What's real AI and what's RPA by any other name? Observations from FATE 2025

Walking the floor at FATE was a bit like being in Groundhog's day with AI rearing its head everywhere you turned. Every booth had it on the screen. Every presentation had it in the headline. Every demo script mentioned it at least three times.

But here's what we noticed after two days on the floor:

most of what's being called AI in finance automation isn't actually AI. It's automation with better branding.

Not to say there's AI fatigue, but repeat something often enough and it beings to lose its meaning.

What Most Vendors Are Calling AI

The "AI-Washing" fell into three camps:

Rule-based automation that got rebranded. If it follows predefined logic: match this invoice to that PO, flag expenses over a threshold, that's not AI, that's deterministic automation. Valuable? Yes. AI? No. It can't adapt or handle exceptions outside its programmed rules.

OCR with a new marketing label. Reading text from invoices and mapping it to fields isn't intelligence. It's pattern matching. The accuracy improved, sure. But it's not AI.

Chatbots that pull from FAQs. A conversational interface running keyword searches is just a search tool with a chat window. Helpful, but not intelligent.

None of these are bad. They automate work and save time. But they're not AI, and calling them AI sets false expectations.

What Real AI Actually Looks Like

Real AI should operate like a team member, not software. The clearest marker is agency. Real AI agents complete workflows autonomously. You interact with them like you would a junior hire. You assign work, they ask clarifying questions, you approve or redirect, and they execute.

The interaction should be conversational. You're not navigating dropdown menus or setting up rule trees. You're talking to the system. Questioning its decisions. Redirecting when needed. The intelligence isn't hidden behind a SaaS interface. It's right in front of you.

Real AI also adapts. It learns from your corrections and improves based on how your team actually works. It doesn't require reconfiguration every time a vendor changes their invoice format. It figures it out.

At FATE, only a handful of vendors had this. Most were still building systems where you do the thinking and the software does the clicking.

Where AI Still Falls Short

It's important to note that while true agentic AI can do a lot - it still has it's limitations. AI is great at patterns and workflow orchestration, but it's not built for precision calculations.

AI can't reliably count the number of "C"s in this sentence, you definitely can't expect it to handle complex multi-step financial calculations.

Real systems combine AI with deterministic code. AI identifies the pattern, routes the workflow, handles the exceptions. Code does the math.

If a vendor tells you their AI does everything, including the calculations, they're either misrepresenting how it works or building something that'll fail an audit.

The best implementations keep these separate. AI makes decisions. Code ensures accuracy.

What This Means for CFOs

The AI-washing problem isn't just annoying marketing. It's creating real risk.

Teams are paying premium prices for capabilities they already have. Or investing in tools that won't scale the way they expect.

The vendors aren't necessarily acting in bad faith. The terminology shifted fast. But your job is to evaluate what the tool does today, not what it might do in two years. The market is ready for AI. The curiosity is there. Budgets are loosening. The operational case is clear. But the technology is uneven. Some vendors built genuinely intelligent systems. Most are selling automation with a new label.

The advantage goes to teams that can tell the difference. Real AI changes how work gets done. It shifts capacity, reduces bottlenecks, and handles complexity that used to need senior judgment.

Automation makes existing work faster. AI makes new work possible.

As finance tech matures and vendors converge on features, the intelligence layer is where differentiation actually lives. The teams that recognize this early will build stacks that scale in ways their peers can't match.