Business Reporter – Management – AI and the productivity illusion

Artificial intelligence is usually sold in one way: it will make you more productive. Board papers talk about efficiency gains. Business cases point to time saved. Vendors promise that teams will do more with fewer resources.

 

The logic seems straightforward. If AI reduces the time it takes to complete a task, costs fall and productivity rises.

 

But there is a harder question that rarely gets asked: does AI actually increase productivity? Or does it simply raise the standard of work we expect from people?

 

For senior leaders, that distinction matters. It affects how investments are judged, how performance is measured and whether promised returns ever appear.

 

In some parts of an organisation, the productivity case is clear. In high-volume, process-driven environments, such as handling claims or answering customer queries, AI can increase throughput in ways that are visible and measurable. If the same team can handle 30 per cent more cases without adding headcount, that is a genuine productivity gain.

 

The picture is less clear in knowledge-based roles, where many AI tools are now being used. When a manager drafts a paper in half the time using AI, the saved time rarely results in twice as many papers. Instead, the paper is more detailed, better structured and supported by more analysis. Expectations quietly rise. What was once acceptable is no longer satisfactory.

 

The output is better. The volume is often the same.

 

Questioning the case for productivity

 

This is where the productivity story becomes blurred.

 

In practice, AI tends to lead to one of three outcomes. Sometimes there is a clear efficiency gain, with more output for the same input. Sometimes the main effect is improved quality, with work becoming sharper, clearer or more reliable but not more frequent. And sometimes AI adds complexity. New systems need oversight. Data needs cleaning and checking. Risks need managing. Controls need tightening. The organisation becomes more capable, but also more complicated.

 

Many investment cases assume the first outcome. Quite often, organisations experience the second. The third is easy to overlook.

 

There is also a human factor. When time is saved, it is rarely removed from the cost base. Instead, it is filled. Sometimes (where management is weak), existing work will expand to fill the time available. But in other cases, it is quality rather than productivity that is affected. Teams use it to go deeper into analysis, spend more time with clients or explore new ideas. That may be a good thing. It can strengthen relationships and improve decisions. But it is not the same as reducing costs or increasing output.

 

In those situations, AI increases capability rather than productivity.

 

Defining appropriate objectives

 

The challenge is that most organisations are not clear about which of those outcomes they are aiming for. They talk about efficiency but behave as though they want better work. They expect cost savings but reward thoroughness and detail. Over time, the benefits of AI are absorbed into higher standards rather than lower costs.

 

This makes it challenging to measure the impact of AI. In manufacturing, productivity is straightforward: units produced per hour. In professional work, it is much harder to define. If a legal team produces contracts with fewer errors but in the same volume, has productivity improved? If software engineers produce more reliable code but release updates at the same pace, is that efficiency or quality control?

 

Without clear measures, organisations fall back on simple indicators such as hours saved or tools deployed. Those figures can look impressive, but they do not always translate into financial results.

 

Governance and leadership

 

There is also the question of oversight. AI systems do not run themselves. They need monitoring. In some sectors, formal controls and clear accountability are required. All of this takes time and effort. A failure to realise this can reduce the potential quality gains from AI. But implementing oversight is a cost which can, in turn, impact productivity gains.

 

None of this means AI lacks value. On the contrary, it can make organisations faster, smarter and more capable. But value does not automatically translate into higher productivity. That depends on the choices made by leadership.

 

If the aim is to improve margins, then processes must be redesigned, and capacity must be managed deliberately. If the aim is to improve quality and strengthen competitive position, then that should be stated clearly and measured accordingly. Confusion between those goals leads to disappointment.

 

Before approving further investment in AI, boards should be able to answer five straightforward questions:

1. Are we trying to reduce costs or improve quality with AI?
2. How are we defining productivity and quality in this context?
3. If we implement AI, how will this affect the way people work, and what are the likely effects on productivity and quality?
4. What additional oversight or control will be required?
5. How will we know whether this has delivered real economic benefitt?

AI undoubtedly increases what organisations can do. It expands the range of what can be analysed, drafted and decided in a given amount of time. But increased capability is not the same as increased productivity.

 

The real issue for senior leaders is not whether AI works. It is whether the organisation is set up to turn that capability into measurable results.