Real Innovation Doesn't Need a Billboard

Real Innovation Doesn't Need a Billboard

On choosing between AI marketing and real automation.

AI is everywhere. It shows up in every pitch deck, every product page, every keynote, doing the exact same job each time: signal innovation, regardless of what's actually underneath it. At some point that stops meaning anything, and I think most finance leaders realised that a while ago.

This isn't going to be another piece whether or not AI is wrong. There are enough of those already.

Instead, I wanted to focus on something more specific for CFOs and finance teams: show where automation is genuinely creating value in finance right now, and where it isn't yet, no matter what the pitch says.

What's actually happening with AI ROI right now

Here's the part that often doesn’t get mentioned. The return on AI investment across finance and operations is nowhere near what most pitch decks promise, and the data backing that up is hard to ignore. Research into enterprise generative AI adoption has found that the large majority of pilots aren't showing measurable financial return within six months of launch. A recent global survey of senior leaders found that only a single-digit percentage have managed to establish AI ROI at all.

Gartner's own forecasting expects a meaningful share of the generative AI projects that started back in 2024 to be abandoned by the end of this year, either because the ongoing cost of running them at scale doesn't hold up, or because nobody can point to the value they're actually producing.

Let’s pause here and look at the pattern. The cost that kills most of these projects is rarely the price given during sales conversations. It's everything that comes after: getting the data clean enough to trust it, checking whether the output is actually right, correcting what’s not and heavy upskilling. The ROI case falls apart when AI is moved from demo into daily use.

Even among the CFOs actively experimenting, the most common three use cases are still reports, accounts payable, and extracting data from documents. None of that touches the operations that actually move money, and better yet, OCR and accounts payable automation are tools that have been available to finance teams long before the AI boom. Payments, reconciliation, the ledger itself, none of it has moved significantly because of AI.

I don't think that's an accident.

Why finance is the wrong place to move fast

Some tasks are forgiving of a rough pass: a first draft gets edited before it’s sent, a summary can be amended. A payment doesn’t work that way. Once it's sent, undoing it means reversing a real transaction, with real friction and real cost attached. Same for a reconciliation, or a ledger entry gone wrong. Those need to be right the first time, not checked repeatedly until someone feels reasonably confident.

And if a person still has to verify every AI output before it can be trusted with company money, those checks are still manual work. It just moves to later in the process instead.

So when finance leaders are cautious with AI, slower to hand over the ledger, more skeptical about the privacy and data-quality implications, I think that caution is well-founded.

Where the real value is

So where does that leave the actual case for automation and AI? Because there is one, and it's already been made for years. Payables automation and OCR have been solved reliably by vendors finance teams already trust, and almost none of it was branded as AI until recently. That's proof there's a real case for automation done properly: accurate, accountable, reliable.

So it comes down to a choice: chase the promise in AI marketing pitches, or put that energy into automation precise enough to actually earn trust with a finance function. We've chosen the second.

This means reconciliation matches every payment automatically, because each customer has a dedicated account behind it rather than an algorithm inferring what a messy reference probably means. It means payment execution that follows rules the finance team set in advance, visible and auditable at every step.

None of this is a black box.

Most of the AI conversations right now are still measured in pilots and quarters, but a well-executed automation gives ROI in weeks. I think about one client we work with, a premium event company, whose accounts payable process used to eat hours of manual work every week. Automating it properly delivered a 90% efficiency gain, saved close to 280 hours a year, and cut fund delivery down to 24 hours.

Implementation follows the same pattern: days to onboard, a few weeks at most to go fully live, not a specialist project with its own budget line and its own project manager. That's a real answer to the learning curve finance teams keep running into with AI specifically. Months spent figuring out how to prompt something correctly or catching mistakes before they became cost. Does your team need to spend time on that? No. The automation should use the workflows they already have.

A final word

Real innovation rarely needs a billboard. Most days, it's working quietly in the background, and it has for years, evolving, getting better, winning customers’ trust because of its accuracy and reliability. Most of it is unnoticed, which is exactly the point.