Transcript: FP&A Clinic – Speed vs. Accuracy 

Liran: Hello. Hey, Ramya. 

Ramya: Hey, Liran. Hello, everyone. Good morning, good evening. 

Liran: Yes, I see that we have a small group, so maybe we’ll just wait another minute or two until other people join. I know that usually, because it’s not very formal, it takes some time for people to join. We see some people joining in, so hello everyone. We will start in just another minute or two to allow others to join. As always, this is recorded, and I think it’s a good time to say hi to those who are listening to this recorded session as well. 

Ramya: Sure. Amazing. 

Liran: So, for those who did join on time—although again, a very small group—hopefully, other people will be joining. Even with a small group, we can have a fruitful conversation. Let me check if there are any technical problems. No, I think it’s a live room; it should all be good. 

Again, thank you to those joining. Because it’s such a small group, if you want to use the chat box to introduce yourself, I have enabled the Q&A and the chat. You can say hi and introduce yourself over there. This is the first time we’re doing this FP&A Clinic, which is amazing. What I would like to do first is introduce Ramya, whom I have known for many years. I’m very lucky to know Ramya and to have her host this in the FP&A Hub. 

Ramya, you have been doing the FP&A Clinic for quite a while. Maybe we can start by explaining exactly what a “clinic” is. We know other terms like roundtables and webinars, so what is so unique about a clinic? You’re not a medical doctor, right? You’re more of a planning doctor. 

Ramya: Maybe I wanted to be a doctor before I got into this profession, so perhaps there’s a linkage there! Thank you, Liran, for the introduction. 

I’m Ramya. I come from a finance background—CPA, cost accounting, and an MBA from Wharton, the University of Pennsylvania. That’s from the academic side. From my career perspective, I’ve been working in finance technology for more than two decades, specifically in Enterprise Performance Management (EPM) for over 15 years. 

So, why a clinic? Coming back to Liran’s question, when you talk about a clinic, you immediately think of a doctor’s clinic. Think about it like this: when we go to a doctor’s office, we do not hear a lecture about medicine. Instead, we talk about our symptoms, figure out together what is really happening, and then apply a solution. That’s exactly the idea behind the FP&A Clinic. 

Personally, I have sat through many webinars and classes where frameworks, best practices, and good ideas are presented. It’s great to get inspired, but at the end of the day, when we go back to our desks, the question is: How is this applicable to me? My situation is very different. It doesn’t mean those frameworks aren’t good, but can I practically apply them? Can I discuss my specific situation and see what direction I should move forward in? 

That’s the idea of this clinic. It’s a peer discussion where someone with experience can provide pointers to inspire the team, diagnose the issues, and figure out the next steps together. 

Liran: Amazing. When we think about the clinic, what are usually the topics? I’ll be honest, today’s topic—Speed vs. Accuracy—is a great idea. I’ve never heard that as a standalone topic for a webinar. We always hear about how to do planning in manufacturing, how a forecast differs from a budget, or how to use AI. Here, you’ve come with something pretty unique. How do you choose the topics? 

Ramya: Thank you for that. There is a reason I chose today’s topic, which I’ll talk about shortly. In general, the theme goes with what the community needs. However, the way I’ve been running the clinic so far is by looking at current business trends. There are common cycles and patterns in every business. For example, by August or September, everybody starts talking about the budget. Right now, everybody is talking about the forecast in one way or another—whether they are in a regular forecasting cycle or preparing for the upcoming budget season. 

Forecasting is one of the biggest conversations happening right now across the industries I work with. So, the overarching theme aligns with what the business is undergoing at a macro level, and we pick specific themes in between, like business partnering or aligning with operations. 

Liran: We don’t have a big group here today. My understanding is that no one needs to come prepared to these meetings, right? What is your recommendation for the audience joining these peer discussions? 

Ramya: There is no preparation needed. Just come with an open mind, be curious, and be ready to put forward your questions and your problems. The topic will be announced in advance, and I might have a few anchor points or common questions to set the foundation. The true metric of success is what we take away after the session, not what we come prepared with. 

Liran: And that’s why it’s so great to have it in the community. This isn’t a one-on-one consultation; we’re bringing together practitioners at different levels to share ideas conversationally. I want to say that as one of the co-founders of the community, this was exactly my vision. Many FP&A professionals feel isolated on an island and want to speak with peers. This clinic is a great opportunity to do that live. 

If anyone has a topic they want to see in the coming weeks, just post it or reach out directly to Ramya in the community. I think it’s a great time to transition to today’s topic. Ramya, do you want to present? 

Ramya: Yes, thanks, Liran. Let me share my screen. Hope all of you can see my slide. 

Liran: It’s showing the PowerPoint interface, not the slideshow. Okay, now we can see it. That’s great. 

Ramya: Awesome. As I mentioned, most organizations are currently having forecasting discussions, especially with budgets around the corner. I felt it was relevant to discuss speed versus accuracy, as this is invariably the challenge every finance organization deals with. After working on multiple planning transformations across industries, I’ve started wondering: Are we asking the wrong question? 

Perhaps the question should be: What decision is the forecast trying to support? Once we answer that, finding the right balance between speed and accuracy becomes much clearer. 

I put together a slide with some observations based on various clients. I’m going to share these perspectives and then open it up for discussion. 

First: Speed. Finance leaders often ask, “How fast can we get the forecast?” My response is, “How fast for what?” If the business makes pricing decisions on a weekly basis, a monthly forecast is already too late. If your board meets quarterly, a quarterly forecast is appropriate. It largely depends on the decisions you are trying to make and how often they occur. 

On the other side of the equation is Accuracy. As finance professionals, we love accuracy—it’s in our DNA. I have seen teams spend days trying to squeeze out that final 1% or 2% of accuracy while the business is left waiting to make a decision. The question is: How accurate is accurate enough for the decision at hand? The idea is not to lower our standards, but to align the required accuracy with the concept of materiality. 

This brings me to my favorite topic: Are we looking at a false tradeoff? Often, the failure point of a forecast is not about how accurate or fast it is; it’s about how it is being used by different parts of the organization. 

For example: 

• The CFO wants the forecast to create confidence with the board. 

• Sales leaders want visibility into the pipeline. 

• Operations needs to know how many resources to allocate. 

• Finance is worried about predictability. 

Depending on the decision you are making—whether it’s launching a new initiative or slowing down hiring—the required level of accuracy is entirely different. 

People often ask what “elite” FP&A teams do differently to balance this. The answer isn’t a sophisticated model or better technology. It’s all about alignment. Are sales and finance aligned? Are HR and finance aligned? Elite teams spend time aligning definitions across the organization (like how they view the chart of accounts or group customers) rather than just worrying about assumptions and numbers. 

Now, I’ll open it up. What are the challenges you are facing with your forecasts today? 

Liran: I’ll chime in for a second. When you used the term “complex modeling,” I think that ties directly into speed. If you want a forecast quickly, it can’t be overly sophisticated or complicated. We often assume a complex model will be more accurate, which isn’t always true. We are dealing with a triangle of speed, accuracy, and complexity. 

Ramya: That’s a very valid perspective. High complexity doesn’t always equal high accuracy, but it almost always leads to a delay. I had a client implementing an employee model who wanted to calculate every single benefit line item exactly the way accounting books it—accruals and all. While it impacts cash flow, you don’t need to go employee-by-employee just to get that final 5% of accuracy. They made the model too complex by over-focusing on accuracy, which eventually impacted speed. 

Liran: This reminds me of running AI models. We’re half an hour in and finally using the term “AI”! When you try to run AI on a highly granular level (like predicting sales for a specific retail store department), the results often aren’t great because there isn’t enough robust history. But if you run predictions at the regional level, it’s much more accurate. How do you see AI plugging into the speed versus accuracy conversation? 

Ramya: I purposefully didn’t bring up AI initially to keep us focused on the fundamentals. Often, the fundamental problems in a planning process only crop up when a company tries to implement AI or automation. If an AI implementation fails to provide accuracy, it goes back to organizational alignment. If the organization isn’t aligned on definitions, the data fed into the AI will be flawed. 

I’d love to hear from the audience. Are you facing issues with speed, accuracy, or assumptions? You can use the chat or unmute yourselves. 

(Audience interaction starts) 

Veronica: (via chat) Accuracy was a problem in my last position. 

Ahmed: Can you hear me? Sorry for the technical issues. 

Liran: That’s fine! We are FP&A experts, not technical experts. 

Ahmed: I wanted to touch on the AI topic. I work at the headquarters for [a large corporation], and we are exploring ways to implement AI within central FP&A. We are facing several challenges. One is governance—reporting calendars, business rules, and allocation exclusions are not the same for everyone. 

The second issue is cost. We don’t fully master the cost drivers of AI agents yet, and it’s hard to project those costs. Third, it’s evolving so fast that we don’t even have time to test new tools before something else comes out. 

Implementing AI on clean data and processes is easy. The hard part is aligning finance and business processes. We are currently trying to re-engineer our processes to be “AI by design,” rather than just slapping a Copilot chat on top of legacy processes. 

Ramya: That is a very real problem, Ahmed. When I attended an AI course at Columbia Business University, the professors reiterated exactly what you just said: you cannot just put an AI Copilot on top of broken processes, because it will only amplify the mess. AI provides an opportunity to rethink your business processes fundamentally. 

Ahmed: To give a concrete example: A few weeks ago, I used an Excel Copilot to do a simple mapping between two master data tables (e.g., SAP vs. another source). The data was the same but labeled differently. I gave it clear business rules. It produced something that looked clean, so I didn’t check it thoroughly enough. We later realized there were edge cases the AI either missed or made silent assumptions about. 

The issue is auditability. I cannot go to my team and say, “This is Copilot’s fault.” I am accountable. How can we ensure humans can easily audit AI work without spending so much time checking it that it kills the efficiency gains? 

Liran: I hosted a webinar on AI in finance last year, and accountability is a massive pillar that often contradicts finance requirements. You are accountable for the data, not the AI. If AI gives you a highly inaccurate result quickly, and you spend more time fixing it than if you had done it manually, you lose both speed and accuracy. 

Ramya: Two points come to mind. First, think of AI like giving instructions to a child on how to make a peanut butter and jelly sandwich. In our minds, we think we are giving perfect instructions, but for the other party, it might be incomplete. That is where human prompt engineering and domain expertise become invaluable. 

Second, every business must strive for “human-in-the-loop” AI. AI performs a task, and a human validates it. Even if you spend time validating, it should theoretically still be faster than doing it manually. 

Liran: I sometimes think of AI as having three enthusiastic, fresh-out-of-college junior analysts working for you. If a junior brings you a report, you challenge it. You ask where they got the data. You have to adopt that same mindset with AI to validate accuracy. 

Also, Ahmed, you mentioned your organization is constantly changing. We have to set realistic expectations: AI may not be perfectly resilient to rapid, constant business changes unless it is constantly retrained. 

Ahmed: The ultimate goal is to mix finance and non-finance data so we can act as true business partners rather than just the “Excel people” who create PowerPoint slides. To do this, finance people need to learn technical skills to become “doctors of AI agents,” understanding how they think. Unfortunately, in large corporations, finance often says, “That’s IT’s job,” and IT says, “That’s a business topic.” 

Veronica: I agree completely. I’ve encountered finance people who refuse to do “technical stuff.” The older generation needs to get with the digital program because AI is being embedded in everything. You have to understand its pitfalls. 

I’ve also noticed that I will teach an AI, it works beautifully for a while, and then it degrades. You have to constantly stay on top of it. 

Ramya: From what Ahmed and Veronica shared, we are dealing with three distinct areas of AI implementation: 

1. Adaptability/Change Management: Rethinking the processes. 

1. Adoption: Shifting the human mindset to utilize the tools. 

1. Maintainability: Updating the AI as business realities change (what Veronica mentioned). 

If I were approaching a client engagement, I would segregate these into three different initiatives so we don’t get overwhelmed. Regarding the mindset shift, it doesn’t always require deep technical knowledge. It just requires openness and discipline—like consistently validating meeting minutes before feeding them into the system. 

Liran: This has been a great conversation, and we are already at the top of the hour. We can definitely continue this in the community discussion groups offline. Ramya, thank you so much for taking the initiative to launch these clinic sessions. 

Ramya: Thanks, Liran. I enjoyed it and learned a lot as well. Looking forward to the next one! 

Ahmed / Veronica: Thank you very much!