DeepSeek and Jevons Paradox: does a lower price truly lead to mass AI adoption?

With the launch of DeepSeek, an open-source language model designed to make AI more accessible, the discussion surrounding the Jevons Paradox resurfaces. This paradox states that when the 'fuel' for a technology becomes cheaper, its use actually grows. But is that truly the crux of AI adoption within businesses?

The Jevons Paradox is named after the British economist William Stanley Jevons. In the 19th century, he observed that as coal became more efficiently used, its consumption did not decrease, but rather increased. In AI terms: now that DeepSeek is an open-source model and the operating costs (e.g., computing power or licenses) are lower, you might expect more organisations to want to use AI. A common metaphor for the Jevons Paradox is that of the car. When petrol prices fall, people drive more. So, wouldn't AI also be massively deployed if the 'fuel' in the form of licensing or usage costs becomes lower?

"Cost is not the biggest barrier"

In practice, however, we see that many companies do not disregard AI because of the cost, but for other reasons. While budget always plays a role, two crucial factors often carry more weight:

1. Data QualityAI is only as good as the data it processes. If the input is incomplete, unstructured, or of low quality, you won't get useful insights. Companies that don't have their data management in order quickly get stuck in implementation and shy away from AI.
2. Clear Business QuestionA common misconception is that AI automatically leads to innovation. But organisations that aren't clear about which problem AI needs to solve will see little return on their investments. Without clear objectives, AI is nothing more than a 'fancy gadget' that adds no real value. It's like buying a car but having no idea where you want to drive.

Read more about AI agents: AI Agents: what you need to know

How can DeepSeek contribute to AI adoption?

DeepSeek is interesting because it:

• Is open source and can therefore be further developed by a broad community.
• Offers lower usage costs, which for some organisations might just be the push needed to experiment with it.

Nevertheless, a new, cheaper technology doesn't automatically mean all organisations are queuing up to embrace it. Especially not if they are still struggling with fundamental issues concerning data and strategy.

Tips for organisations wanting to start with AI

1. Start with a concrete business problemClearly formulate which challenge you want to solve. Think about optimising customer service, predicting machine maintenance, or personalising marketing campaigns.
2. Ensure the right dataSet up your data infrastructure in such a way that you can guarantee quality and consistency. This sometimes means investing in better data pipelines, data warehouses, and good governance.
3. Choose the right AI solutionDetermine whether open-source models like DeepSeek are suitable. Sometimes a standard SaaS solution works better, sometimes you need customisation. It depends on your objectives, time, budget, and available expertise.
4. Invest in knowledge and cultureAI is not simply a matter of 'installing a model'. It requires employee training, adjustments to business processes, and a culture that is open to data-driven decision-making.

The Jevons Paradox focuses on costs and usage but ignores that AI adoption is far more complex. Costs may fall, but if your team doesn't know what AI can do for them or if the data environment isn't in order, the organisation won't suddenly embrace AI. Without a clear destination, cheap petrol is of little use, just as lower AI costs contribute little if you don't know for which problem you want to use AI.

Read more about AI agents: The end of SaaS? How AI agents are shaping the future of software

DeepSeek, as an open-source language model, can lower the barrier to AI. But lower costs alone are rarely the decisive factor for mass AI adoption. Companies that successfully deploy AI have their data in order and know precisely which business problem they want to tackle.

In short: if you truly want to benefit from AI (and models like DeepSeek), first check whether your organisation is ready in terms of data, strategy, and knowledge. Only then will the path be cleared for the 'fuel savings' of AI, and you can gain a significant advantage in the next digital wave.

• Understanding AI token costs