AI Strategy Finance Leaders Must Use to Avoid Siloed Tools and Failed Adoption with Vikram Bhandari
In this episode of FP&A Unlocked, Paul Barnhurst and Glenn Snyder are joined by Vikram Bhandari to explore how organizations should approach AI in finance and FP&A. They discuss why most companies fail with AI adoption, the importance of enterprise-wide strategy over siloed tools, and how finance teams can evolve from reporting to forward-looking decision-making.
Vikram shares practical insights on AI readiness, forecasting, and how leaders can leverage AI to drive better business Outcomes.Vikram is the Chief Technology & Innovation Officer at Riveron, where he leads AI, digital transformation, and finance modernization initiatives. With nearly 25 years of experience, Vikram works closely with CFOs to transform finance functions through technology. He previously served as the President and CEO of Yantra, a company he founded and led for over 15 years before its acquisition by Riveron.
Expect to Learn:
Why most AI implementations fail and how to avoid common mistakes
The importance of enterprise AI strategy vs. siloed tool adoption
How FP&A is evolving into a forward-looking, decision-making function
What it takes to prepare your data, processes, and governance for AI
Here are a few relevant quotes from the episode:
The real question is not which AI tool is best, it’s whether your finance architecture is ready to absorb AI.” - Vikram Bhandari
If you treat AI as a feature, you get incremental gains. If you treat it as a strategy, you get competitive advantage.” - Vikram Bhandari
Vikram emphasizes that AI should be treated as a capability multiplier rather than a shortcut. He highlights the importance of strong data foundations, governance, and enterprise strategy to unlock real value. The episode reinforces that the future of FP&A lies in combining machine intelligence with human judgment to drive better, faster decisions.
Campfire: AI-First ERP:
Campfire is the AI-first ERP that powers next-gen finance and accounting teams. With integrated solutions for the general ledger, revenue automation, close management, and more, all in one unified platform.
Explore Campfire today: https://campfire.ai/?utm_source=fpaguy_podcast&utm_medium=podcast&utm_campaign=100225_fpaguy
Follow Vikram:
LinkedIn: https://www.linkedin.com/in/vikrambhandari/
Company: https://riveron.com/
Website: https://www.tutorialspoint.com/
Follow Glenn:
LinkedIn: linkedin.com/in/glenntsnyder
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] – Intro
[02:08] – Introducing Vikram Bhandari
[03:16] – What Great FP&A Looks Like
[05:21] – AI in Finance Today
[05:09] – Why Most AI Implementations Fail
[11:05] – Enterprise AI Strategy vs. Siloed Tools
[15:16] – AI Readiness: Data, Decisions, Governance
[17:54] – AI Forecasting: Calibration & Trust
[23:14] – Skills Finance Needs in an AI World
[34:24] – AI Agents, Excel, and Productivity vs. Strategy
[41:40] – Final Advice on AI Strategy
[47:52] – Closing Thoughts
Full Show Transcript
Host: Paul Barnhurst (00:00):
Welcome to another episode of FP&A Unlocked, where Finance meets Strategy. I am your host, Paul Barnhurst, AKA the FP&A guy, and this week I am thrilled to be joined by my co-host, Glenn Snyder. Glenn, welcome back. Thanks,
Co-Host: Glenn Snyder (00:14):
Paul. Always love being here. Always
Host: Paul Barnhurst (00:16):
Enjoy it. Glad to have you brack. So each week we strive to bring you conversations and practical advice from thought leaders, industry experts, and practitioners who are reshaping 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. Speaking of impact, our title sponsor for FP&A Unlocked is Campfire, the ERP. That's helping modern finance teams close, fast, and scale faster. So today I'm thrilled to be joined by Vikram Bhandari. Vikram, welcome to the show.
Guest: Vikram Bhandari (00:57):
Thank you very much Glenn and Paul for having me on the show.
Host: Paul Barnhurst (01:01):
Excited to have you, Glenn. I know you know Vikram much better than I do, so I'm going to go ahead and let you do the introduction of him here.
Co-Host: Glenn Snyder (01:08):
Yeah, well, Vikram and I both work at Riveron. Vikram runs our technology group, and he came over from a company that he had started called Yantra. But Vikram, you're better at introducing yourself than I am. So I'll turn it over to you and you can talk about your phenomenal background.
Guest: Vikram Bhandari (01:24):
Yeah, thanks Glenn. So as Glenn said, I am the Chief Technology and Innovation Officer at Revon, being in the industry for close to 25 years, working at the intersection of finance, technology, and transformation, helping CFOs modernise their finance functions. A big part of my focus today is AI and how it's reshaping businesses, including FP&A.
Host: Paul Barnhurst (01:48):
Very excited to get into the topic of AI and FP&A. I think everybody's excited, sometimes a little scared by it, and everybody's trying to figure it out and how they can get the most benefit from it. So we're really excited to be able to pick your brain and get some of your thoughts on the subject. I know you're talking to, as you said, the CFO and finance leaders all the time about this and seeing what they're doing to try to get that benefit.
Guest: Vikram Bhandari (02:16):
Absolutely.
Host: Paul Barnhurst (02:16):
Before we get into the topic, I like to start, as Glenn knows every episode with this question. So you've run your own company, I'm sure you've done budgets and forecasts and worked with the FP&A team. From your perspective, from your seat, what does great FP&A look like? What is great FP&A?
Guest: Vikram Bhandari (02:33):
Paul, as you must have seen, the responsibility of FP&A teams have changed over the years. A good FP&A team does not just report numbers, they shape good decisions or they help us make decisions. They translate financial data into forward-looking insights, actionable scenarios, and again, they're commercially sharp, wide into the business, not stuck in spreadsheets. When I was running my own company, my team would actually give me models and scenarios and clarity under uncertainty. Those are good FP&A teams where the CFOs and CEOs can rely on them during uncertain times and for building the right models to be successful. Again, this is increasingly, technology is helping this, especially AI is accelerating this shift.
Host: Paul Barnhurst (03:27):
Agreed, and I really like something you said of having great FP&A help you make better decisions, not just report. Because to me, that's one of the biggest differences between what I'll call average FP&A is just doing the reporting, doing the basics, not helping the business make smarter decisions.
Guest: Vikram Bhandari (03:49):
Yeah. Again, that's where I've seen evolving in the last 10 years, FP&A was focused just on reporting, so it was not forward looking. It was backward-looking reporting. Now it's forward-looking decision-making. They're embedding themselves into the business rather than just playing in the spreadsheets. I'm seeing that shift happening lately. Yeah,
Host: Paul Barnhurst (04:11):
No agree. We've definitely been seeing that shift and it's exciting. It's a good one. So I appreciate you sharing that. Alright, Glenn, I'm going to turn it over to you to get us started on the AI topics.
Co-Host: Glenn Snyder (04:21):
Yeah, well actually, and before we go there, I just want to say I echo what you guys were saying about to me, great FP&A. It's about influencing the next decision, and that's the thing is what decisions that have already happened in the past, you can't change it. You could evaluate, you could analyse it, well, what's the next decision and making sure it's the best one for the company. And you do that through data, you do that through analysis, and so a hundred percent agree, but let's dive into ai. That's what we're really here to talk about and AI is such a big topic right now and Vikram, I'd love to get from your perspective in thinking about not only what you have seen but also kind of what riveron has been doing for some of its clients and how did AI really help enhance finance departments today? What are you seeing for some of the best applications and where AI is really making a difference?
Guest: Vikram Bhandari (05:09):
Most companies evaluate AI solutions the wrong way. They compare features, they run flashy demos, maybe do a pilot and wonder why adoption fails, right? So the real question is which AI tool set is the best? It's not the real question. The real question is having a finance architecture mature enough to absorb AI without breaking, right? So there's a few things which we are doing at Riveron for our clients, right? We're trying to evaluate AI solution problem alignment, not feature density. Again, most clients end up looking at features that don't align their problems and then evaluate AI data and integration fit. Again, how does it integrate into an existing data ecosystem? How does it cleanly connect to ERP CRMs and planning platforms? Does it require heavy data transformation? All of these need to be thought through when we are evaluating AI tools. And third, can the output be trusted and defended?
(06:22):
Again, we were just talking about this is the model explainable? Is there transparency in assumptions? Are there audit trails? All of these are very, very important and they're not optional for finance teams, they are mandatory for finance teams. So these are a few things which we are doing at riveron. This is where our teams are helping our clients and go through the entire AI journey. One thing which I personally feel is the most overlooked question is who owns it? Is it the finance team owning it? Is it it owning it? Do I need a data scientist to own it? As soon as you have, if the finance teams have dependency on a vendor or on it, we are seeing that adoption fails or adoption is slower. So the biggest mistakes companies make in thinking AI selection is a technology decision. It's actually an operating model and governance decision. To me,
Co-Host: Glenn Snyder (07:28):
It's funny because Paul and I had a conversation a few months ago. One of our topics was around analysing FP&A solutions and we talked about the same thing and it was making sure you're understanding what the needs of the business are, not just going over and fitting in a solution because it has certain bells and whistles to it. I think that it, it's funny, I never thought of looking at AI in that way, but you're right, it's because you've got to go over and solve the business problem, not just go over and look at, well, which one could go over and produce this report for me because they could all end up doing that. But it's about how it interconnects and makes the company overall more efficient and effective.
Guest: Vikram Bhandari (08:11):
And again, everybody wants to engage with ai, but nobody's thinking of readiness. When our teams run these AI readiness workshops for our clients, what we are typically seeking is identifying high value use cases, conducting data readiness assessments, mapping the workflows where AI would live. Most of the leaders I'm talking to are saying, how do I plug in AI into finance? They're not thinking that they're trying to think feature, they're under the pressure of the board to plug in AI in some shape or form. They're not thinking enterprise wide strategy at this point, which is where again, there's a difference between forward looking leaders and leaders who just want AI embedded in their tools.
Host: Paul Barnhurst (09:03):
And I'm curious, why do you think that is? I mean, you mentioned pressure from the board, but we see this so often, not just AI, but technology in general. It often seems like you focus on features and solving a problem versus a broader strategy and asking yourself why it often seems a challenge with any transformation, whether it's ai, digital, whatever you want to call it, implementation. Why do you think that's such a struggle to really step back and focus on the why and what you're trying to accomplish and governance before just jumping in and, oh, I need something that does this and I want these features.
Guest: Vikram Bhandari (09:41):
AI is one thing where I think it is going to be the most transformational. In the last decade, we have seen anything like this's impacting technology. It is the hype or is that it's going to build efficiency. The hype is that it'll help reduce manpower needed. The hype is that there are human workforce and AI workforce. So as soon as these multiple things come into play, everybody's struggling to say, where do I start? And everybody and most people are starting at the wrong place by saying, oh, let's do something with ai. Let's experiment to one scenario in finance. And what happens suddenly is you have pilots going on in different functions of the firm without any enterprise strategy, without any enterprise governance, without any enterprise level risk management happening. And that's where I think the pressure on the CEO and the CFO is, let's just start. They're not thinking, let's take a pause, have an enterprise strategy, let's fix our data sets, because AI will actually sprawl a huge output, which you'll get drowned into if you do not do ai. Right,
Co-Host: Glenn Snyder (11:05):
Vikram, so we had a conversation a couple weeks ago, the three of us when we were talking about this podcast and you brought something up that goes right along the lines of what you're saying is that another issue when you're not thinking about the AI strategy is you go out and you have different systems with their own AI components but are not working together to go over and solve that broader thing so that you could go over and have an ERP solution with an AI component. You can have an EPM solution with an AI component, but if they don't know about each other, they could actually be looking at similar data come with two different answers. And you need that overarching AI AI strategy to connect everything. So not only can AI work better in each solution, but it can actually learn from the mistakes that it makes in a more aggregate solution. Can you talk a little bit more about that, why that's so important?
Guest: Vikram Bhandari (11:57):
One of the things people are trying to shortcut is if I have a platform, whether it's E-R-P-C-R-M-E-P-M, what are that may be? Each of the platforms today are actually having their own AI components, and rightly so, right? Each platform will they have ai. So I have an ERP, which is AI enabled. I have a CRM, which is AI enabled. I have an EPM, which is AI enabled, but all these AI agents do not talk to each other. So what is happening is you are still stuck in siloed data sets with AI agents working. Am I able to do a conversational ai? Do I have a conversational AI which cuts across the data sets and give me a response, right? I will not be able to have a decision-making AI if I'm not cutting through multiple data sets. So decision intelligence platform is very important. Again, integrating predictive models is very important.
(13:01):
Instead of we asking what happened or what will happen, are we capable of asking what should we do and that what should we do? Response can only happen if you have an enterprise AI strategy. It'll not happen if you have a platform AI strategy in a platform AI strategy, you can say what happened in an enterprise AI strategy. You can say, what should I do if this happens? There's a huge difference, and if you look at it, in my experience, everybody's talking about gen ai, the world of AI is companies cannot be thinking AI in a single category. It has to be predictive models, it has to generate signals, it has to have optimization models, it has to recommend actions. Generative is probably the third or the fourth layer and where it communicates insights and together they need to create a decision system platform, AI's siloed platform ais will not be able to create a decision system.
Host: Paul Barnhurst (14:15):
I really like what you said there on the whole idea of decision system and really getting AI to go to the next level. The other thing that I appreciated is reminding people that there's more than generative ai. I talk to a lot of these different tools and almost all of them, especially in finance, start with deterministic and machine learning and then they use jet AI in certain use cases typically with the human in the loop for that part to validate it. Yet everybody runs out, wants to use the gen AI side. And so it's so much more than that and I'm grateful you brought that up because you really got to think holistically and speaking to that, you talked about the enterprise, but how does a company even know they're ready for ai? What are those things they should be doing from a prep? You mentioned about data and AI readiness. What are the things that tell them I'm ready to move forward with an enterprise AI strategy,
Guest: Vikram Bhandari (15:16):
Whether we are ready or not, every firm needs to, or every company needs to go through an AI strategy. Now there's three or four things I would typically want companies to think of. One is your data ready, right? The biggest piece of AI readiness is identifying high value use cases tied to measurable business impact. Not just having to do something. I want to do something with the ai. Did I identify that? Conduct a data readiness assessment, map the workflow where AI will live, establish probably a governance or a validation protocol upfront. We normally tell our customers that if you can't define the decision you're trying to improve, AI won't help you. You need to understand or define the decision you are trying to improve. That's one. AI is going to generate a lot of output. We also tell them that AI will expose every inconsistency in your data and forecasting process, and your teams will have those uncomfortable moments in the near term and they should be ready for it. So when somebody says, are we ready for ai? Again, we want to get in, understand what AI in your industry, in your company, in your function areas mean, what decisions you are trying to improve, what data sets are you trying to consume from? We'll come in, we look at that holistically and then say, you need to do these three things or four things before even you embark on an AI project or select an AI tool.
Co-Host: Glenn Snyder (17:11):
Vikram. One of the things that I think a lot of people, especially in net p and a, when they start thinking about ai, they think about forecasting. I remember I was talking to a representative from a particular vendor and he said to me, you know what? You're going to need to run your models probably for about six quarters or so before you're actually going to get some meaningful forecast data. It'd have to keep on learning through each iteration. What are some of the tips that you could give a company when they're thinking about moving forward with AI and thinking about AI for forecasting that can either speed that up or just improve the process and make sure that you're setting the right expectations for what AI can do and what AI can't do.
Guest: Vikram Bhandari (17:54):
One of the things which I tell most of my customers, and I want to clarify with most of my customers is AI does not learn in abstract. It learns within the boundaries of your data quality, your business stability, your model design. So if I was to put a timeline on how long it takes, probably I'll break it into a couple phases. Phase one would be calibration, right? Take a few forecast cycles, again, two to four forecast quarters depending on data history and volatility. During this period, we are benchmarking AI outputs against human forecast and historical accuracy. So that to me is more or less phase one calibration validation happening in that the next phase typically is trust building, right? Even if the model performs statistically well, confidence comes from explainability. We have been talking about it repeatability. This usually takes additional cycles. So again, in my mind it is two to four quarters on calibration, couple quarters on trust building. At no point in time AI is going to be plug and play. It's iterative and it has to be gone, right? Where companies go wrong is assuming more data equals better results. It actually is better structured data and tighter feedback loops, which we will have to incorporate into these two phases of calibration and trust building, which will actually help them get accurate results. And as I said, the role of fp n is changing from reporting to forward looking decision making, and all of these needs to happen pretty quickly and AI is helping that.
Co-Host: Glenn Snyder (19:52):
Just to expand on that, do you recommend that companies go back, let's say if they had six quarters of prior forecast to start running some of the stuff historically and doing the comparison to maybe cut down on some of that lead time of when people could start getting confident. That's why they say, Hey, look, we get our forecast from a year ago and said, ai, what can you do? And here's what it learned. Here's how accurate. Is that a good approach to maybe save some time?
Guest: Vikram Bhandari (20:21):
Absolutely. Because we have historical, we already have that historical data, so what we do is we run AI on top of historical data instead of looking on forward-looking data, right? So you run AI on historical data, you save a lot of time by doing that. You see if the models match, you see the data sets match, so the reporting pieces are taken care of by running it on historical, you can validate both. We need to stress this based on current business climate also, that cannot happen on historical data unless you feed in a lot of what happened around the business, macro, micro economy and so on and so forth. That can happen only forward looking based on current business, current business climate, what's happening, what's not happening in the industry. AI is smart enough to get that couple IT with your data set and help you make decisions, and that's where the real power of AI comes in. The power of AI is not just generating reports and validating that more accurately.
Host: Paul Barnhurst (21:26):
Leads me to a couple questions. First, something you mentioned is the importance of transparency, flexibility. We've all heard the term black box in finance and I think that's one of the things that has allowed the spreadsheet in Excel to be such a prominent tool is you can always open it up and trace it all out and figure out what's going on. You just have to put in enough time depending on how messy it's, what skills do we need to get comfortable as we go more and more to AI and more machine learning forecasts and things? Do finance people need to understand data science or what's your advice there? What level of skills do you think we need to really be able to understand as we go more and more to machine generated forecasting? And now, a brief message from our sponsor, Campfire. Build a best-in-class AI-native ERP designed for modern finance and accounting teams. With Campfire, you can automate revenue, streamline your close, and centralize accounting, all in one unified platform. And now they've hosted Finance Forward, a first-of-its-kind AI summit in San Francisco. They brought together the sharpest minds in finance and operations, all focused on how to actually use AI to move the needle. You don't even have to travel to learn the tactical skills to put your team ahead. You can stream the whole thing on demand, totally free, anytime after the event. So if you're ready to get tactical about AI and finance, head to campfire.ai and register for the on-demand content for free. That's Campfire, where finance teams go to think forward.
Guest: Vikram Bhandari (23:14):
I don't think we want to convert the finance leaders as data scientists, right? Two different skillset sets. Each one does each one.
Host: Paul Barnhurst (23:25):
I agree.
Co-Host: Glenn Snyder (23:26):
Everyone's going, alright, good.
Guest: Vikram Bhandari (23:31):
Let's start at that. I do not want finance leaders to become data scientists. That's one. Models are built, AI tools are available. The whole science and the magic, whatever you want to call it, are happening within the tools. I want the finance leaders to not even be prompt engineers, but to understand how AI works so that they can leverage AI in decision making. Today everybody's leveraging AI on what happened, kind of questions. I see my customers saying that, why is my sales down? Oh, it was down because X, Y, Z reasons. I can get you a narrative, but the next question they should be asking is, what do I need to do to fix the sales in the next quarter? Right? Again, things like those, so not data scientists, probably the right prompting needs to happen by finance. Probably the mindset change needs to shift from digging into history versus forward looking, and I think the skillset is a little bit more fungible on data. Scientists need to know a little bit of finance while they're building the product and the finance guys need to know a little bit of prompting so that they can leverage the product better.
Co-Host: Glenn Snyder (24:55):
It's funny when you were saying that, I was thinking about when I joined a new company and I'm a finance business partner, I have to go out and understand the people that I'm supporting, how they think, how they want to work, and how do I best work with them. In a way, AI is the same thing. I'm going to have to go over and think of being a business partner to AI to say, what's the best way for me to get what I need out of AI and how do I best work with it? So it's the same almost business partner skillset that you have in FP&A, you're now applying it to a tool rather than a person.
Guest: Vikram Bhandari (25:31):
Yeah, absolutely. So even if we implement, so I keep AI on the side for a minute. When the best ERP developers or even in our team at riveron Glenn, the best ERP implementation consultants are those who understand technology and finance both because they are able to relate to a finance problem and apply a technology solution to it, and the best finance users of ERP are those who know 90% finance, but 10% on how to leverage technology to get their financial outputs. So that skillset does not change. You keep changing the technology in between that skillset sets on. Both ends need to coexist how to do business and business and technology, and on this side who are developing it's more technology and business. That combination has to coexist.
Co-Host: Glenn Snyder (26:32):
And the funny thing is, is that it also then you layer in the best finance fp a, people are also able to learn and grow, start seeing a bigger picture and develop more strategic questions and strategic solutions. And the same thing would be the case in AI when you're working with us. So I think it's a great parallel there because it is the same skillset. It's something that we most people in FP&A do naturally to build out that trust in. The relationships start expanding, and it's the same thing you got to do is you might start smaller with how you're going to prompt AI to figure something out, but then you want to move into that bigger strategic questions about where are you going and how are you going to impact decision making to really get the most value out?
Guest: Vikram Bhandari (27:17):
Absolutely. Yeah.
Host: Paul Barnhurst (27:19):
One thing you mentioned is the importance of getting better at prompting something we've seen a lot of with ai. I think a lot of people say they're AI experts and they really just, they've got good at prompting, right? They really don't understand the underlying technology, and two, we've seen a tonne of what I'll call people sharing just a prompt, prompt packs and come see the prompts to use. What's your thoughts there of how do people get good at prompting? It feels to me like prompt packs are kind of the old way of saying just give me the answer whether it's really right or not. AI feels like it's much more of a process to prompting, but I'd love your thoughts there.
Guest: Vikram Bhandari (28:02):
One thing I need everybody to understand is AI gives you a response based on the persona you are. Again, define your persona first. Am I asking this question as the CFO of the firm? Am I asking this question as an FP and expert? Am I asking this question as an AP analyst? Define your persona to the AI because if you ask based on the persona, it'll get you the slice accurate for your role within the company, right? So that's one. So what I typically do is my AI is trained based on, it knows I'm the CTO, it knows what my role is, it knows what kind of work I do. So when I ask my AI any question, it actually gets me a response which has a technological angle to it or it has got a different slice than a CFO would ask. That's number one.
(29:06):
Two is as you keep prompting, you need to keep training the AI saying that, Hey, you got me the wrong answer. You got me the right answer. You got this is right, this is wrong. Refine this for me. Do this. As you keep doing it, the AI understands what you are ultimately looking for. So yes, prompting, it'll take a few iterations or a few weeks or a few months for the AI to learn who you are to AI to understand what kind of prompts you use. Because you may be asking for example, what was my sale in Chicago market versus I may be asking how much technology I sold in Chicago market versus Glenncould be asking how much accounting advisor has sold in account three different questions. It could give me three different answers or it could give me one answer with one blurb saying that Chicago was a hundred thousand dollars this month. So the slice, the thought process, the persona AI is learning all the time. Again, I think most clients think that they have one right prompt and it's going to work every time. That's not true. AI will learn from your prompts. I dunno, it was a long wandered answer, but that's my take on it.
Host: Paul Barnhurst (30:23):
I really liked the answer at the end where you said they're like, there's not one prompt answer. There's not just a can prompt you can use for everything because like you said, it's going to learn, it's going to adjust. It's like with business and Glenn's example, I can't go to every business partner and ask 'em the same question.
Co-Host: Glenn Snyder (30:42):
I
Host: Paul Barnhurst (30:42):
Got to customise them. And I think there's that with ai depending on the dataset and what you're trying to accomplish and your persona. So I really like how you said always start with the persona. That's something I try to do. Tell it who I am, tell it what I'm trying to accomplish, give it the constraints. You don't want to go overboard and give it five pages of information in your one question, but you want to give it enough that it can tailor what it's giving you. As I've always heard vague question, get a vague answer, And now, a brief message from our sponsor. Is your AI thinking like a generalist or like a finance leader? Here's the thing: most large language models weren't built for finance. They're trained on everything from Reddit threads to recipe blogs. So when you plug them into your accounting workflows, the results can be, let's just say, not balance sheet accurate. That's why Campfire, the company pioneering AI-native ERP, just made a huge move. They've raised $65 million in Series B funding, co-led by Excel and Ribbit, bringing their total to $100 million raised in just 12 weeks. And they're putting that momentum to work with LAM, the large accounting model. It's a proprietary AI trained specifically for finance and accounting, already hitting 95%+ accuracy on accounting tasks. That's redefining what ERP can do for modern finance teams. Unicorn companies like PostHog and Ripple are already using Campfire to automate their finance operations and seeing real results. You can see it for yourself, sign up for a live demo of LAM at campfire.ai. That's campfire.ai.
Guest: Vikram Bhandari (32:34):
Hallucinations are real within the bound. Even if you're working within a set data set within the boundaries of your systems, you still can get hallucination if the prompts are not right. A lot of CFOs ask me, why is this hallucination? I'm not going to the web and asking that question. I am within my data set, I still get different answers. Different people have different answers with similar questions, and you have to train AI on what you're looking for. You have to tell AI this is wrong versus this is right and it'll keep learning over time.
Co-Host: Glenn Snyder (33:12):
I think that's actually one of the biggest changes because for people in FP&A, we've worked with systems and data for decades, and we would go into our system and we would run a query and say, show me to your example. Show me Chicago broken down by service and what sales would be and we could run a query and it gives us the data, but the systems we were working on weren't learning as we were going. It wasn't trying to understand what we were trying to do. It was answering the direct question we were effectively asking. And AI is a little different where it's going to go and learn and start to anticipate and think about where you are, what you're trying to do to give you a better answer and to give you a more complete answer. And it's a different mindset than just going over and saying, Hey, I'm just going to query. I need this data. You got to be thinking almost more like you're talking to a person. So that person, if you have someone on your team and you keep on going to them with the same question, after a while, that person on your team is going to know, Hey, on Tuesday, Vikram's going to come over and ask this question. I better have here.
Guest: Vikram Bhandari (34:20):
Yep. It's similar to that. It's very, very similar to that analogue lens.
Co-Host: Glenn Snyder (34:24):
It's a very different mental approach and how people go over it, start thinking about how they're going to get the data they need and it's almost like having another person. Which kind of leads me to some of the next thing, which was like agents and how agents are working. I know certainly we were talking right before this stuff about Excel agents and that's something that's coming up a lot. Tell us a little bit about your thoughts on what you are seeing and what makes agents work well and where maybe sometimes people have trouble working with agents.
Guest: Vikram Bhandari (35:00):
As we were talking earlier, right? People we need to think of a few things when we are talking of agents. Again, Excel agents is one piece where Excel, I call it more plugging plugins on Excel, whereas gen AI and other agent to agents are a little bit different where we just talked about, let me break this down into a few things. There's a gen AI layer which stops to an agent. There's a predictive and machine learning model, which is there. There is anomaly detection and pattern recognition models. There are process and embedded intelligence models. And finally there is this whole decision intelligence platform. So I may be talking a little bit too technical here, but when I talk about agents, these are four or five stacks which companies need to think through, not as finance and FP&A guys, but from the technology teams.
(36:06):
All these four or five things need to start working together for a good FP&A or for a good business tool to be working. Again, AI workflows in finance, where I see typically is revenue forecasting, demand forecasting, churn prediction, working capital modelling, so on and so forth. When we talk about anomaly detection and pattern recognization, it's about we typically throw in unsupervised learning models, which are extremely powerful in surfacing what humans can't see. And then process automation, again, embedding RPA and ML together to actually transform close processes, reconciliations, reporting workflows, and then finally the whole decision intelligence intelligence platform where I can ask what we should do rather than asking what happened. So these are the three or four and each one is, think of it as each one is an agent. The reason I went through this whole thing is each one of these things are agents, and these are so different than a plugin on an Excel spreadsheet, which is different.
(37:30):
Again, we just talked about that chat g PT came up with an Excel plugin or an Excel agent, but these are different. These are enterprise level, industrial grade enterprise strategy governance. If you have these five, six things, then you are done. You can't just have saying that I have gen AI or I have an agent to agent talking or a platform agent. So together this is a decision making system. I said that earlier that if you don't have all of these, then you're working in silos. So all the agents need to talk to each other.
Host: Paul Barnhurst (38:05):
Listening to you say that, it makes me fear that what we're doing in Excel, these plugins are basically creating another silo. Would that be kind of your take all these Excel agents in some ways because they're really standalone, they're not part of a broader strategy, they're not hooking into other layers. Now we're starting to see some of 'em create workflow tools and bringing different data in which starts to get into that broader strategy. But do you have concerns with these Excel agents that we're almost taking a step back in the sense of we're not really integrating them into the strategy we should,
Guest: Vikram Bhandari (38:44):
We are not, but we should also be practical, right? Excel is powerful because it's the operational backbone of finance today.
(38:54):
Why is so much money being thrown into building these Excel agents or Excel plugins, right? It is the backbone of finance. It's not going away anytime soon. So that's the practicality of it. So embedding AI directly into an environment and analyst, again, there's a low friction for any analyst to use an Excel plugin, high adaptivity, low friction, but what is it doing if you double click on it? It's automating your formulas. It's generating models from prompts, it's summarising your data. It is running quick scenario analysis for FB and a, it's like a productivity tool to me, right? It's not an architectural solution. So the day the entire enterprise strategy has these Excel plugins part of the strategy, that's the day it's going to actually start making sense. Today it's in a silo, it is being used in a silo, and I'm pretty sure in the coming years, I think this will be part of the whole architectural solution where it'll be more traceability, model integrity, reproducibility, all of these things will start happening with Excel models also.
Host: Paul Barnhurst (40:17):
And I would agree with you. That makes sense to me. I mean, I think we're already starting to see some of that of how do we layer in the different tools and make this much more, as you call it, enterprise than just a productivity tool. At the same point, we all know spreadsheets are the backbone of finance. We've all seen one image or another that says, if Excel goes away, the whole financial backbone of the world is in trouble if we pulled Excel out tomorrow. So I really appreciate the practicality. I've seen a tonne of stuff I can do and benefit I can get using the agents, but I just wanted to get your take of how to balance that as we've talked a lot about the importance of that enterprise approach, but also there's times you have to be practical and go where you can get a lot of productivity as you work through things.
Guest: Vikram Bhandari (41:04):
Yeah, absolutely.
Co-Host: Glenn Snyder (41:06):
So Vikram, as we are coming up on about 45 minutes here into the podcast, want to kind of think back about everything that we've discussed here, and I mean I know you've hit several times on the importance of a company-wide AI strategy. So thinking about all the different things that we've covered here, what advice do you have for companies who are thinking what all my competitors are going to be in ai? I got to get in there, but I don't even know how to go over and start thinking about an AI strategy.
Guest: Vikram Bhandari (41:40):
So again, if I was to summarise this and leave the audience with a few thoughts here is there's a couple themes which come to my mind. Think of AI as a capability multiplier, not a shortcut. Majority of the discussions I'm having with customers is shortcut. I need something to be done on ai. How can you help? Right? Then not thinking of it as a capability multiplier. AI will amplify whatever foundation you have. So fix your data, get your processes right? Governance has to be strong. This will all accelerate performance. So think in those lines before embarking on it. If they are weak, we are accelerating weak, whatever we have, we are accelerating that by just throwing in ai. The second piece is start with decisions, not tools. Everybody's thinking, what tools should I buy? Everybody's evaluating tools. They're not seeing what kind of decision making do I need for my business and then trying to fit in a tool. So don't ask how do we use ai? Typically, I tell customers ask, which critical decisions do you want to improve, automate, or have forward looking thoughts on? And finally don't treat. If you're treating AI as a feature, you get incremental gains, absolutely you'll get some gains, but they are incremental gains. If you treat it as an architectural shift, you'll get competitive advantage. I'll tell my fp and F friends that the future of FP&A isn't autonomous finance. It's finance leader who can combine machine intelligence with human judgement will survive everything else. I think the days of reporting on spreadsheets is gone soon. So again, those are some of my final thoughts. Glenn and Paul,
Co-Host: Glenn Snyder (43:47):
I think it's great, and I think back to again the parallels, and Paul and I have said this many times about FT and a solutions that are out there. You don't build the solution for what you need today. If you do, you're going to build the wrong solution. You build it for where you're trying to go in the next five, seven years, and you've got to be thinking about that bigger picture because it could just go over and say, I just need this to go faster. Yeah, you're going to solve the problem, but then you're going to end up spending more money down the road because you didn't do it in the right way and you didn't build it out in the right way. Absolutely love what you're saying. Paul, what do you think?
Host: Paul Barnhurst (44:21):
I mean, I think this is a great conversation. To me, what really stood out is just that importance of starting with the strategy, starting with the why, and be careful to avoid tool sprawl in the sense of focusing on features. We've seen that with SaaS and we're already seeing it with ai, and often what it means is we're spending more than we should. We have many tools that we're only using a part of, we're not getting full capability. And with ai, you run the risk of also creating more silos of data similar to SaaS. So I think there's a lot of lessons we can take from that as we think about AI and really approaching this from an enterprise point. So those are a few of my thoughts, and I really enjoyed hearing your thoughts, Vikram, it was great to hear you talk about it and share what you're seeing because everybody's trying to figure it out. This is not new territory in the sense that we've all had to deal with technology, but it's new in the sense of what AI can do compared to prior technology. So it's a little scary and a little exciting for most of us.
Guest: Vikram Bhandari (45:31):
No, absolutely. And again, as you said, this is not new territory, but I think this is a totally different territory, and when I say not new but different is the impact of this is unknown and can be huge compared to the lot of technological evolutions you have seen in the last decade or two decades. So when people talk about human workforce and AI workforce, that thought process itself is new, that there could be AI workforce, which is new, and so it's going in a different direction. It's not only a technological evolution, it's an evolution which companies will go through, which will help efficiency, but will actually help them run a business more efficiently, more predictable, more profitable, more scalable. All of those, whatever objective we want to use, we can keep adding to it, but that's where AI is going to take all of us.
Co-Host: Glenn Snyder (46:36):
And the thing is just to add to that, because it is going to make such a big impact, but it's an unknown impact. We don't know. As you said, it is just that much more important to take a strategic approach and that because you need to be thinking about all the different nuances, all the ways things could come together and how they should be connecting to make sure you're going to have that right solution. Because I mean, I think we've all seen companies where they go over and they try and solve something. They think they have a great idea, but they weren't thinking about how everything fit together and what the strategic impact would be, and those companies usually aren't around for much longer.
Guest: Vikram Bhandari (47:13):
That's true.
Co-Host: Glenn Snyder (47:14):
It's the same thought and skillset process that we've gone through, but at a different level, which just again requires that strategic thinking. It just makes it that much more important. And so I would highly recommend anybody who's even thinking about going down the path of AI or any kind of technological solution, take a step back, make sure you're mapping things out in the right way. You're bringing all the people in. It's not just a finance solution, it's a whole company solution to make sure everybody's kind of weighing in before you move forward.
Guest: Vikram Bhandari (47:45):
Absolutely, and it is existential. People need to recognise that not adopting AI is not a choice.
Host: Paul Barnhurst (47:52):
It's something we all need to figure out. The sooner you figure it out, the better, because if it's done right, it can be a competitive advantage. If it's done poorly, you can be at a detriment to others who have done it well. Well, thank you, Vikram. We loved having you on the show. Great conversation, Glenn. Thanks for inviting 'em, and we're really excited to share this with our audience. I think they'll get a lot out of it, and it's definitely an exciting time. So thanks again for joining us, Vikram.
Guest: Vikram Bhandari (48:18):
Thanks a lot, Paul. Really enjoyed the conversation with you and Glenn and exchanging insights on AI and where it's going.
Co-Host: Glenn Snyder (48:26):
Yep, Vikram, thanks. Always great talking with you Paul. Always love being here. Looking forward to next month.
Host: Paul Barnhurst (48:33):
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.