Theo Wasserberg, Head of UK&I, Embat.
Discover top fintech news and events!
Subscribe to FinTech Weekly's newsletter
Read by executives at JP Morgan, Coinbase, Blackrock, Klarna and more
At Google's Gemini Founders Forum this year, they had a term for what happens when AI looks impressive in demos but never changes how work gets done: AI theatre.
Finance and fintech teams know this performance well. They sat through it all year. The pilot that would "transform cash visibility." The dashboard that would "revolutionise forecasting.” The platform that would "finally connect everything.” Exciting presentations. Polished decks. Then back to hunting through spreadsheets for yesterday's cash position.
2025 wasn't the year AI-infused fintech modernisation failed. It was the year we learned what separates the show from the work.
What Actually Happened in 2025
2025 didn't deliver the fintech AI revolution everyone predicted. Instead, we saw something more valuable: growing awareness of what modernisation actually requires.
AI Pilots Are Easy, AI-First Workflows Are Not
Teams stopped pretending AI deployment was simple. Painful experience helped people distinguish between "AI pilots" (exciting to demo, impossible to scale) and "AI-first workflows" (boring to build, essential to operations).
Whether its payments, banking or treasury, everyone can list a dozen AI use cases they were pitched or watched a demo of. The challenge wasn't innovation, identifying a problem on paper or imaginative solutions. It was turning the exciting demo into something banking operations, payment processing teams, and treasury departments could trust and use daily.
Top-down AI mandates produced impressive presentations that never changed how work got done. We learned that real progress happens when the people closest to the work can design and own their agents.
The lesson: without the right infrastructure, AI can't scale beyond the team that built it.
The Board-Level AI Gap
A recent Think & Grow report reveals that only 32% of UK startups and scaleups have AI expertise at board level, trailing the 40% of FTSE 350 tech firms that have appointed specialists. This gap risks stifling growth amid strong investor interest in AI.
However, there's a stark divide by company size. Scaleups with over £50m revenue are more than three times as likely to have AI expertise (50%) compared to their smaller counterparts (15%), though 32% of companies overall plan to make appointments in the next year.
As fintech evolves toward AI-first workflows, board-level AI knowledge becomes essential to distinguish hype from scalable operations, ensuring startups maximise funding and compete effectively.
Breaking the 30% Automation Ceiling
For twenty years, finance automation followed the same principle: if X happens, do Y. That approach created value - automating maybe 30% of manual work - but it also created a ceiling [Deloitte Survey, 2024; McKinsey Report, 2024].
Real life rarely follows rules perfectly. A customer pays two invoices in one transaction. Someone mistypes a reference number. A file format changes. Suddenly, the system freezes, and a human must step in. The promise of efficiency evaporates in exception handling.
AI, by contrast, doesn't require every rule to be predefined. It understands intent. You tell it the desired outcome, and it figures out how to achieve it. That's the difference between 30% automation and 99%. It's also the difference between a system that merely saves time and one that transforms how finance operates.
The real breakthrough in 2025 was understanding this wasn't just incremental improvement. It was a different category of capability.
The Contained Value Breakthrough
One of 2025’s most important lessons came from understanding what didn't work, and why.
Early AI mistakes involved treating it as a cosmetic upgrade atop legacy systems. Teams that succeeded took a different approach: contained value.
Contained value means specific, auditable use cases where you know what AI will do, who it serves, and how success will be measured. Not "transform X process or industry." Instead: automate reconciliation first, then forecasting, then cash visibility. This builds confidence, one use case at a time.
The teams that flipped from cost centre to strategic partner did so by making AI agents accountable for specific outcomes. Not vague efficiency targets, but measurable work removed: reconciliation time cut by 75%, forecasting accuracy above 90%, audit prep compressed from weeks to days.
They stopped talking about AI strategy and started retiring manual processes. They killed familiar workflows when data showed better paths, even when it made people uncomfortable.
We Watched Finance's Role Fundamentally Shift
Something deeper was happening beneath the surface of failed pilots and stalled initiatives. Finance itself was evolving.
In a revealing shift, HSBC found that 64% of CFOs at large organisations now consider treasurers part of the C-suite, reflecting a mindset change: finance is no longer viewed purely as a cost centre, but as a catalyst for insight and strategic agility [HSBC Corporate Risk Management Survey, 2024].
This wasn't just finance functions - whether treasury, payments operations, or banking teams - getting better at their traditional tasks; they were becoming something entirely different.
Digital adoption is central to solving this challenge: 80% of CFOs expect digital tools to dominate operations by 2025, while 30% of finance tasks are fully automatable [Deloitte Survey, 2024; McKinsey Report, 2024]. By modernising tools and processes, companies can both attract top talent and unlock productivity gains.
Over the next five years, 69% of CFOs expect greater emphasis on data analytics, 60% anticipate more scenario planning, and 55% say finance will become a more embedded strategic partner across the business [Cherry Bekaert CFO Survey, 2025].
But you can't advise strategy while drowning in exception reports.
The Real Shift Isn't Technical
What 2025 ultimately revealed is that modernisation isn’t a technology contest, it’s a cultural reckoning. Finance operations are not evolving just because the new tools are powerful. They're changing because leaders finally confronted how much of their operating model depends on institutional memory, heroic manual work, and processes that only "function" because people quietly filled the gaps.
Modernisation begins not when teams deploy agents, but when they stop accepting complexity, exceptions, and rework as the unavoidable cost of doing business.
The show is over. The gap between those stuck in the front row and those who have traded theatre for measurable outcomes is widening fast.
