AI in Finance: Cutting Through the Hype with Bogdan

AI is everywhere, and finance is no exception. Every week, it seems like there’s a new headline about how AI is revolutionizing the industry, replacing jobs, or transforming the way we work. But how much of this is real, and how much is just noise?

If you’ve spent any time scrolling through LinkedIn, you’ve probably seen bold claims like “AI will replace the CFO in five years” or “This one AI tool will automate your entire finance function”. But if you’ve actually tried using AI in a finance setting, you’ll know that it’s not quite that simple.

To get a clearer picture of how AI is actually being used in accounting and finance today, I sat down with Bogdan Năforniţă, CEO and founder of Profluo, an AI company that works directly with finance teams. We talked about what AI can realistically do right now, the biggest misconceptions about AI in finance, and what CFOs should be paying attention to as the technology evolves.

If you’ve ever wondered how AI fits into your finance function—or if you’ve rolled your eyes at yet another overhyped AI promise—this one’s for you.

AI in Finance: What’s Real and What’s Not

Right now, everyone’s obsessed with Large Language Models (LLMs), but Bogdan didn’t hesitate to call out their limitations.

“In finance, we’re storytellers—but our stories are based on hard numbers,” he said. “If AI is going to play a real role in finance, it needs to be focused on accuracy, not just generating text.”

So where does AI actually fit in finance? Broadly, it falls into two categories:

1. Pre-ERP AI – This is where most AI tools live today. Think document processing, automatic accounting, statutory and commercial validations, 3-way-matching and securing approvals before data enters your ERP system.

2. Post-ERP AI – This is where AI analyzes existing ERP data to extract insights, predict trends, and improve decision-making. It’s happening, but we’re not quite there yet.

For now, pre-ERP AI is more practical and widely used. The technology to automate

The Reality of AI in Financial Forecasting

LLMs are great at summarizing and generating narratives, but they aren’t built for financial forecasting. That’s a job for predictive algorithms, which use historical data to identify trends and make projections.

“A lot of people assume AI can do everything,” Bogdan pointed out. “But when it comes to forecasting, you need specialized models—not an LLM trained to guess the next best word in a sentence.”

One promising development? Reasoning models. Unlike standard LLMs, these models can show their chain-of-thought process, making them useful for reconciliations, risk monitoring, and decision-making. One of the biggest mistakes companies make? Throwing LLMs at every problem.

“AI isn’t magic,” Bogdan said. “It’s about choosing the right tool for the right problem.”

A quick breakdown of AI’s role in finance:

• Document processing? Use classification models (not LLMs) to make decisions about relevant financial entities that were found in documents.

• Forecasting? Use predictive machine learning models. A special type of predictive behaviour is automatically filling in accounting and reporting entries. The more sophisticated a company is, the more expanded the reporting hypercube could be - think profit centers, projects, expense categories, SAF-T codes, commodity codes.

• Risk monitoring? AI-driven analytics.

• Process automation? RPA (Robotic Process Automation) has become a bit too rigid and brittle in our super-volatile world. This is now being replaced by Agentic workflows, which is an exciting area, but only when tackled right. Agent autonomy is highly appreciated, but Agents need to run on auditable guardrails and their behaviour needs to respect enterprise security and privacy. Also Agents need to understand and comply with enterprise contexts. Nobody wants an Agent that decides to pay invoices outside the company policies

“If you’re asking ChatGPT to pick your top 10 stock investments for 2025,” Bogdan joked, “you deserve whatever happens next.”

Security & Privacy: The AI Blind Spot

AI adoption comes with a huge but often overlooked issue: data security.

“I see people uploading confidential financial documents into public AI models without thinking twice,” Bogdan said. “That’s a massive risk.”

The challenge? Ensuring AI tools are secure, auditable, and role-based, meaning employees only see what they’re authorized to see.

“It’s not just about who can access the data,” he explained. “It’s about what the AI itself is allowed to answer.”

Bogdan’s team is tackling this by embedding confidentiality controls directly into AI models, making sure sensitive financial data is not disclosed to inquisitive prompters across the enterprise. This is quite hard to do, considering how LLM architectures actually process information by using tokenization, and then embedding into vectors in order to assign semantic meaning to each token. All these steps need to carry security and confidentiality dimensions: to which legal entity does the data belong? What departments or user roles are allowed to work with it?

With AI moving so fast, how do finance leaders keep up? Bogdan kept it simple:

“As a CFO, you don’t need to be an AI expert. But you do need to understand which AI tools solve which problems—and which ones are just hype.”

His advice?

1. Be specific. What problem are you solving? Automating transactional accounting? Enhancing forecasting? Reducing manual data entry? Define it first.

2. Track accuracy. If AI can’t get over 90% accuracy, it’s not worth implementing. Fixing AI’s mistakes shouldn’t create more work than it saves.

3. Think security first. Make sure your AI solutions protect confidential data and are fully auditable.

Choose the right partners. The best AI tools are specialized. Work with vendors who actually understand finance.

AI and the Future of Work in Finance

With all this automation, are finance jobs at risk? Not exactly.

“It’s not AI replacing people,” Bogdan said. “It’s people who use AI replacing people who don’t.”

Instead of eliminating jobs, AI is creating AI-assisted finance professionals—people who use AI to be more strategic, efficient, and valuable to their companies.

“I don’t buy into this ‘AI will replace us all’ narrative,” Bogdan added. “People don’t want an all-knowing AI God. They want tools that make them better at their jobs.”

And that’s exactly where AI in finance is headed.

AI in finance isn’t about replacing people—it’s about making finance smarter, faster, and more secure. The key takeaway? AI is only as valuable as the problems it actually solves.

CFOs don’t need to follow every AI trend—they just need to understand how AI can make their teams more efficient and effective. With the right approach, AI isn’t a threat. It’s a game-changer.