AI does not automatically increase productivity, TNO concludes

AI tools save time, but that time saving does not automatically translate into higher productivity. That is the central finding of TNO following research into the use of artificial intelligence in the workplace. The organisation concludes that the way in which companies incorporate the technology into their processes is decisive for the ultimate effect.

Without deliberate steering, the time gained disappears. Employees fill freed-up hours with other tasks, meetings or administration, without any noticeable increase in total output. According to TNO, that pattern is recognisable across multiple sectors and types of organisations.

The finding calls into question an assumption widely held among companies: that purchasing AI software in itself delivers a productivity gain. TNO argues that technology and process change must go hand in hand in order to achieve tangible results.

What TNO measures and why that is difficult

Measuring productivity in knowledge work is notoriously difficult. TNO points out that common metrics, such as revenue per employee or hours worked, reveal only a limited picture of what AI contributes. An employee who uses a language model to write a report more quickly appears just as productive as before on paper, unless the time freed up can demonstrably be directed towards other value creation.

This phenomenon is well known from earlier waves of technology. When email and office software were introduced in the 1990s, it took years before the productivity gains became visible in macroeconomic figures, a phenomenon economists call the 'productivity paradox'. TNO draws a comparison with that period and emphasises that the current AI wave may involve a similar lag.

Concretely, TNO advises organisations to decide in advance what will happen with the time freed up when deploying AI tools. That requires deliberate process redesign alongside the technology: which tasks fall away, which are expanded, and how do you measure the difference?

How you integrate the tool into your work makes the difference

TNO distinguishes two types of AI use in the workplace. In the first type, employees use a tool for discrete tasks, such as summarising documents or drafting emails, without the broader workflow changing. In the second type, the tool is woven into the daily process, so that colleagues, schedules and responsibilities shift along with it.

Only the second type consistently leads to measurable productivity improvement, according to TNO. That has implications for how companies approach AI implementations. Rolling out a software licence is not enough; guidance is needed on how teams reorganise their work once a tool takes over part of it.

For HR and operations managers, this means that AI adoption more closely resembles an organisational change programme than an IT project. Training focused on operating the tool falls short if the underlying work processes are not reconsidered.

BP reorganises and refocuses on fossil fuels

Separate from the TNO research, there is notable corporate news from the energy sector. BP is undergoing a far-reaching reorganisation: the company is cutting divisions and explicitly shifting its strategy back towards oil and gas. The move marks a reversal of the course the company has followed in recent years, during which renewable energy was central to its public communications.

The reorganisation follows sustained pressure from shareholders who felt that BP's investments in sustainable energy were delivering too little return compared with competitors that stayed closer to their fossil fuel core activities. Exact figures on the scale of the reorganisation, such as the number of jobs to be cut or the size of any write-downs, had not been fully confirmed at the time of publication. Consult the source for the most up-to-date details.

For the energy transition debate in Europe, BP's move is significant. In recent years, the company was regarded as one of the western oil majors that went furthest in formulating climate targets. A strategic shift towards greater fossil fuel production carries symbolic weight for the broader debate on the pace and feasibility of the energy transition.

What this means for AI practice in the Netherlands

TNO's findings align with signals that are also being heard in the Dutch startup and scale-up world. Investors and founders note that AI tools are being widely purchased, but that the expected efficiency gains often fail to materialise in practice or are difficult to demonstrate. TNO now provides a research-based substantiation of that observation.

For policymakers seeking to stimulate AI adoption, for example through subsidies or innovation programmes, the research suggests that support for process change is at least as important as access to the technology itself. Funds that finance only the purchase of tools may therefore achieve only part of the desired effect. That is a practical consideration for programmes such as those run by RVO or regional development agencies that are currently seeking to accelerate AI adoption among SMEs.