AI? Are they willing to make AI automate their own process?

So, they’re trying to experiment with that. And this process is going on in a lot of companies at this point of time. But we’re still barely scratching the surface of all the things that we can do in finance. It’s going to take some time for us to actually go explore, maybe we can use it here. Maybe we can use it there. All those things to come out, all those things to surface, and start applying AI there.

Sebastian Taylor (06:37)
Yeah.

So it sounds like we’re in an experimental phase, at least from the perspective of most companies, where they’re just starting to dabble.

Ashok Manthena (06:57)
I would say yes, it is experimental phase, but it’s not the experimental for whether AI can work. It’s more about experimental about how my teams are going to react to it. That’s going to be a biggest hurdle to it, right? Let’s say if I want to automate a reconciliation process, if my existing reconciliation team is not willing to actually accept this new process, then what’s the point of having a new tool that’s doing it? So that’s what…

leaders want to really try and see how they can adapt faster.

Sebastian Taylor (07:30)
Since you mentioned that as a barrier, is that one of the biggest barriers you come across is people not being willing to adopt these new tools?

Ashok Manthena (07:40)
There are a lot of hurdles, right? One is probably there is a hostility towards AI. There’s always three countries. There’s early adapters of technology. There is some people who’s like, okay, well, we can see what is going to happen here, or maybe I’ll do it in the next few quarters, not right now, but next few quarters, I’ll take it up. And then there are always late adapters who are saying, they already analyzed the present situation and they’re saying, I think we’ll wait. We’ll wait until…

someone has tried it, someone actually publishes papers on it and say, hey, this works, this doesn’t work. So there’s always these three data sets, who we are dealing with is early adapters in this technology, who’s jumping in, taking the benefit of it or exploring the benefits of it at this point of time.

Sebastian Taylor (08:25)
In terms of the benefits that people can see when they do adopt these tools, you mentioned in a recent podcast, I think, that you estimated that roughly 70% % of an analyst’s time can often be taken up by operational tasks. What do you see as the scope for that to change with these tools?

Ashok Manthena (08:47)
When I say 70%, I’ve seen a lot of people telling me it’s more than 70% that time just goes in the manual work and getting the data extracts and doing all that. I think we should be able to reverse this. 30% manual tasks and 70% is all the inside generation and all the things that really matter. We should be able to do it in the next few years in reversing that. I think AI is going to help us in achieving that pretty soon.

Sebastian Taylor (09:15)
Yeah, that’s a huge time saving to be able to flip that ratio.

Ashok Manthena (09:21)
It is going to be. But there will always be something for us to do. It’s not that we’ll have a lot of time, if you’re an employee, it’s not that you’ll have a lot of time. You’ll actually do something else as well, in addition to what you’re doing here.

Sebastian Taylor (09:27)
Yeah, exactly. The never-ending list. Yeah. And besides time as the obvious benefit of these tools, are there any other benefits of using these AI tools that you think people should be particularly aware of?

Ashok Manthena (09:37)
The never-ending list.

I think the real time innovation that’s going to happen in the processes that we do. For example, we have at the month end, we have our data. Once our month end closes down, we have our latest data that came in. We finalized it. Now look at that. We have a new data set sitting in there in terms of how our revenues have been and how our expenses were. And we can take that data and if we can generate insights right away.

That’s going to help make us decisions for the next one month. So that thing is going to be significantly impactful in terms of decision-making. So it’s not just the time-saving. It’s the real-time analysis, how when things become faster, that’s going to impact our decision-making as well in organizations. Now they have a chance to react much faster than wait till you see the whole trend playing out there. So that’s going to be our biggest benefit.

Sebastian Taylor (10:20)
Mm-hmm. Yep.

Right.

Yeah. Okay, so not just the speed of the task itself, but the speed of the decision that can be made off the back of it. Yeah, okay, got it. Yeah. When people are using these tools, I think…

Ashok Manthena (10:53)
Exactly. Right.

Sebastian Taylor (11:01)
Probably one of the biggest buzzwords that’s flying around is this idea of prompt engineering. And from what I’ve seen, even with prompt engineering, there’s still a huge scope for people to, or for almost a weakness in the process to be the analyst asking badly worded questions or questions that aren’t specific enough. I’m curious, how do we overcome that problem? Is it a case of we teach people to use AI better?

Or do we teach AI to better understand the bad ways we phrase things?

Ashok Manthena (11:38)
So prompt engineering is one way of interacting with AI. And this is going to evolve pretty soon. So for example, if there is a prompt that worked for you, the ideal way is to actually save it somewhere. So that you can, of course, right now we don’t have places where it’s easily accessible for you to just go there. But that model of UI is going to change pretty fast.

Sebastian Taylor (11:55)
Yep.

Ashok Manthena (12:06)
It’s going to be a combination of chat, dynamic UI, where you know you can already save that process. For example, you have already done your variance analysis, and you know this prompt work. You should be able to actually save that as a button server. And it should be as dynamic as it is. You can switch between your chat, which is kind of an ad hoc request.

You want to change something, then using a standard prompt, you should be able to do it. But if you already know something that works, you should be able to do it. So.

Sebastian Taylor (12:21)
Yep.

Ashok Manthena (12:33)
This dynamic hybrid kind of UI is going to evolve pretty soon. And prompt and prompt engineering is not going to be a problem as we evolve in various dimensions, not just in terms of AI, but also in terms of the user interface and how we store it, how we save it.

Sebastian Taylor (12:37)
Yeah.

Right, okay, yeah. I guess we’re already seeing that with various AI tools that they’re becoming, it’s not just a text or voice interface anymore, but they’re more all -seeing. They can see what you’re doing on your screen and they’re able to respond and take what you’re saying in the context of what you’re seeing.

Ashok Manthena (13:12)
Right. Right. I think about, if you look at all our existing software or ERP, it’s always a fixed UI. If you open it, I open it, it’s the same UI, even if our roles are different, right? What if it can actually change as per the role, as per my need on that day, my BD plus three need is probably different from my BD plus 15 day, right? So what if it can change that and it knows what I usually select and it can show things like that?

Things will be much faster and easier for us to access those things.

Sebastian Taylor (13:44)
Yeah, yeah, definitely. How do you think either adoption already or adoption in the future will differ depending on the size of companies? Is that something you see differences in?

Ashok Manthena (14:01)
For sure. I think the AI adoption by size of the company and their motivation to do it, their intention to do it always varies. For large companies, there’s a lot of pressure that’s coming from the board, CFOs, CEOs, to see if they can incorporate some of the AI into the process. So that will be like money saving, cost saving, as a speed of decision making. This is what is driving in large companies. How do we…

How do we get things done with less number of resources or how do we efficiently use our existing resources? That’s what drives in large companies. But when we’re talking about mid -size companies, that usually the number of people is less. They don’t have resources. So if you give them a tool and they say, hey, I can do this process for you, they’re very happy to adopt that product. But they need to have the clear ROI, clear use case available so that they can go adopt to it. That’s with the mid-size.

But small size companies are different. Small size companies needs totally end to end, end to end working of a use case for them to even actually do it. For example, and also there’s also a lot of differences at a small size companies. One company of probably a $10 million company, where it’s totally different from other $10 million company. So that kind of changes how the small size company is going to adapt. I think we need to get more tools

and also more customized tools for specific use cases, specific industries that will come into the market that will serve these various needs of small companies.

Sebastian Taylor (15:40)
Yeah, that’s interesting. I’m conscious we’re running out of time, but there’s one more question I wanted to throw at you, which is, I guess from our listeners perspective, most of our learners are working in Excel, day in, day out. Some have played around with some of these AI tools, some haven’t. If you have one recommendation of a tool or a skill that you suggest people should learn first, you know,

in order to get onto this AI revolution, for want of a better word, what would that suggestion be?

Ashok Manthena (16:19)
Learning is different, right? The idea of using AI is that so that we don’t have to learn at all. Any tool, we should be able to use it very easily rather than having the steep learning curve to start using a product. AI is going to help that. ChatFind as a generative AI tool works very well with Excel. For example, you have all your logic built for your modeling in Excel.

It can actually help you populate the data and also take the data from there. It will help you with visualization and all that seamlessly with Excel and also with other ERPs. So Chatfin.ai is the one tool I would suggest people try it and learn more about it.

Sebastian Taylor (17:05)
Awesome. Well, thank you for the suggestion. This feels like a good place to end, so thank you so much again for joining us and thank you everyone for listening.

Ashok Manthena (17:16)
Great conversation, Seb. Thank you.

Sebastian Taylor (17:18)
Thank you so much.