techbullion.com

The More We Automate, the more Human Judgment Matters

By Anamta Shehzadi·Published November 20, 2026·Artificial intelligence

The More We Automate, the more Human Judgment Matters

As artificial intelligence reshapes financial services and business operations, an overlooked question emerges: If technology performs the work through which professionals once gained experience, how will tomorrow’s experts develop the judgment needed to oversee it?

By Dr. Sushant Rajput

Imagine two banking professionals beginning their careers a decade apart.

The first spends months reconciling transactions, investigating settlement failures and examining discrepancies that do not fit neatly into established procedures. The work is often repetitive. But through difficult cases, corrected mistakes and conversations with experienced colleagues, the analyst gradually learns something that cannot be captured entirely in a process manual: why things happen, not merely what to do when they happen.

The second joins an operation in which intelligent systems already match transactions, classify exceptions, and recommend corrective actions. Routine activities that once occupied hours now require considerably less manual intervention.

The second professional may become productive sooner. But when an unfamiliar problem emerges, will that professional possess the contextual understanding needed to challenge the system’s recommendation?

These are hypothetical scenarios, not accounts of institutions. They illustrate a question that deserves attention as artificial intelligence changes how organizations operate.

What happens when we automate the very experiences through which people once learned to exercise judgment?

This is not an argument against automation. It is a question about what organizations must preserve, redesign or deliberately develop as technology takes on more work.

For years, businesses have invested in technology to improve efficiency, consistency and operational performance. Those objectives remain important. Yet an intelligent system and an intelligent organization are not necessarily the same thing. A system can become more capable while the people responsible for overseeing it face new demands on their expertise.

The next challenge in business transformation may therefore be less about whether technology can perform a task and more about whether professionals can understand, question and take responsibility for its outcomes.

The Automation Story Has Another Side

Artificial intelligence is increasingly part of financial services operations, supporting activities such as document interpretation, anomaly detection, information extraction, workflow prioritization and decision support.

The Bank of England and Financial Conduct Authority’s Artificial Intelligence in UK Financial Services 2026 survey reported that 86% of responding firms were using AI, compared with 75% in the 2024 survey. The regulators cautioned that differences in survey respondents and scope mean the figures should not be interpreted as a directly comparable measure of adoption growth.

The 2026 survey also reported that operations and IT accounted for approximately 31% of identified AI systems.

For financial institutions managing large transaction volumes, complex information and demanding control requirements, the potential benefits are substantial.

Automation can reduce repetitive processing, support more consistent execution and allow employees to concentrate on activities requiring interpretation or investigation.

But one distinction is often overlooked.

Improving the capability of a process does not automatically improve the capability of the people responsible for it.

An employee may oversee more transactions without understanding every dependency behind them. A recommendation may be generated faster without becoming easier to evaluate. A workflow may require fewer interventions while becoming more difficult to investigate when something unexpected occurs.

These are not inevitable consequences of automation. They are risks and design considerations that organizations can address through appropriate training, governance and operating models.

The real question is whether human capability is evolving as deliberately as technological capability.

The Experience Hidden Inside Routine Work

Consider a hypothetical reconciliation process in an investment bank.

An automated application compares transaction records across two systems. Most transactions match successfully. One item, however, remains unresolved because the settlement amount differs from the expected amount.

The application identifies the discrepancy, assigns it to a predefined exception category, and recommends a standard investigation. A junior analyst reviews the recommendation. The classification appears reasonable, the supporting information is available, and the suggested action follows the established workflow.

But suppose the discrepancy is connected to an upstream booking issue affecting several related transactions. Resolving the individual exception without recognizing the broader pattern could leave the underlying cause unaddressed.

The automated recommendation may have been appropriate for the information and rules available to the system. The concern is not necessarily technological failure.

It is whether the analyst understands enough about the transaction lifecycle to recognize when a seemingly isolated issue deserves wider investigation. An experienced professional might examine related transactions, review upstream activity or consult another team before deciding how to proceed.

That response would not necessarily come from intuition alone. It could reflect knowledge developed through earlier investigations, exposure to unusual cases and an understanding of how different operational activities interact.

Now consider what happens when automation removes much of that earlier exposure. If new analysts rarely investigate straightforward exceptions, how will they develop the foundation needed to understand complex ones?

The answer need not be a return to manual processing. Structured simulations, supervised investigations, process walkthroughs and targeted learning can provide alternative opportunities. But those opportunities must be intentionally designed.

When organizations automate a task, they should ask whether they are eliminating unnecessary effort, removing a learning opportunity, or doing both. That distinction could become increasingly important in professions where practical judgment depends on understanding how processes behave beyond their standard operating conditions.

When Technology Changes How We Think

The relationship between artificial intelligence and critical thinking is attracting growing research interest.

A 2025 study by researchers from Microsoft Research and Carnegie Mellon University examined 936 examples of generative AI use reported by 319 knowledge workers.

The researchers found that greater confidence in generative AI was associated with less reported engagement in critical thinking. They also observed that AI-assisted work could shift critical-thinking activity towards verification, integration and oversight.

These findings require qualification. The research relied on self-reported experiences and does not establish that using AI causes a deterioration in critical-thinking ability or long-term professional expertise.

Nevertheless, it raises a relevant question about how employees engage with technology-generated recommendations.

Consider two professionals using the same AI-enabled application.

One accepts the output because the system has produced it.

The other examines the evidence, considers whether relevant information may be missing and determines whether the recommendation is appropriate for the circumstances.

Both are using technology. But their approach to the decision differs.

The distinction is not between people who embrace AI and those who resist it.

It is between using technology as a source of assistance and treating its output as a substitute for informed evaluation.

This is particularly important when systems operate in environments where decisions may involve financial, operational or regulatory consequences.

Human judgment does not mean questioning every automated decision. It means knowing which decisions deserve scrutiny, what evidence to examine and when intervention is necessary.

The Illusion of Human Oversight

As organizations introduce AI into operational workflows, governance arrangements often include human review, approval and escalation.

These controls can be valuable. However, their effectiveness depends on more than the presence of an employee at a particular stage of the process.

Consider a hypothetical client onboarding case.

An AI-enabled application reviews customer documentation, identifies inconsistencies, and recommends whether additional due diligence is required.

A reviewer receives the recommendation and approves the next step.

On paper, the process includes human oversight.

But what if the reviewer does not understand the factors behind the recommendation? What if important information is missing? What if the reviewer lacks sufficient time, authority or expertise to challenge the outcome?

The approval may satisfy a procedural requirement without necessarily providing the intended level of scrutiny. This does not mean automated recommendations are inherently unreliable. Nor does it suggest that every reviewer must understand the technical architecture of an AI model. It means the organization must establish what knowledge is necessary for meaningful oversight and ensure that the people responsible possess it.

The Bank of England and FCA’s 2026 survey provides relevant context. Among respondents mentioning, how they obtained sufficient understanding of high-materiality AI systems, 77% reported relying on deep knowledge for internally developed systems, compared with 24% for third-party systems.

The figures do not establish inadequate governance. They illustrate how approaches to understanding AI can differ depending on whether technology is developed internally or obtained from external providers.

For organizations operating complex technology ecosystems, the implications extend beyond model design.

They involve accountability, process knowledge, vendor governance and the ability to investigate unexpected outcomes.

A reviewer does not necessarily need to explain every technical calculation performed by a system. But the reviewer should understand the decision being made, the relevant business context and the circumstances that require further examination. Human approval is not automatically evidence of human judgment.

The Next Transformation Challenge Is Developing Expertise

The World Economic Forum’s Future of Jobs Report 2025 estimated that an average of 39% of workers’ existing skill sets would change or become outdated between 2025 and 2030, based on expectations reported by surveyed employers. The report also identified analytical thinking as the most sought-after core skill, with approximately seven in ten surveyed employers considering it essential.

These findings highlight an important challenge for organizations investing in automation.

Technology may change the activities employees perform, but it does not eliminate the need for people capable of interpreting information, solving unfamiliar problems and exercising responsibility. Indeed, in some roles, automation may shift human attention away from routine execution and towards exceptions, ambiguity and consequential decisions.

That shift creates opportunities for more meaningful work. It also changes the requirements for professional development. A new employee cannot be expected to acquire deep operational understanding simply because a sophisticated system makes the work appear easier.

And an experienced employee cannot assume that knowledge developed under an earlier operating model will remain sufficient without adaptation. Both may need to develop new capabilities. For business leaders, this suggests that workforce development should be considered during transformation design, not only after technology implementation.

If the operating model changes what employees do, the learning model may need to change how they acquire expertise.

From Process Training to Judgment Development

Traditional operational training often concentrates on procedures, system functionality, service standards and defined escalation paths. These remain essential. But an AI-enabled workplace may also require deliberate opportunities to interpret incomplete information, evaluate recommendations and investigate unusual outcomes.

In financial reconciliations, for example, analysts could be presented with simulated exceptions where the obvious resolution is not necessarily the most appropriate one. They would be asked to examine related activity, identify potential root causes and explain their reasoning.

In client lifecycle management, employees could review hypothetical onboarding cases involving conflicting documentation or incomplete information. Rather than simply following an automated recommendation, they would assess what additional verification might be necessary.

In operational risk and control functions, professionals could examine AI-generated incident summaries and identify which conclusions are supported by evidence and which require further investigation. These are proposed training approaches, not claims of demonstrated effectiveness. Their value would need to be assessed through appropriate testing and performance evaluation.

The underlying principle is that employees should learn to understand the business decisions represented by system outputs, not merely the steps required to process them.

Experienced professionals have an important role in this transition.

Their contribution can extend beyond resolving difficult cases personally. They can explain the reasoning behind decisions, help design realistic learning scenarios and mentor colleagues who are developing independent judgment. This creates an opportunity to move from training people primarily to execute processes towards developing their ability to understand and oversee them.

It also suggests a different way of thinking about expertise.

Experience should not be measured only by the number of transactions processed or the years spent in a role. Its value also lies in the understanding developed through those experiences and the ability to apply that understanding in unfamiliar situations.

Measuring What Automation Leaves Behind

Imagine two financial institutions implementing comparable automation capabilities. Both reduce manual intervention, improve processing efficiency and meet their intended operational targets.

The first evaluates success primarily through productivity, service performance and technology adoption. The second considers those measures alongside workforce capability. It examines whether employees understand critical process dependencies, can investigate unusual outcomes and possess the knowledge required to challenge automated recommendations.

The comparison is hypothetical, and neither approach guarantees a particular business outcome. However, it illustrates why transformation may need to be assessed through a broader set of considerations.

Processing efficiency, accuracy, cost and control performance remain important. They can be complemented by measures of operational understanding, quality of exception handling, effectiveness of escalation and readiness to manage unfamiliar situations.

Such measures would need to reflect the specific risks and responsibilities of each process. They should not be treated as a universal framework or a replacement for established governance requirements. The objective is not to slow automation or preserve manual work unnecessarily.

It is to ensure that productivity improvements are accompanied by the capabilities required to sustain reliable operations. A transformation that reduces repetitive effort while strengthening professional judgment would offer a valuable combination of technological and human capability.

The Question Leaders Cannot Afford to Overlook

AI offers organizations opportunities to redesign processes, improve access to information and redirect human effort towards activities requiring greater interpretation and problem-solving. These opportunities deserve serious attention.

But as technology becomes more capable, organizations must also reconsider how professional expertise is acquired, retained and transferred. When routine tasks are automated, learning opportunities may need to be redesigned. When systems generate recommendations, employees need the ability to evaluate them.

When decisions involve both human and technological contributions, accountability must remain clear. And when operating models evolve, leaders must consider not only what technology can perform, but what people must understand to oversee it responsibly.

The challenge is not to choose between automation and human judgment. It is to design organizations in which each strengthens the other. Perhaps the next important question in business transformation is not simply, “How much more can we automate?”

It is, “What must our people become capable of as we do?”

Because the future of work will not be defined only by what machines learn to do. It will also be defined by what humans must never stop learning to understand.

About the Author

Dr. Sushant Rajput is Vice President at eClerx Services Ltd., with more than two decades of experience across financial services, business operations, transformation and business development. He is also an author, TEDx speaker, educator and mentor whose work explores professional development, leadership and organizational change.

Sources and References

  1. Bank of England and Financial Conduct Authority (2026). Artificial Intelligence in UK Financial Services 2026. Published 26 October 2026.
    https://www.bankofengland.co.uk/report/2026/artificial-intelligence-in-uk-financial-services-2026
  2. Microsoft Research and Carnegie Mellon University (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers.
    https://www.microsoft.com/en-us/research/publication/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers/
  3. World Economic Forum (2025). The Future of Jobs Report 2025.
    https://www.weforum.org/publications/the-future-of-jobs-report-2025/