I Would Never Let AI Touch a Number and what to do about that with Alok Ajmera
In this episode of FP&A Unlocked, Paul Barnhurst sits down with Alok Ajmera, CEO of Prophix, to explore how AI is reshaping FP&A, financial planning, and enterprise decision-making. Alok explains how modern finance teams are evolving from spreadsheet-heavy workflows to AI-enabled systems, while still relying on structured data, deterministic processes, and human oversight to maintain accuracy and trust in financial outputs.
Alok Ajmera is the CEO of Prophix, a leading financial performance management platform helping CFO organizations modernize planning, budgeting, and forecasting. With over 22 years at Prophix, he has supported more than 3,000 finance teams globally in transforming how they close, consolidate, and plan. Alok has led the evolution of FP&A systems from spreadsheet-first to system-first, and now toward AI-enabled finance operations.
Expect to Learn:
What makes FP&A a true force multiplier inside organizations
How data stewardship, collaboration, and forecasting define great finance teams
Why AI must stay deterministic when handling financial calculations
The difference between probabilistic AI and structured finance systems
How AI agents support FP&A without replacing financial control
Here are a few relevant quotes from the episode:
"The biggest opportunity is to take inefficient manual workflows and simplify them with AI agents, delivering value quickly while building confidence in the team." – Alok Ajmera
"AI democratizes capabilities, but there must be governance and discipline. It’s easy to build things, but maintaining them is hard." – Alok Ajmera
Alok explains that while AI is transforming FP&A through speed and automation, the core of finance still depends on structured systems, clean data, and human judgment. He emphasizes that the future of finance is a hybrid model where AI enhances productivity while finance leaders maintain control, context, and strategic oversight.
Follow Alok:
Website - www.prophix.com
LinkedIn - https://www.linkedin.com/in/alok-ajmera-93896617/
YouTube - https://www.youtube.com/@prophix
Earn Your CPE Credit For CPE credit, please go to earmarkcpe.com, listen to the episode, download the app, answer a few questions, and earn your CPE certification. To earn education credits for the FPAC Certificate, take the quiz on earmark and contact Paul Barnhurst for further details.
In Today’s Episode
[00:00] – Trailer
[04:10] – What great FP&A looks like
[09:13] – Revenue and qualitative goals
[18:38] – Why AI in FP&A needs structure
[29:36] – AI in forecasting and planning
[33:45] – Risks, governance, and AI sprawl
[45:40] – Soft skills in FP&A
[49:11] – Travel and personal reflections
[51:25] – Closing thoughts
Full Show Transcript:
Host: Paul Barnhurst (00:00):
Welcome to another episode of FP&A Unlocked where finance meets strategy. I'm your host, Paul Barnhurst , FP&A guy. Each week we 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. World. Together we'll uncover the strategies and experiences that separate good FP&A professionals from great ones helping you elevate your career and drive strategic impact. Today's guest is someone who's earned the coveted seat at the table as the CEO of Prophix. I'm thrilled to welcome on the show today, Alok Ajmera, welcome to the show. Thanks,
Guest: Alok Ajmera (00:40):
Paul. Excited to be here.
Host: Paul Barnhurst (00:41):
Yeah, really excited to have you. And so the audience know, it's kind of funny how this one came together. You had shared a post on LinkedIn about AI and the whole idea of A CFO, not wanting to let AI touch any of the numbers, right? That fear of it's going to hallucinate, it's going to mess up my numbers. And I responded to the post, you said something back and sent you a DM and said, Hey, would you be interested in talking about the conversation? And you said yes. So not the typical way I get my guests, but I'm really excited to have you. So thanks for responding. We'll start there. Yeah,
Guest: Alok Ajmera (01:11):
No, I appreciate it. Well, first of all, I appreciate the thoughtful comments you made on the post. I followed your work over the last couple of years. And yeah, excited to talk about my favourite topic, which is actually FP&A.
Host: Paul Barnhurst (01:23):
Well, Alok Ajmera is the CEO of Prophix, the financial performance management platform purpose-built for the Office of the CFO. With more than 22 years at Prophix, he has helped over 3,000 CFO teams across North America, Europe, and beyond modernise how they close, consolidate, and plan. Backed by global software investor Hg Capital, Prophix serves finance leaders across industries ranging from financial services to manufacturing, bringing rigour and speed to the full FP&A lifecycle. Alok is one of the few SaaS CEOs who has watched the Office of the CFO evolve from spreadsheet-first to system-first, and now to AI-first. He believes the finance teams that thrive won't just adopt smarter software; they'll learn to work alongside intelligent agents that transform what FP&A can do. They'll learn to work alongside intelligent agents to transform what FP&A a can do. So I like to start every episode with this question and I'm excited to hear your answer. Not only being a CEO, but a CEO of an FP&And a software. What does great FP&And a look like to you? What does that mean? What do you expect from your FP&And a department?
Guest: Alok Ajmera (02:36):
Yeah, great question. Look, at the end of the day, FP&And a done really well, I think is a force multiplier in the business. And I think there's three things that really in my mind really demonstrate or articulate a world-class FP&A function. The first really is data stewards. I think financial operational data that's existing in the business, really bringing it into one spot and making sure everyone is looking at the same information, making sure there's governance, there's controls, et cetera. So I think that's the base foundation. It really resides in good, crisp, clean data, and they're making sure everyone is looking off of that information. The second really is really deep collaboration with the business stakeholders across the organisation. Oftentimes FP&And a is just providing reports and analysis, but I think really world-class FP&A is when you're partnered with the business, you're collaborating in a bi-directional way.
(03:27):
You're supporting with data, you're supporting analysis, you're involved in decision making, and that's key. And then the third one really is collaborating with the broader C-suite and board around the forecasting and planning. It's really understanding where the business has been, but most importantly, where is it trending and what's actionable and not actionable. So in those kind of three things, if you can master those capabilities, I think it puts you in the operational on of what FP&A professionals are doing. And again, going back to the force multiplier, like businesses that have world-class FP&A tools perform and outperform their peers. It's such an important function.
Host: Paul Barnhurst (04:03):
Yeah, the way I like to say when you talk about tools, the one I say is when you have the right people, the right process, and then you bring in the right technology, that's really when magic happens, it's an enabler. You need the right people, you need the right processes, and you can do great things with subpar technology. I've seen it, I've lived it unfortunately more than once. But you could do so much more when you're not constantly fighting against the data and fighting against the technology and the tools and you actually have time for the important stuff is what I found.
Guest: Alok Ajmera (04:39):
Yeah, Paul, just to add to that, you and I were talking about this previously. Accounting and FP&A are slightly nuanced. One is rooted in a science. It's like this process or structure. There's rules. Fp a is a little more art. There's creativity that's kind of enabled in terms of what you can do. And I think the challenge sometimes is we take accounting functions and accounting skill sets and accounting people and we naturally move them into FP&A and we don't unlock the open-ended curiosity and creativity that actually makes FP&A, I think really special and ultimately leads to better FP&A organisations.
Host: Paul Barnhurst (05:17):
I love that you said that. I had a boss that goes, he had the opinion. He is like, I'm not sure accountants, most accountants make good FP&A professionals. And that was part of his reasoning. And I always give people reaching out to accounting, I want to move into FP&And a and one of the first questions, if I've ever talked to him, I'll be like, are you sure? Recognise, here's the difference in the roles. Are those the things you like? Is that what you're wanting to booing? And I still remember I had a great accounting partner and I was looking for an analyst role and I went to him, I said, are you interested? I think you'd be great in this role. He's like, no, FP&A is just a bunch of making up numbers. You guys just do funny math. I like it all tying out, I'm good where I'm at. And I said, that's fine. You're great. I love having you as a partner there. And so it was that reminder of you got to know your personality. It's not one's better than the other or one, they're different. And there are benefits to both roles
Guest: Alok Ajmera (06:04):
A hundred percent.
Host: Paul Barnhurst (06:05):
But I loved his answer where he is, you're just all funny map. I'm like, Hey, I take offence to that. I want to ask question. On your LinkedIn profile, you talk about how you're building profits into a dominant global office of the CFO platform and something you mentioned several times on there is A-B-H-A-G, right? The big hairy audacious goal. So how do you measure something like that? How do you know when you're the dominant global office of the CFO platform? Have to ask because I love seeing that on your platform. I can tell you dream big and have big gold.
Guest: Alok Ajmera (06:36):
I came across the concept of A-B-H-A-G maybe a decade or so ago, and the concept really resonated with me. And I'll get to your question in a moment, but I'll take a step back and talk about BHAGs for a second. One of the challenges, I'll speak in the context of software companies, but I think it applies to everyone. When you're early in the lifecycle of a company, maybe you're in startup mode, et cetera, there's a unifying purpose. And that oftentimes is not necessarily the mission that you're on, it's survival. So at the beginning, everyone in the organisation is unified to the single goal of making it. We just got to survive and prove that we have a viable business. And as you continue to be successful and as you grow, and then at some point you get very large, it's like that connection for everyone.
(07:22):
The purpose oftentimes can get dissipated. Now if your business is mission driven where you're like, we are saving babies, it's like that is unifying and that's purpose built, right? In my world, it's like we're building software for corporate finance groups now I love it and our customers love it. But if you're the average person who's an engineer who's not connected to FP&A, it's like there's no mission in that directly. And you've lost the mission of, Hey, we're trying to survive and prove that we can do it. And so I found when teams are really motivated, they're focused exclusively, and they really, and you get the most out of them and they enjoy the work that they're doing, purpose is at its heart. And so I stumbled upon this concept of A BAG, and I started playing with it a decade or so ago and I found it was a rallying call.
(08:15):
It's like, Hey, we're trying to do something really hard and here's exactly what that is. It's very uncomfortable when I say it out loud. You probably, when I roll out these BHAs on every couple of years, I update them. My team usually has a visceral like, Ugh, here we go again. How are we possibly going to do that? And that becomes unifying. And then when you can figure out a way and you hustle and you iterate, and then one day you cross the line and you can go back and say, Hey, remember that thing that we said we were going to do and it was really difficult and daunting, but we did it. You get the check mark and the team feels amazing. And so I've just found these are so good. They're aspirational goals. So now let me get back to your question for me, B hs have no timelines.
(09:00):
It's not like I want to get to this place by this time. It's like, Hey, this is what we're trying to do. We don't know how we're going to do it. It's a little bit uncomfortable. Like I said, it's a little no clear path to get there, but it's a directional guidepost. So we as an organisation can then continuously iterate to figure out how we work closer and closer at Prophix. I've always had two BHAGs, like a quantitative and a qualitative. The quantitative one is the easier one because it's like it's a number and you can measure it. Half of my team needs a number that can be measured. The other half is like, Hey, who cares? It's just a number. And so the other BHAG is always qualitative. And the qualitative one is the more mushier one where it's like, well, what does it exactly mean?
(09:44):
And everyone has a slightly nuanced version. And because it's an aspirational goal, I think it's okay. And so for profits, it's always been, there's one revenue number. It's like when we had zero revenue, it's like, Hey, we're going to get to 5 million in revenue or 10. And once we crossed off 10, we were like, let's get to 20, let's get to 50, let's get to a hundred, let's get to 500. And each time that we crossed one off, I've been doing it for like I said, a decade or so, and it's like you feel this visceral, the team that's been on that journey feels like, I can't believe we did it. And then I put the next one up and it's like, oh man, we got to do it again.
(10:23):
And the qualitative one is what you alluded to, the global offices CFO platform, the qualitative one for profits is we want to build software that fundamentally changes how our customers work in a positive way. So if you think about FP&A or even broader CFO, a lot of the processes and the day-to-day work that happens is rooted in technology from 20 or 30 years ago. And I look at it and I'm a technologist more than I am a practitioner of FP&A. And it's like these processes don't make much sense. There's a huge opportunity, especially now with AI and I'm sure we'll get to that to really change how we do these things for the better. With the goal of, if you think about CEO, sorry, CFO, it's like actually what they're there to do is optimise enterprise value. So if you start with that and then you work backwards, it's like, okay, well where can technology update these processes with that as the objective? How do we drive more enterprise value? Sorry, I've gone on a long roundabout. No, I
Host: Paul Barnhurst (11:27):
Appreciate that was a great answer. I liked it. I was definitely listening close and even thinking in some areas, what would mine be for my business? I have goals. I don't necessarily call it BHA, but it's a little different when it's one person versus company. But same idea, you need that unifying. What's the key to getting a team to achieve those? You've obviously hit the revenue. I think I had a goal of a hundred million at one point. You're now past that. You're heading, I think 500 was the goal I saw on LinkedIn that you said, but how do you keep the team focused and unified? Obviously as the leader, you need to help keep 'em focused on the goal. And there's all kinds of distractions. We all deal with them. So how do you make sure people remain focused on that kind of big goal when there's so much in between that has to be done?
Guest: Alok Ajmera (12:13):
Yeah, that's a great question. Let me answer in two ways. One, look, I'm not an FP&A practitioner, but I love FP&A. I spend most of my time with my FP&A team. So it's like, okay, how do we have a big goal that we don't have from a timeline perspective, but then how do we start backing into, okay, how do you take a B HG and break it down into five or six individual measurements and then as a long-term success measures, and then how do you take those success measures and say, okay, well how do you roll that into a three to five year plan? And obviously year 3, 4, 5 are kind of really high level tops down, but year one and two are very granular bottoms up. And then again, how do you reassess the success measures, the KPIs that you're going to measure along the journey so you can keep everything focused and relevant, and then how do you cascade those down through the organisation?
(13:01):
So that's step one in my mind, it's like how do you take a big BHG, start working backwards from that b hg and be like, okay, for that to be true, these six things have to be true for these six things to be true. These are the things that we need to actually be working on. And you go all the way down to what has to happen next month. So that's step one. This step two, and this is the piece that I find organisations miss sometimes is you have to be constantly communicating with the team. And so I do every Monday, I do a town hall every Monday, and then quarterly we do really big town halls in a cascading management week. And I find on the day before the quarterly town hall, the team is at its most disorganised. Everyone has splintered off and going in different directions. And the moment the town hall ends, that is the most aligned. The entire team is exactly to your point. Everyone gets like, right, right, this is exactly what I need to do. I'm charged ready to go, let's go. And then three months later, it's like all these little things come in, this customer problem or this thing or this deal we're trying to close. And then you kind of get frayed and it's like you got to bring everyone back up to 30,000 feet. Get them aligned, get them focused, and then let them go. Again.
Host: Paul Barnhurst (14:17):
You
Guest: Alok Ajmera (14:17):
Make
Host: Paul Barnhurst (14:17):
It sound also simple there with that answer. I love it. But I heard kind two things there. Like I said, the planning, you got to take it from the top and really figure how do you filter that down so people have something manageable they can focus on that helps 'em achieve the goal that you can measure. And then two, I'm going to simplify it. Communicate, communicate, communicate. You got to keep it front of mind.
Guest: Alok Ajmera (14:40):
A hundred percent. People are busy. Think about an FP&A team times, how often does an FBA team Friday afternoon being like, I got nothing to do. Let me go back to strategy and reflect. It's like, no, we're slammed.
Host: Paul Barnhurst (14:55):
Yeah, I'm going to go read the mission and vision again. It's been a few weeks. Yeah, no, I don't think I've ever heard somebody say it that way. So appreciate you sharing that. So I want to transition to a little bit of an AI discussion like we talked about. So as I mentioned before at the beginning, you and I first interacted on a post, and I'm sure you've heard this story before I've heard it, is how A CFO would not let AI touch the budget or any numbers for that fact, but was using it in other areas. Maybe can you set the stage by telling a little bit about that story, your response, set it up and we'll go from there?
Guest: Alok Ajmera (15:28):
Yeah, for sure. Listen, I was on a call with a Prophix customer, I guess this would've been a month or so ago now, and we're just chatting and catching up and she's a pretty active FP&A team, and they're pretty active. They're on the spectrum of good to grade, they're closer to the great side of an FP&A team interviews. We were chatting about AI and they were avid users of ai, both in profits and all the capabilities that we have, but also just in general, they're deeply committed to Claude and cowork and so more on the progressive side. So we were chatting about what do you use it for and how do you get value? And so right away it was like you love using AI for reporting, for analysis, for creating narratives, for synthesising content, for creating board update emails, and so anything analytic or analysis centric, she was like, oh my God, I love it.
(16:20):
My team is getting so much value from the capabilities. And I was like, oh, that's really great. So I'm making notes and I'm like, great. Profix had launched a budgeting agent in September of last year, and the budgeting agent I kind of love because I find business users don't like to actual budget. And when I dig into why, it's because they don't want to go into a spreadsheet and they don't want to go update numbers. And so oftentimes what happens to a financial analyst, this is probably your audience will resonate. It's like there's like five laggard executives and you have to be like, fine, let's just get a meeting, get in a room together and just tell me what you want to do and I'll do it for you. Right? So great. We're like, let's solve for that. We created an agent where you can just articulate all the changes you want to make up these numbers, down these numbers, add this headcount, give this merit increase, and then the agent can actually do all the work on your behalf.
(17:10):
So I brought this up, I'm like, Hey, you're using all this amazing reporting analysis capabilities. Why are you not using the budgeting capabilities? And the entire tone of the conversation changed. It was like, oh, no, no, I would never let AI touch a number. And for me, it was actually an interesting moment and I've been now going away and talking to as many people as I can to pressure test this, and there seems to be a divide. Don't get me wrong. There's super progressive CFOs that are currently a small minority. I think of AI for everything, but there's a very big chunk of the CFO and FP A teams which are very apprehensive of ai, whether it's a budget or forecast or even on the close side, it's apprehensive around actually touching or modifying numbers. And it makes sense if you think about where AI today is today, it totally makes sense. I was just surprised at first.
Host: Paul Barnhurst (18:11):
Yeah, no, I am not surprised that you got that response. I am surprised at how many and how big it is sometimes. I think if you look back to the early days of ai, first time you tried it, you asked it what two plus two was and they told you five, and you're just like, okay, no numbers again. Or the examples everybody would give is, I asked that Sally was four and Bob was eight. What were they 30 years later? The answer was just like what? And so I think a lot of people have a fear and obviously hallucinations, but as you and I talked and as I talked to a lot of people, what I always try to say is, and these tools aren't just using generative ai, I think people sometimes have the idea that, okay, it's all being done by generative ai.
(19:00):
And I've talked to a lot of software vendors and almost every one of 'em says, no, the numbers we're using a deterministic process here. If you give me the same inputs, you're going to get the same output every time. It's no different than I think, I mean different in little ways, but very similar to almost every FP a tool that has an algorithm that will do a statistical model for you. There's a lot of that behind the scenes. You're just now using ai and so that different term, a little bit of different technology and automation creates a real fear. What would you say to that? That's how I think about it. Is that a fair assessment here of how to think about a lot of this technology with agents? Yeah,
Guest: Alok Ajmera (19:38):
For sure. Look, AI as a stands today is a probabilistic engine. And I think part of it is not everyone really understands how, look, actually I have to take a step back. No one really understands how AI works. They're like, this is really great. If you think about ai, it's really simple as core it'ss like grade 12 maths, multi-variable regression analysis at monstrous scale, it's matrix math. That's really what it is. And not many people, even the researchers that are building these capabilities even understand why this math produces the output and outcomes that it produces, but it's good. And so we figured this out. Now you take that to the non-technical finance person and there's just a lot of natural fear. One, it's probabilistic. So everyone has had a situation where they've gone to chat GPT or code asked the same question three times and got three slightly nuanced answers.
(20:33):
And it's like, well, that's a problem. Fundamentally, a probabilistic model don't work in finance. Second, everyone is afraid of losing information, putting information out into, and all of a sudden one of these frontier labs is going to be training off of your data. And so there's a opaqueness, they don't understand how to put guardrails and to protect data. And so there's fear that if I start putting in data, it's going to go away and all of a sudden chat, GPT will know all of my financial information, which to be fair is a legitimate concern. There's things you can do to protect yourself, but it is a concern. So there's just this overarching fear, but your answer is a hundred percent correct. And this is what I try to get across to people as well is we're not asking a generative AI predictive model to do anything with your numbers.
(21:24):
What we're actually doing is saying we are decomposing our technology into hundreds of thousands of little tools and skills that AI agents can use. And so the AI agent is not actually doing the math, it's just passing the right parameters to the tools that we have so that we can do the math and return the number. So the math is a hundred percent deterministic. The piece that I think software vendors are missing still is the process has to be deterministic as well. And so if you think about how agents work today, it's first you query data, then you reason, and then you execute. And what happens actually is, and this brings up a whole nother conversation we can have is it's not efficiently done. So the querying is wildly inefficient. An agent will just bombard every access point it has with queries and it'll collect as much information as it can.
(22:23):
And oftentimes it's orders of magnitude too much, and then it tries to apply reasoning skills on top of that to figure out, okay, what do I do with all of this? And then how do I reason out the steps? Now if you go to an FP&A person and you're like, Hey, I need to do a four plus eight forecast, it's like it can't be a different process than when you did the three plus nine. It actually actually follows some basic methodology. So even though process has to be deterministic, so if you think about going back to what the agent does, it bombards every system, every access point, collect as much information, then it tries to make sense of it all and then come up with the steps that he needs to do and then it executes. And so for what I think software vendors have to do, especially in FP&A, you absolutely have to make the calculation engine deterministic. You can't have it guessing in making mistakes. You have to create more context to have deterministic workflow processes and then you can let it go off and execute. That
Host: Paul Barnhurst (23:17):
Makes a lot of sense. And one thing I've seen that, I don't know if you've seen it copilot, they just released this week in Excel is they have a mode they call plan mode where you tell it what you want and you build all that plan stuff and you agree on everything before it goes off. And the more that happens, and then you have deterministic tools as well, I think the better you're going to be. Because especially in Excel, if you're just in a spreadsheet, it doesn't have context unless you've given it, maybe reviewed your spreadsheet or whatever. But often when you're building Excel, there's no context. So having a plan to start with is a lot better than, Hey, build me this model and then you're done and you're like, that's not what I wanted. Or that has 37 mistakes, you're going to get a lot better output.
(23:56):
Kind of like you're talking about those workflows. And so I agree. So what would you say to the finance community out there? Where should they be sceptical? When should they be using it? How should they be thinking about this? Just all moving so fast? I think everybody's at a different point in their journey. Everybody's struggling with something, whether it be, hell no, it's not touching my numbers or sure, I'll let it do everything, but if it screws up once I'm out to, I'm not using AI at all. I think we have a little bit of everything
Guest: Alok Ajmera (24:27):
For sure. I was just telling you, I got back from a big Prophix conference. We met hundreds of CFOs and every conversation was about AI. It's on everyone's mind. Of course. And look, the people are at different places in the journey. There's definitely people that are like, we have a corporate mandate not to use ai. There are people that are on the one end that don't use ai. And some people came to me and I was kind of surprised. People were like, we have an ethical concern around ai. It's like, okay, walk me through your ethical concern. It's like, well, we are really concerned about our stewardship for the next generation of employees. It's like, oh, okay, that's interesting. People have come to me with environmental concerns around using air. They're like, we're really concerned about the energy consumption, the land footprint, et cetera. And it's like, okay, look, I'm not in a place to tell you what your moral compass should be, right?
(25:20):
First of all, I just want to acknowledge, right, for all whoever's listening here, wherever you are on the spectrum, it's totally okay. There's just too much rhetoric around AI, too much hype around AI and a lot of it's not real. And so it's okay, wherever you are on your journey, just understand where you are and then start mapping out where the next couple steps are for you. Oftentimes when you're on the early end of the spectrum where you're like, I don't know if I should use it or I don't want to use it, some of it is just take some small steps. Don't try to go and be like, I have found this a tonne of agent that we can start building or using. It's like, don't go up to the extreme. It's difficult. There's work that has to get done to do that properly.
(26:02):
Small steps could be like, we're going to start, we're going to add a corporate licence of one of the frontier labs so we can have some protection and we can start playing around with it for analysis. If small steps could be like, Hey, there's really specific workflows that are very inefficient. Our CFO, he's interesting, right? He's a CFO, but he also works in a company that builds software for cfo. So he's on our product advisory board. He's a marketing guy as well. It's kind of a unique spot for him. But anyways, he and I were chatting buddy. He's like, ah, man, I need an AP agent that just helps me do accruals. At the end of the month, you're looking at your expenses and then it's like, oh, shoot, this vendor didn't send me the invoice and time and all of a sudden there's a variance.
(26:44):
And it's like, but we know all the vendors, we have all their contracts. We just need to reconcile it with which ones we got and make the accrual adjustment. And it's like, well, that's a very small problem, but it's easy to solve. And it's like, okay, so there's so many things that your organisations are doing that are manual, that are slow, that are extraordinarily repetitive. It's like just taking some of those and chewing through them. See if you can experiment, get a tiny bit of value, just take little steps and then before it, you'll see some value, you'll get more comfortable, you can build some confidence and capabilities in the team, and then you can start taking bigger steps. Sorry, Paul, I went again on a tangent. I dunno if that answered the original question. No,
Host: Paul Barnhurst (27:26):
You're fine. I think that definitely helped a little bit. We're all at different points and I think that answers some of it. I like the example of the CFO, so the CFO's, everybody's at a different journey. Where do you think AI makes the most sense right now? Maybe step back a little bit there. So assuming everybody's comfortable with using it, which obviously we know they're not, where do you think it makes the most sense for finance right now? We hear the hype of using it everywhere and it can change my life and it does everything and there are amazing things it does, but we all know reality versus marketing hype are usually never at the same point. And so I'd love your thoughts of where do you think the best use cases are? Where does it make sense and where would you tell people? I'd have some real pause here, proceed with caution or don't use it. I'd love to get a little bit of your thoughts.
Guest: Alok Ajmera (28:19):
It's a tough question. I'm hesitating for a reason. The technology is moving so quickly and my thoughts from last week have changed from this week and they'll change again.
Host: Paul Barnhurst (28:28):
I recognise that it's a great caveat. We ought to put it out there by the time we release it, the answer could change and it is so hard to talk about this.
Guest: Alok Ajmera (28:36):
Yeah, yeah, totally fair. Look, there's definitely value on the reporting and analysis side. If you think about generative AI, generative AI is a content engine. Anything that's content related, it's a natural use case. So if you think about the biggest use cases that are being monetized today, it's agentic code writing, it's agentic content creation. It's in the legal space, which is a lot of content. How do you understand the terms and case history, et cetera. And so that's a natural use case. So anywhere in your workflow that you're creating or analysing content, it's a hundred percent ready to go. And by the way, that fits with a lot of what CFOs are doing today, right? It's like, oh, I want to build some reports. I want to do analysis, explain what's happening. I want to build a board pack. I want to build a variant and explain.
(29:21):
It's like those are all slam dunk, take it to the bank use cases right now and they're getting better and better. If I then go from that point and start breaking away from that, your audience is more FP&A, so I'll focus a little more FP&A as opposed to the controller close slide. It's really modelling. Now the next major kind of hurdle is like how do you do better forecasting? How do you increase the velocity of forecasting? How do you get more people collaborating on forecasts? AI is an enabler for all of this. So if you can use capability, assuming you have really good data as a starting point, you have some decent technology. Like you said, it's like if you think about forecasting as well, there's two schools of thought on forecasting. I think I'm curious to get your perspective on this, Paul, there used to be a really strong voice around machine learning.
(30:06):
It's like, Hey, how do you do more predictive forecasting, both operational but also financial and that's kind of died down a little bit. I don't think, especially for mid-market organisations, I don't think machine learning based time series forecasting is there yet. There's just too much of a variance in terms of how accurate it is. Oftentimes with time series forecasting, 10, 15, 20% variation is okay, but it's like I've just never met a CFO's like I'm totally good with 20% variance on my forecast. It's like that's just way too wide, right? The cool thing I think about generative AI is not about forecasting in terms of, Hey, let's produce models and algorithms that are going to spit out time series numbers. How do we facilitate broad scale collaboration and make it really, really easy? And I think ag agentic AI actually is really good for that. So I think that's another area that's emerging on the FP&A side that I think is going to be great.
(31:04):
Here's the analogy: right now everyone does an annual budget. It's really time consuming. You got to get all these stakeholders, you got to wrangle them in and they got to do a plan. And then you come out with a plan whatever time it takes, months potentially by the time you get there to the end, you finish Q1, you're like, this is all out of date. Now you got to do a refore cast. And it's like, well, I'm not doing that again. I'm not going back to a hundred people and wrangling them. They don't even want to be part of it. So then we go to these tops down models and we're like, well, the CRO told me we're probably going to be around here, so let's go tops down and drive this out. And what I've been challenging people recently is like, well, what if we could just take that entire process of wrangling 200 people in a budget process and make it so easy that it could be done in a couple of days? Would you then do a monthly forecast with everyone actually contributing and to getting really focused and really clear accuracy because people still have a lot of institutional knowledge that's really important in these processes, but because it's so taxing, we don't bother including everyone, and I think generative AI is perfect for that. Yeah,
Host: Paul Barnhurst (32:06):
No, I appreciate that answer. So on the flip side, we talked about some of the real risks. Obviously we have hallucinations, there's data risk, but what are the watch outs and the, Hey, pause and slow down a little bit. Maybe talk about that side for just a minute.
Guest: Alok Ajmera (32:21):
Look, one of the cool things about AI is that it's democratising all knowledge expertise. So you probably have people that are on the forefront or your audience members that are on the forefront of AI and they're off and they're like, oh my God, I'm using replica to build tools. I'm writing code with Claude Code and I'm building my own stuff, and it's like, oh, I'm integrating it to Excel and I'm doing this integration. I'm doing that integration. And I think it's great by the way, and I'm a huge supporter. I love Tools Rep. Obviously we're a massive Claude Code organisation, so I really think it's great. The challenge is you could really quickly build a whole bunch of stuff that you don't understand and that has no governance or control, and now all of a sudden all the problems that we've been trying to avoid over the past where it's like we're looking at different sets of data or the integrations are the pipelines are not all clean and it's like all of a sudden we can take problems and make 'em like we can explode them.
(33:18):
Everyone's running around with their own really complicated Excel model that they built with Code Cowork and it's like, oh man, this was one of the problems we were trying to harmonise. Everyone's one pan of glass, I'll built all these applications using Claude Code or Rept or something like that, and it's like, actually, I don't know how to maintain any of this stuff and it's getting messy and it's getting complicated. So I think the exciting part is we're democratising capabilities. We're democratising technology and we're putting it into everyone's fingertips, which is really good. The word of caution is like there is structure and discipline that needs to be put in place to govern these things properly, to control them properly, to make sure that you're going to maximise your benefit, not just immediately but down the road. And it kind of touches on another thematic that we were talking about, but building things is getting easier, but maintaining things is actually quite hard, and oftentimes we're allowing people to build things who have never had to build things in the past and don't necessarily understand the maintenance part of the equation yet.
Host: Paul Barnhurst (34:19):
It's like you talked about a little bit the whole idea of AI tech debt in finance on LinkedIn, and I've talked about this a little bit from a different angle, but I think a lot we're heading into what I call AI sprawl. We all went through SaaS sprawl where all of a sudden, how did we go from two tools to 500 and from 30,000 a month to 500,000 or whatever, you're going to see the same thing with tools for ai, but now you're going to multiply that by the fact of not only are people adding a tonne of AI tools, they can build their own tools and they've never, like I said, they've never maintained it. They've never had to match the digital security, the governance privacy rules and what are going to be the ramifications of that in a couple of years. There's probably going to be a lot of debt and there are going to be some companies that's going to be a cluster, for lack of a better word.
Guest: Alok Ajmera (35:10):
Look, it's not just finance. This is everywhere. And just to make it, to take one step further, this is a problem in engineering right now. My engineers can often write up to 20 times more lines of code than they used to be, so just put better context for the effort. It used to take them to write a hundred lines of code. They can now write 2000 lines of code, so that's wonderful. We're writing way more lines of code, we're creating more applications, we're building new things, but what's happening is the velocity of build is being outpaced by the velocity and our capability of understanding and maintaining, but engineering already has that discipline and functions. We have dedicated QA processes, we have dedicated people that's job is to maintain. We have documentation people that all they do is write up exactly what they've been doing and why and how.
(36:02):
Now you go outside of engineering and you go to a finance function and you're like, Hey, you can start building your own applications, but the muscle memory and the discipline around, well, let's document all this. Let's have QA processes. Let's actually understand the governance of data. Let's think through infosecurity, all these points you made, the discipline isn't there because they've never had to have that discipline. And so I think this is the word of caution, right? It's really easy to build things. It's really easy to just try things out and go. I think we're really going to have to be thoughtful around what we put into production, how much rigour we put around it. I think there's going to be new roles that we're going to hire. I'm convinced that FP&A teams are going to end up having to hire engineers. You're going to need, in your finance ops team, you're going to have to go hire actual engineers because there's benefit to do this. You just have to have the discipline to know how to manage all this stuff.
Host: Paul Barnhurst (36:55):
I hadn't thought of engineers, but I've definitely thought of AI architects or a system person. So what you're saying makes sense. It doesn't surprise me. I just hadn't thought that far ahead before.
Guest: Alok Ajmera (37:05):
Everyone's like, oh, the skill you're going to need is prompt engineering. Like, oh, you're going to need to know how to prompt really well. And it's like, well, that's easier. That's getting easier and easier. You don't even have to know how to prompt it. You can even just ask Claude to create the prompt for you. You don't even have to know these things. It's just the danger if you lose the understanding of what is being done and why and how it works and that's the danger. I think the same thing by the way with Excel right now, you can create these really amazing models in Excel quickly, and we're making it more robust. You can create plans so you can understand what's being built, but I'm worried that actually what's going to happen is we're going to propagate more and more monstrous, cumbersome Excel models through the entire enterprise that are kind of disconnected.
(37:50):
They're working off a little bit of different data, which is some of the problems you have when you're building large scale, cross-functional, take it, they're going to sell. By the way, I'm a little biased, so you should understand that. The audience should understand that to be biased, but actually what I think is happening is you build these complicated models and it's easy to do and actually you're just going to throw them away and every time you need, you just build a new one. It's too much to maintain and remember, AI actually doesn't have memory. It doesn't train. Everyone has this idea that it's like, oh, it's learning and it's going to get better, and it's like, no, it doesn't actually, the only time really it trains is when a new version of the model comes out. So when you're on Opus four and it goes to Opus four six, that's when it trains. The only thing we are doing really is providing incremental context. Every time we prompt, we're not training it on anything. It's another tangent. We can go on one day, but
Host: Paul Barnhurst (38:45):
I've heard that before when we were talking on financial bias. Someone's like you ever, it doesn't really train. Yes, you give it. There's context. Yeah, so I've heard that before. That could be a whole episode by itself around that whole thing. But when you mentioned the modelling, I'll share one thing here and then we'll move into the FP&And a and get to know you section. I know we're coming up on time, we're running here, so I don't want to keep you too long, but the one that I talked to a guy that works in modelling and he's one of those trusted for the audit, the verification. When a big deal is happening, a lot of the banks and the big companies, they come to him to validate the model. He goes, I'm seeing a lot of stuff that concerns me. The way I have to validate is different now, different risk ai he even mentioned, which I thought was really interesting, is a lot of the banks are struggling to find good modellers earlier in their career. They think they can just use AI and everybody thinks it's going to take away the job. They interviewed someone else that said, right now the banks haven't been able to eliminate any jobs. They're getting pressure to, but they really can't. The AI is not good enough yet. So there's this fear of 50% of all jobs are going away, and then there's reality, and yes, there are places, jobs are going away, don't get me wrong, but I think you got to step back sometimes.
Guest: Alok Ajmera (39:58):
I a hundred percent agree with this, right? I don't think jobs are going away. There's going to be disruption. Some jobs are going to change in your jobs. Well, there was with
Host: Paul Barnhurst (40:04):
The internet, there was with the computer. That's just life.
Guest: Alok Ajmera (40:07):
But lemme give you a really good example. The number one use case for AI today is ag agentic code writing. It's where Philanthropics is making most of their money, and it's a great use case. Like I said, I can get 10, 20 times productivity lift on my engineers. I'm still hiring engineers right now. I can't hire them fast enough. I'm like, let's go. I need more engineers. And so that's a litmus test for every other function. It's like, yes, AI is making us more productive or can make us more productive, but as humans, I feel like we have an insatiable ambition, and so if we can actually produce more, we will just do more things. If we're that much more productive, it's like, great, let's expand the scope of the number of things they're going to do. So I fundamentally believe it's not at least AI in its current format in the underlying technology as today, I don't think is going to wipe out. We're not going to have unemployment. It's just sensational height building. A hundred percent of jobs are going to be destroyed. 50% of white collar jobs are going to be gone. It's like, yeah, I just don't see that world.
Host: Paul Barnhurst (41:11):
I'm with you. I don't see that world. The idea of we're all going to be at home with a universal basic income. Do I think there's going to be disruption? Are there jobs? Are there companies that may be able to take out a lot of jobs? Are there industries that are completely changed? Yes, it'll have massive disruption. I don't think it'll have massive job displacement, which I think are different things.
Guest: Alok Ajmera (41:30):
Your audience member, right? They're sitting in a corporate finance group and they're already understaffed. I can pull up all these stats. The corporate finance groups have been getting squeezed for years. They're already overrun. The world is more complicated. They're requested to do way more sophisticated things. There's just not enough headcount today. The idea that we're going to use AI and then wipe out more people from FP&And a, it's ludicrous to me.
Host: Paul Barnhurst (41:58):
Historically, numbers have been roughly 75% of FP&And a's time is on non-value add activities, and most you ask any company if they're getting as much as they would like out of FP&And a, the answer is almost always no. Have I asked you, would you like to get more out of your FP&And a team?
Guest: Alok Ajmera (42:16):
Yes, a hundred percent
Host: Paul Barnhurst (42:17):
Probably yes, right? They can always do more. They're doing a bad job. So okay, if I take out the now, I can get 75% more of what I want. Okay, yeah. There may be some companies that displace a person here and there. Yes, there'll be some of that, but the idea that 50% of FP&And a jobs are going away as ludicrous to me as well, I don't see those
Guest: Alok Ajmera (42:36):
Type
Host: Paul Barnhurst (42:36):
Of numbers.
Guest: Alok Ajmera (42:37):
Yeah, I agree.
Host: Paul Barnhurst (42:38):
Alright, well real quick, I have an fp, a section to get to know you, so we'll run through these. What do you think in today's environment and kind of going forward is the number one technical skill for FP&A professionals to master?
Guest: Alok Ajmera (42:51):
Yeah, look, I will answer it slightly nuanced. I thought about this a lot in kind of prepping for this. It's not necessarily a technical skill, but I think everyone really needs to up their AI literacy, and that's not just playing around with ai. I think it's really important to understand what is being built, how it's being built, why it's being built, what are the mechanics, what does training mean? What is, what is context? How are tokens used? I just think if you don't understand these things, then you're just kind of running around in the dark trying to follow or catch up, or you might trip yourself up because you fundamentally don't understand how this technology is what it's doing. So it's maybe not a technical skill, but I'll put it under that category. I think everyone would be,
Host: Paul Barnhurst (43:35):
I'll give it to you. I think it qualifies.
Guest: Alok Ajmera (43:37):
It's like if we can all up our just general AI literacy, listen to some podcasts, you can read about it and honestly, you could just ask AI to help you, which is pretty incredible.
Host: Paul Barnhurst (43:51):
I'll put a plugin. You can listen to my Future Finance podcast about ai. Glenn Hopper is an expert. I learn from 'em all the time. So one thing I was going to say, which interesting, your answer is similar. I've heard a lot, and these are more smaller companies. The number one scale is starting to become system thinking design data and data modelling. And those are answers from FP&A. People that are getting a lot out of ai. They realise how important the data and the system and the design becomes. And so I think that kind of goes into your AI literacy. I think that all goes into what's that skill becoming. I don't hear near as much as I used to, which was almost always the answer six months ago, and now it's a lot less. Excel and financial modelling still very important skills. I'm not going to say they're not important. You have to know them, but it's been very interesting to watch the shift on the show of that answer.
Guest: Alok Ajmera (44:46):
Yeah, I totally agree. And I like the shift.
Host: Paul Barnhurst (44:49):
So what about soft skill or human skill?
Guest: Alok Ajmera (44:52):
Look, I'll answer it in two different ways. One, I think as AI helps us kind of streamline a lot of that 70% non-value add. I think the real value add, and this is what I mentioned at the top when it comes to best in class FPA tools, is actually integrating with the rest of the business. And so this is around how do you speak the language of your business stakeholders? How do you understand and empathise for what they're doing? And then how do you support and collaborate? So building the bridges outside of finance is going to be critically important. So I don't know how you want to categorise that as a soft skill, but I think that's number one. I think if as you free up more of your 70% of mundane time, really good FP&And a professionals are actually outside of the finance group working in the operations.
(45:40):
They're working with their sales team, they marketing team, they're in the plant talking to the managers that are running the machines, et cetera, and they're collaborating. And you can't do that as a finance person come here to, you can't go to our CRO and say, Hey, look, I think you can, I ran some numbers and I think you can squeeze up the quotas because of the number. Here's my book. And it's like, they're not going to want to listen to you, but so you have to understand their world, what they're going through, and then you have to communicate your analysis to them in the way in which they're going to understand and then they're going to partner with you on, just as an example,
Host: Paul Barnhurst (46:13):
It's a bad idea to hear the CRO and say, we think you can raise all your quotas.
Guest: Alok Ajmera (46:17):
I ran a model and I think we can get a 18% price increase if we do it this way. And it's like, okay. But you have to empathise for them. They're the ones that are going to have to be on the call with every customer trying to squeeze out an 18% price increase. And then there's, you know what I mean? So that understanding that empathy. By the way, as an aside, ambitious FP&And a professionals should leave FP&And a for some period of their career and go work in the other areas. If you really want to go and become a CFO, it's like go work in sales operations for a year or two and understand what they're going through. Go work in marketing, go work in a plant, understand what's actually happening, make that context is invaluable. So there's just a side tip for people. If you're ambitious and you want to grow, get out of the function and then come back.
Host: Paul Barnhurst (47:07):
I have talked about that more than once of the importance of having some kind of operational, I worked in a business analyst role, I started my career in procurement and I've run my own business for four years and I can guarantee you if I had to go back to an FP&A role today, I would be better than I would've been five years ago before I started my own business. Because of the experiences and seeing things from a different lens, you get an appreciation you just don't get, or it's very hard to get if you get it. But very few really get that appreciation without those type of roles. I'm with you. That was something our American Express, CFO, we had Jeff, I can't think of Campbell I think was his name, but he mentioned one of the best things he ever did was go into operations for a year and a half before he came back out as a CFO. Alright, so we'll two get to know you questions. I'll let you go. I know we're running over on time. What's a book you'd recommend for our audience?
Guest: Alok Ajmera (47:54):
Yeah, I thought about this one too. I am going to go off on a tangent. I'm an avid reader, but I read fiction.
Host: Paul Barnhurst (48:00):
I'm actually a big fiction reader as well. I don't do a lot of, I do some, but I stardom and never finish 'em is my problem with the business books.
Guest: Alok Ajmera (48:08):
Yeah, I'm the same. And then honestly, I find I read for stress management and relaxation. So reading a business book just winds me up more.
Host: Paul Barnhurst (48:16):
Well, we're in the same, I think we have a similar thing there. So
Guest: Alok Ajmera (48:21):
I'm going to go fiction and I'm going to go, I'm reading right now, it's the kind of adventure fantasy book, but it's the Stormlight series by Brandon Sanderson.
Host: Paul Barnhurst (48:30):
Okay. Yeah. I've read all Brandon Bull stuff. I hadn't read Brandon Sanderson. He's from Utah. Brandon Sanderson.
Guest: Alok Ajmera (48:36):
Yeah. Yeah,
Host: Paul Barnhurst (48:37):
So I have a friend that actually is his chief of staff, runs everything for him, used to live up here and moved to become his chief for his business.
Guest: Alok Ajmera (48:46):
He's just a super interesting guy, both as an author but as an entrepreneur as well. And yeah, I think the first book in the series is called The Way of Kings.
Host: Paul Barnhurst (48:54):
Okay. Yeah, that's been on my list to read some of his stuff. He's one of the few I haven't, I do a lot of young gets, I call that kind that young adult fiction category. It's actually probably what I read the most. If you could go on vacation anywhere in the world tomorrow, where are you going?
Guest: Alok Ajmera (49:11):
My happy place is on a mountain any season. If it's winter, it's skiing. If it's summer, it's hiking me on a mountain. Disconnected from my phone. Is it is my happy place and the place that I've been trying to get to or get back to is actually in Peru. So I would say Peru, getting back to the Sacred Valley and the hikes, the mountains that you can do there. It's just something, I had been a fortune to go there a couple years ago with my girls and we hiked the Machu Picchu Trail and it just has something so magical about it. So I would get back to Peru, get back onto a mountain, get disconnected from my devices.
Host: Paul Barnhurst (49:48):
There is something magical about it. I've done a mountaineering course. I used to do Boy Scouts as a leader for years. I've done a lot of hiking. I've done a 50 mile backpacking trip for a week with a bunch of young boys, and I love the outdoors. So similar to you. I love a good hike in the mountains, so that's when people say beach, your mountains. I'm mountains.
Guest: Alok Ajmera (50:07):
Yep. I'm the same. And any season, if it's winter, it's cold. No problem. Then I ski. Yeah.
Host: Paul Barnhurst (50:12):
Funny. I never picked up skiing and I live in Utah. That's a shame, but I never had money as a kid. We did a tonne of what we called extreme sledding. I don't think I've ever shared this from the show, but you'll appreciate it. My dad would take us up the mountain, drop us off at a spot, and we probably went down over a mile and he'd pick us up at the bottom and take us back up and we'd ride our slide. There were parts, we were going 30, 40 miles an hour. We'd have to walk across the road and keep going. I still remember we had, my sister broke her hand one time. We had some pretty good, it was some of the more extreme sliding I've ever seen, so that was my closest 15.
Guest: Alok Ajmera (50:44):
Parents won't let their kids do anymore, like fly down the I of a mountain that goes down a mile with
Host: Paul Barnhurst (50:53):
No helmet. No
Guest: Alok Ajmera (50:55):
Helmet. That sounds
Host: Paul Barnhurst (50:57):
Amazing. Yeah, totally. But that was what we called extreme sledding and we did quite, quite a bit of that as a kid. So that was my poor man's version of skiing.
Guest: Alok Ajmera (51:08):
Very cool.
Host: Paul Barnhurst (51:10):
Alrighty, well, thank you so much for joining me. It's been an absolute pleasure to chat with you. I hope the audience enjoys the conversation as much as I did, so thank you for carving out some time elope. I really appreciate it. Thanks,
Guest: Alok Ajmera (51:20):
Bob. Topic that I love. Great conversation. I hope everyone got some value out of it.
Host: Paul Barnhurst (51:25):
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.