Why the AI accountabilty gap is a real risk for organisations
Can a manager truly own their decisions if these decisions are increasingly influenced by technology, asks Mandy Laurie
Artificial intelligence is becoming increasingly capable of influencing decisions that shape careers, livelihoods and futures. It can identify patterns humans miss, process vast quantities of information and make recommendations with remarkable speed. For many organisations, that promise is irresistible. But there is a question that deserves far more attention than it receives: If a decision affecting someone’s career is influenced by AI, can a leader truly stand behind it if they do not fully understand how it was reached?
Imagine informing an employee that they have not been promoted, have been selected for redundancy, or identified as a low performer. The employee asks why. What if the answer is based, wholly or partly, on the output of a complex system that is difficult to interrogate? The decision may be lawful. It may even be statistically sound. But is it ethical?
Leadership has traditionally required more than communicating decisions. It requires owning them, defending them and being prepared to explain the reasoning behind them. That becomes more difficult when decision-making is increasingly influenced by technology which creates distance between individuals and responsibility. Leaders may find themselves implementing outcomes they trust but cannot fully explain. Organisations may become comfortable relying on recommendations they cannot meaningfully challenge. Accountability becomes fragmented between algorithms, software providers, governance frameworks and human decision-makers, until it is no longer clear who truly owns the outcome.
This challenge also touches on one of the less discussed causes of burnout: values misalignment. Employees can tolerate long hours and difficult work. What they struggle with is a disconnect between an organisation’s stated values and its actual behaviour. If leaders claim that people are at the heart of decision-making, yet cannot explain the rationale for decisions because they originated from a system they do not fully understand, that disconnect risks becoming impossible to ignore.
If leaders are expected to embody organisational values, they must be able to demonstrate those values in practice. That requires confidence not only in an outcome but also in the process that produced it.
This is where the concept of moral injury becomes particularly relevant. Moral injury arises when individuals are required to act in ways that conflict with their values, professional standards or sense of fairness. In an AI-enabled workplace, that tension may arise not because leaders disagree with a decision, but because they are unable to satisfy themselves that it is fair, despite being asked to endorse it.
The critical question is not whether AI can make decision; it clearly can. Nor is it whether AI can outperform humans in analysing data. Increasingly, it already does. The real question is whether generating an outcome is the same as earning the right to make one.
As AI becomes more embedded in workplace processes, organisations will naturally focus on accuracy, efficiency and legal compliance. Yet those may prove to be the easier challenges to solve.
The harder challenge is deciding where accountability truly sits. If a leader cannot explain a decision, cannot interrogate the reasoning behind it and cannot satisfy themselves that it reflects the organisation’s values, on what basis are they taking responsibility for it?
At some point, organisations may have to confront an uncomfortable possibility: the greatest risk presented by AI is not that it makes the wrong decision, but that it allows humans to stop owning difficult ones.
Mandy Laurie is Head of Employment, Burness Paull