By João Pedro Almeida, CEO and Co‑Founder of Noxus AI  

UK financial services firms handled 1.85 million complaints and paid out £283 million in redress, the highest level since the end of the PPI cycle, in the first half of 2025. Also, given the continued financial uncertainty and events such as the car financing scandal there is no sign that the situation is improving. The latest Financial Conduct Authority (FCA) complaint volumes and redress levels have risen significantly in recent reporting periods that highlights a shift that should concern both incumbents and fintechs.

At first glance, this looks like a familiar volume story, however, it is not. The more important warning sign is operational: complaint handling is slowing down, despite sustained investment in automation. What is needed is not more of the same, but new AI-driven approaches and processes for handling complaints, including emerging AI-driven solutions such as Noxus.ai designed to improve complaint handling efficiency.

Resolution gap is widening

In 2022, 51 percent of complaints were resolved within three days. By 2025, that figure had fallen to 45.3 percent. More than half of all cases now take longer than three days to resolve, with nearly one million complaints sitting in what can be described as the “resolution gap”.

This gap is not just about time. It represents a different class of work. Cases resolved quickly tend to be simple and rules based. Those that fall beyond three days require investigation, policy interpretation, and multi-system coordination. They are operationally complex and disproportionately expensive.

The data shows that 54.7 percent of complaints now sit in this category, with the majority clustered between three days and eight weeks. This middle ground has expanded steadily since 2022 and now accounts for nearly half of all complaints.

At the same time, 57.9 percent of complaints are upheld. Firms are not only slower. They are also, more often than not, getting the initial decision wrong.

Two insufficient models

The firm-level data reveals a structural split in how banks are responding.

Large incumbents have optimised for speed. Many resolve well over half of complaints within three days. But this speed comes with a high rate of complaint upholding, suggesting a reactive approach where redress is paid out early to close cases quickly.

Digital-first and challenger banks show the opposite pattern. Resolution times are significantly longer, but upheld rates are materially lower. These firms are investigating more thoroughly, which trades speed for accuracy and lowers redress costs.

This creates a clear industry trade-off. One model is fast but expensive. The other is accurate but slow. For financial services firms, this is the key competitive insight. The opportunity is not incremental improvement while still choosing one model or the other, but breaking out of this paradigm entirely.

Why the first wave of automation has stalled

Between 2020 and 2022, complaint handling improved rapidly. Resolution within three days doubled from 26 percent to 51 percent. This was driven by a familiar stack: chatbots for intake, RPA for simple processing, and case management systems for workflow visibility.

That progress has now plateaued and reversed, due to a structural issue. Existing tools are effective for the easy 50 percent of cases. They do not address the remaining workload, which is of a fundamentally different nature.

A complex complaint requires navigating multiple systems, checking firm-specific policies, verifying transactions, reviewing customer history, calculating redress, and producing auditable outputs under Consumer Duty regulations. This is not a single task, but multiple, spanning four to six systems for a single case.

Current tooling does not execute this work. It either deflects it, automates fragments of it, or has to be completed manually.

Consumer Duty raises the bar further

Sector regulation is compounding the challenge. Consumer Duty requires firms to demonstrate consistent, explicable outcomes. This shifts complaint handling from a time-based SLA problem to a decision-quality problem.

Manual processes struggle to meet this standard. Variation between handlers, fragmented data points, and an inconsistent application of policy all increase the likelihood of errors. For fintech operators, this is where AI infrastructure becomes a serious differentiator for financial viability and firm success.

From complaint handling to operational execution

One of the more important findings in the report is that resolution speed varies more within sectors than between them. This suggests the problem is not driven by product complexity, but by operational capability.

In other words, the resolution gap is solvable.

Closing it requires a shift away from front-end automation towards systems that can execute complex workflows end to end. That includes orchestrating actions across systems, applying firm-specific policies consistently, and generating compliant audit trails as part of the process itself.

This is where the next wave of fintech infrastructure is likely to emerge. Not as better interfaces or incremental efficiency tools, but as execution layers that sit across existing systems and handle the operational work directly.

A structural opportunity

Complaint volumes are unlikely to decline. Redress costs are rising, and regulatory demands continue to increase.

The firms that can resolve complaints both quickly and accurately will not just improve customer outcomes. They will unlock a structural cost advantage, through reducing redress, lowering handling costs, and minimising regulatory risk.

The FCA data makes one point clear. The industry has optimised the front half of complaint handling. The remaining gap sits in execution.

Whoever closes that gap first will define the next phase of operational advantage in financial services.