The Resolution Gap: Where Automation Has Stalled and AI Can Step In

Businesses across industries have spent much of the past two decades investing heavily in automation, and those efforts have delivered measurable progress. Routine queries are handled more efficiently, transactions increasingly require minimal human input, and some workflows now operate with a high degree of independence.

However, despite these advances, legacy systems continue to dominate much of the operational landscape. This is especially true in sectors like financial services and healthcare, where the complexity of existing infrastructure has slowed deeper transformation. While meaningful gains have been made, significant work remains to modernise core systems and fully realise the potential of automation.

In UK financial services, for example, complaint resolution speed has gone backwards since 2022 despite sustained technology spending. Furthermore, redress costs have hit their highest point since the PPI era. In healthcare, clinical staff are buried in administrative work that consumes hundreds of hours a month and diverts attention from patients.

The problem in both sectors is the same: the complexity of each individual action, not the volume of actions, is what compounds costs and produces poor outcomes. This is the resolution gap, and it will not be closed by the tools that created it.

A trade-off that should not exist

The resolution gap is clearly illustrated in the FCA’s complaints data. Our analysis at Noxus AI of firm-level figures reveals something that has not been widely observed: the UK’s largest banks have split into two distinct operational models. Neither is delivering what customers or regulators should expect.

Barclays, Lloyds, and HSBC resolve complaints quickly. Barclays closes 62% of its banking complaints within three days, well above the industry average. But its upheld rate sits at 78%, meaning three-quarters of resolved cases end with a finding against the firm. These firms have optimised for speed, but the speed is reactive: acknowledge the error, pay the redress, move on. That is an expensive method to prioritise efficiency.

Monzo, Starling, and Capital One sit at the opposite end. Monzo resolves only 25% of complaints within three days, but its upheld rate is only 30%. These firms investigate more thoroughly and get it right more often. However, they do so at the expense of time that customers and regulators are increasingly unwilling to grant.

The industry has, in effect, accepted a binary choice between fast-and-reckless or slow-and-accurate. However, that trade-off is not inevitable. It is a symptom of tools designed for one category of work being stretched to cover another entirely.

The ceiling that no one wants to acknowledge

Healthcare tells a similar story. At CUF, one of Portugal’s largest private healthcare providers, staff were manually processing more than 3,000 patient communications each month. These include appointment requests, prescription queries, clinical administration, and documentation that needed to be digitised, interpreted, and followed through on.

Each of those exchanges carried an obligation of speed and accuracy. Each sat in the same operational territory as the banking complaints above. Hundreds of clinical staff hours a month were consumed by tasks that required attention and judgement, but not clinical expertise.

The cost is felt financially, and in the quality of care patients receive — but the deeper problem is structural. The work exists in a space that lies above what automation handles and below what skilled professionals should spend their time on. That is the ceiling, and most organisations have stopped questioning why it is there.

What the resolution gap actually involves

Across business sectors, and in particular in heavily regulated areas such as financial services and healthcare, where complex case resolution is part of the operating model, the pattern is the same. An agent opens a case, consults the relevant handling policy, and logs into the appropriate platform to verify the underlying transaction or record.

From there, they cross-reference the CRM for the customer or patient’s history, check the document management system for prior correspondence, and apply the organisation’s criteria to determine the outcome.

They calculate what is owed or required, draft the outcome communication, update the case management system, and record every step to satisfy regulatory or compliance requirements. A single case can span four to six systems and require dozens of separate actions, each dependent on the accuracy of the previous one.

The automation tools that firms and providers have invested in over the past decade were not designed for this kind of work. Chatbots handle the front end: triaging queries and sending cases to the appropriate team. Robotic process automation handles the back end: updating records, writing letters, and processing straightforward outcomes where the data is clear. Case management platforms sit in the middle, tracking each case’s status and flagging when deadlines approach.

Each of these tools does its job well. But none of them can execute the full sequence described above, and the gaps between them are where the bottleneck sits. Agents fill those gaps manually, moving between systems, interpreting policy, making judgement calls, and stitching the process together by hand. That manual bridging is not a temporary workaround. It has become the process itself, and it is a key reason why the resolution speed has stalled.

What closing the gap requires

When Noxus was deployed at CUF, the platform automated that full sequence: digitising incoming documentation, interpreting content, routing cases with over 95% accuracy, and redirecting more than 600 staff hours per month back to patient care. In financial services, the same principle holds.

A system capable of reviewing a case file, consulting internal policy, verifying transactional history across platforms, assessing eligibility, calculating the appropriate outcome, drafting compliant correspondence, and documenting every step under a full audit trail is doing something qualitatively different from any chatbot or RPA script. It is completing the work, not supporting the person who completes it manually.

The distinction matters because it changes what becomes possible. The compound cost of nearly two million complaints per year sitting in that expensive middle ground between three days and eight weeks will not shrink through incremental refinement of the same toolset. Nor will clinical staff in overstretched healthcare systems find relief in another workflow layer that still routes the hard work back to them.

The resolution gap will not close through incremental improvements to the same toolset. It requires systems that can execute complex, multi-step work autonomously, within existing infrastructure and under existing policies. Consumer Duty in financial services and rising patient safety expectations in healthcare are both pushing in the same direction: regulators and consumers want faster, more accurate outcomes with a clear audit trail.

The firms and providers that get ahead of this pressure, rather than scrambling to meet it retroactively, will set the standard for good operational performance over the next decade.

---

Noxus is a London-based startup building the first scalable, enterprise-grade AI operating system that enables large organisations to deploy teams of AI co-workers across their operations. Through a no-code platform, businesses can build, manage, and scale AI colleagues that handle repetitive tasks and complex workflows, from claims processing and dispute resolution to knowledge management and customer communications, without building AI infrastructure from scratch.

Founded in 2023 by CEO João Pedro Almeida, CTO Jorge Pessoa, and engineers Gonçalo Ferreira and Miguel Ribeiro, Noxus has secured six-figure enterprise contracts across healthcare, finance, FMCG, and legal sectors in multiple European markets. A leading European healthcare provider has saved 700 hours per month using Noxus AI co-workers, while a global accounting group reduced client churn by 50%.

The company raised $1.5 million in pre-seed funding led by SFC Capital, with participation from Antler, Bynd VC, Caixa Capital, and AltaIR Capital, and has been selected for Google’s UK Growth Programme.

The post The Resolution Gap: Where Automation Has Stalled and AI Can Step In appeared first on Enterprise Times.