For thirty years, the unsung heroes of the finance department have spent their careers matching hundreds of items a day within the cells of a colour-coded, multi-tabbed, Frankenstein’s monster of a spreadsheet, a creation stitching together their firm’s entire operation.

Countless generations of new-starts have been trained to tiptoe their way through it, and some have even tried to cut past the mess of nested formulas, only to have half the workbook light up with errors.

“Maybe in the noughties,” you might scoff, “but no chance are big businesses still propped up by those ancient Excel files now, not in the age of AI.”

Sorry, but the uncomfortable truth is that those creaking spreadsheets and manual workflows never went away, they just got more tangled and more critical – to the point that, last month, 90% of finance managers admitted spreadsheets are still an integral part of their operations.

But those same firms neglecting back office innovation are the ones ploughing millions into slick front-end apps and dashboards, customer service chatbots and automation pilots, with almost all (96%) having rolled out AI elsewhere in the business.

Back Office Bottleneck

“A huge number of firms have at least done some adoption of AI,” said Jim Sadler, CTO of Glasgow‑based reconciliation platform AutoRek, the quiet giant pushing to replace the antiquated tools and controls still keeping the financial world upright.

“That’s great, but firms now need to deploy as much attention to streamlining and automating the back office as they have in innovating the front office.”

For many organisations, critical background functions have been left behind, sometimes by design, sometimes by neglect, in the all-out rush to adopt AI, multi-cloud, IoT, and edge networks. 

But those choices have resulted in a now desperate need to replace failing legacy systems and manual processes, problems that, according to AutoRek’s 2026 Payments Survey, more than two‑thirds of businesses say are actively holding back their growth. 

According to AutoRek, these gaps are also creating serious regulatory and competitive risks. For example, the FCA is about to introduce a tougher safeguarding regime, coming into force on 7 May, meant to tackle long‑standing flaws in how funds are protected in the event of shortfalls or business collapse.

But most firms aren’t ready. Despite there being barely weeks to go, only a third say they’re fully prepared to meet the compliance deadline, while 84% expect their controls will need more updates within the next year to stay up to standard.

“With the ever-growing number of transaction volumes and data sources on one hand, combined with the regulatory need for safeguarding customer money, you get to the point where the way you manage the back-office, at a lower scale with lower complexity, ultimately means it can’t survive anymore,” said Sadler.

But the fault lines run deep, rooted in both data readiness and workforce capability, problems that no amount of automation can paper over. 

The Culture Gap

Every enterprise firm is painfully aware that its grand ambitions for AI are dependent on staff who, until just a few years ago, had spent their careers working with tools designed for a world that suddenly no longer exists.

That’s particularly true in the back office, where a single AI tool is capable of handling functions that once needed specialist teams, tasks like bookkeeping, invoicing, and payroll processing, a sea change that has resulted in business leaders reporting up to 20% overcapacity from automation.

While the AI skills gap is a ubiquitous problem, back office teams are different in that they are also struggling with a cultural shift.

The likes of accountants, admin assistants, payroll clerks and procurement managers have, for decades, been trained for stability and compliance, not speed or experimentation. Now they’re being asked to adapt to AI tools and real‑time data flows, with most left reeling and unprepared. 

That has slowed progress, but for Sadler, it’s an opportunity for the back office to finally be freed from decades of manual matching drudgery and repurposed for higher‑value work. It’s a transformation that AutoRek claims its AI-enabled platform has already helped more than 120 firms achieve.

“Wouldn’t it be great if all of that tedious work could be removed? Wouldn’t it be great if your agent, working alongside you, could eliminate them for you, and you could just approve that that’s acceptable?”

“We’ve got to automate reconciliation to remove it from being a human task. Maybe then you could get involved in bringing on new data schemes, maybe you could actually innovate faster because you’re not encumbered by a finance operations team or a set of processes that are holding back growth. 

“Maybe that human workforce can actually be put to use in driving further innovation.”

Garbage In, Garbage Out

But even if that skeleton crew of humans left working in the back office is ready and raring to integrate AI, the foundational, critical data they need to do that is stored in conflicting formats, locked inside scattered, third-party systems never designed to work together. 

Trying to layer AI, or any automation, on top of that kind of fragmentation doesn’t just limit its impact, it risks making worse the inefficiencies firms are trying to eliminate.

“Simply deploying AI into the back office doesn’t solve the problem. If you train a large language model on garbage, then it amplifies the garbage. AI is not going to resolve the fundamental problems of data control and fragmentation at the core,” said Sadler.

“In a finance ops team, we’ve had rules-based software for thirty years that helps firms bring in data sources, build rules, and match items. But the ingestion of data sources and building match rules can be labour-intensive for a human when you’re looking at maybe fifty different data sources all in different formats.”

The vast majority (80%) of companies report operational disruption from this fragmented data, especially around reconciliation and monitoring, an urgent issue given the need for real-time controls as AI agents are increasingly handed the power of a credit card.

Recommended reading

• Finance Leaders See 2026 As the Year to Scale AI
• Report: UK Firms Losing £27m Per £100m in Transformation
• Most UK Banks Use AI, But Consumer Concerns Remain

How urgent? Well, the finance industry had reckoned on being much further ahead by now. As far back as 2021, it was widely predicted that, by 2025, 80% of high-value payments would be made using the ISO 20022 protocol, a new global standard replacing limited message formats with a more detailed and universally understood “language” for payments data.

ISO 20022 was meant to bring order to the chaos by standardising payment data end‑to‑end, but its arrival is exposing just how unprepared many firms are, with less than a third fully migrated and live.

This slow adoption is a symptom of a deeper malaise, with core financial functions having already reached their limit, but left facing the complexity of integrating new payment rails, like instantaneous transactions and blockchain, making issues like data fragmentation that much worse.

“The firms that continue to address front-end innovation at a faster rate than back-office modernisation, they probably won’t notice anything breaking immediately, but they’ll see a gradual slowing down of their ability to grow and innovate,” said Sadler.

“And at that point, they’re going to have to stop their front-end innovation, fundamentally restructure and address the back-office, before they can carry on.”

“This isn’t a new problem, it’s been a growing concern for firms for a long time. And we’ve effectively reached the inflection point where the speed of innovation in the back office now needs to catch up.”