The Jevons Paradox offers insight into the inability of Generative AI Slide Tools to give you back your afternoons

Generative AI reduced the average amount of time taken to complete a professional writing task by 37% in a 2023 MIT study with 444 college-educated workers. This is a significant time saving, but how do we account for the Jevons Paradox? The Jevons Paradox indicates that lower-cost slides will not enable you to save hours designing slides; rather, they will enable you to create more slides.

Savings from efficiency gains may lead to an increase in overall consumption.

The Jevons Paradox states that as technological advances allow greater efficiency in producing or using a resource, the total amount of that resource will actually increase instead of decrease. In 1865, William Stanley Jevons published his book on coal, in which he demonstrated that a Watt engine used far less coal for each unit of work it did. British coal consumption grew as a result, because the new lower-priced steam engine provided the opportunity to generate power in places where previously expensive steam engines had been unavailable.

Simply put, with coal as an example of an energy resource, the same logic applies to the time expended on white-collar jobs. Historically, the cost associated with creating slide presentations kept the total number of slides produced to a minimum.

The scale of the rebound depends on how far demand for presentations can stretch

The total amount of rebound will depend on the price elasticity of demand for slide presentations (how much the quantity demanded responds to a change in cost). If you save hours creating slides, whether those saved hours are spent is dependent upon the demand for slide presentations and how much consumers are willing to pay for that time. In situations where the organisation has a fixed requirement for a number of presentations per annum (e.g. when there are four meetings of the Board of Directors, and there is a dependency on the sales cycle containing a defined number of stages), a quicker production tool gives back to the end-user the difference between the amount of time spent developing the presentation and what would have been produced with traditional methods. Once again, if increased capacity is created when prices are reduced, the gain is absorbed by the volume of output that can be created.

The evidence points squarely at the second point: most presentation decks do not contain statutory information; they are created because an individual believes that the meeting is important enough to warrant preparing for, based on a determination of the costs involved in creating that deck. The paradox arises only in circumstances where demand can grow, which is why the situation exists with respect to internal corporate communication.

When production costs decline, a presentation deck will become a viable product, based on the assumption that the need for the deck was there all along but went unmet because of cost.

Opportunity cost serves as the determining factor. If preparing a deck takes the majority of an analyst’s workday, the implicit question posed at each request is whether the time would be more effectively allocated to working elsewhere, which is often the case, and thus the request would not be made. An ai presentation maker can convert a prompt or an uploaded document into a formatted presentation in less than one minute, effectively eliminating this decision-making process for the end-user.

The concept of waste is not simply a linear one. In many cases, some of those additional decks may ultimately be used; however, the previous cost structure prevented them from being produced. The broader point is that the number of hours freed up by the tool will be available for work that was made feasible by the tool itself.

The highest profile illustration of the “rebound” effect is AI’s electricity usage.

The trend in the technology layer can be seen one layer further down. The International Energy Agency indicated that electricity demand from data centres increased by 17% in 2025, significantly outpacing the increase of 3% globally in overall electricity usage, and that centres specialising in AI saw even higher increases. Each successive generation of chip performs greater computational work per watt consumed, thereby resulting in an increase in overall consumption; the reality is that the increased availability of lower-cost computation has produced greater demand for computation in more locations.

Even where the hours are not saved, an increase in measured productivity still occurs.

All of this demonstrates that the tools do not provide a negative experience, and it is useful to differentiate between the two claims that often get run together. Although the time to completion of a presentation has dropped significantly, the total time spent on presentations by the entire organisation is a distinctly different measurement, and efficiency gains cannot be definitively attributed to a specific increase or decrease.

The question to watch is whether the organisation has a natural cap on presentations

Individuals who expect a less busy week as a result of automating this process should focus on the demand side of the equation rather than the tool. Rather than focusing on the absolute speed increase for each presentation, the key question is: Does the organisation have an inherent cap regarding the number of presentations it wishes to produce? Where a cap exists, as it does in some organisations, the time savings realised are tangible and retained; however, in the majority of cases, the absence of a cap will result in an increasing quantity of output rather than fewer hours spent at the point of use.