By João Pedro Almeida, CEO & Co-Founder at Noxus AI.
AI has earned its reputation by doing the predictable things well: automating repetitive tasks, cutting manual effort, and improving speed and cost. But now many sectors require far more of this important tool, with its next stage lying beyond repetition.
AI is now being pushed to handle complete workflows that require judgement and coordination, and progress has slowed. While the financial services, insurance, and e-commerce industries continue to struggle in moving beyond automating isolated tasks, healthcare, perhaps more than any other sector, is beginning to show what that looks like in practice, with companies like Noxus AI are at the centre of it. . The question is no longer whether AI works, but how it is applied. And more importantly, what other industries can learn from this success.
Where automation has delivered
Across highly regulated industries, automation has followed a largely consistent pattern. Chatbots can manage intake, process automation can handle transactions, and case management systems can direct information to the relevant channels. The success in this area shows in the numbers, as between 2017 and 2020, the number of firms using AI-driven automation doubled, followed by a further 50% increase in sectors like financial services and insurance between 2020 and 2023.
These tools perform particularly well in structured environments, where tasks are governed by clear rules and contained within a single system. In financial services, this has improved areas such as initial complaint handling, customer onboarding, and basic transaction processing. For these use cases, AI has reduced turnaround times and lowered operational costs.
The limits in financial services
Despite these gains, progress has now plateaued in many sectors. The remaining workload in financial services is not defined by simple, repeatable tasks, but by cases that require interpretation, cross-system coordination, and consistent policy application.
This is particularly visible in complaints handling. While AI can assist with first-contact and categorisation, more complex cases still require manual intervention. Resolution times remain slow, and redress costs continue to rise. The issue is not simply volume, but the growing proportion of cases that fall outside the capabilities of current automation.
There is also the strict regulatory route that defines how these processes have to be handled. Under FCA guidelines, firms must resolve complaints within a defined timeframe, yet the first generation of AI tools consistently struggle to manage the full arc of a complex case, from categorisation through investigation to redress, without handing back to a human.
Processes must remain consistent and compliant, limiting the ability of fragmented legacy systems tools to track the progress of intricate cases. AI currently improves the front end, but does not change the whole outcome.
What healthcare is doing differently
Healthcare presents a clearer view of the effects of this, as administrative delays directly affect frontline delivery. Time spent managing patient communication and documents is time taken away from clinical care, an issue that needed a fast solution.
Recent developments show a shift in how AI is being applied. More than one million patients have accessed appointments through NHS digital tools that were not previously available. Digital triage can now be completed in an average of 3.5 minutes, with around 70% of patients seen within a week.
However, the more significant change is happening deeper in the workflow. Rather than focusing solely on intake, newer systems are being designed to manage entire operational processes, including classification, coordination, and compliance.
At CUF, one of Portugal’s largest private healthcare providers, Noxus AI’s Healthcare OS was deployed to handle patient communication, and administrative workflows end to end. The system consolidated incoming data, converted unstructured input into organised frameworks, and automatically directed cases across departments. Only the most complex cases were escalated to human staff.
The results were immediate and measurable. Within two months, over 6,000 tasks were processed automatically, approximately 600 hours of staff time were redirected to higher-value work, and per complaint costs were reduced by 95%, from £1.50 to £0.08.
What financial services can learn
This contrast highlights the broader shift in the evolution of AI that must take place if automation is to make this next leap. Progress has not stalled because the underlying AI technology is incapable, but because, thus far, it has been applied incorrectly: to isolated tasks rather than full processes.
Healthcare is demonstrating that AI can operate across multi-step workflows with consistency, while adhering to strict and variable industry policies. Financial services face similar structural challenges; the difference lies in how the problem is being approached.
Firms across all sectors must move from seeing AI merely as a tool that can assist in completing tasks, to one that can execute them end-to-end, if they want to maintain the competitive advantage adopting first-generation AI gave them. Successes in healthcare show that this is possible, and firms in other sectors must acknowledge this reality, before they are left behind by the potential of this natural next step.
