Patrick Noël is an engineer, strategy expert, and advocate for frugal AI. He teaches the ethics and practice of AI for sustainable resilience at several leading business and engineering schools. Convinced that technology must remain at the service of people and the planet, he supports organizations (SMEs, mid-sized companies, and large international groups) in their transition to sober and sovereign digital models.
Beyond the hype, how can AI be made a sovereign, viable, and robust tool?
For a long time, artificial intelligence was seen as a futuristic technology, or even a technological demonstration gadget reserved for tech giants. In 2026, the reality is quite different: AI is becoming part of organizations’ value chains. However, while media hype is at its peak, a growing fringe of experts is speaking out to warn of the drift towards a model that could prove as costly as it is ineffective. Faced with the promises of large generalist models, a strategic, viable, and sovereign alternative is emerging: Frugal AI.
The illusion of “Super AI” and economic reality
We are currently experiencing a period of euphoria surrounding Large Language Models (LLMs) and so-called “agentic” AI, which are being presented as the miracle solution to all productivity problems. However, this vision masks a brutal economic reality. The majority of current agentic AI offerings are being sold at a loss, supported by monopoly business models that cannot last indefinitely. Many observers predict that this bubble will burst within the next eighteen months, leading to a drastic correction in costs and the sudden disappearance of many market players.
For businesses, this volatility represents a major strategic risk. Relying on “financial black holes” without a sustainable business model is like building productivity on sand. The question is no longer whether to use AI, but how to do so without jeopardizing financial health.
The quality crisis and the “slop” trap
Beyond the financial aspect, it is the very quality of the responses provided by these models that is at issue. The industry is rushing headlong into a race for raw power, at the expense of accuracy. We have entered the era of “slop”: as models train on content generated by other AIs, the quality of responses deteriorates, leading to a visible loss of accuracy and relevance, including in search engine results.
For a business, this inaccuracy is unacceptable. While it may be tolerated in entertainment applications, it becomes a major obstacle for industrial applications, customer service, or sales tracking, where trust and accuracy are paramount. Caution regarding generalist LLMs is therefore not a barrier to innovation; it is a sign of maturity. It reflects a desire to favor AI that knows how to say “I don’t know” rather than inventing answers, thus placing the value of clean data and industrial precision at the heart of the matter.
Sovereignty and Ethics: Breaking Free from Dependency
A third major pitfall lies in data sovereignty. The current offering is largely based on a form of appropriation of human knowledge by a few hegemonic players, often located in the United States. The extraterritoriality of US law still allows these giants to train for free on global publications, writings, and creations without compensating the authors.
This situation poses a massive ethical and legal problem for companies concerned about their independence. Adopting a frugal approach also means rejecting this dependence. It means choosing to control your value chain by using open source models, hosted locally or on sovereign infrastructures, ensuring that company data remains the exclusive property of the company.
Frugal AI: doing more with less
This is where Frugal AI is a game changer. Unlike an approach that consumes infinite resources, Frugal AI prioritizes efficiency over technological excess.
- It is pragmatic: it tackles repetitive tasks to free up human time, but with controlled energy consumption.
- It is precise: By using Small Language Models (SLM) trained on clean, contextual data, it offers far greater reliability for industry and service professions.
- It is creative: It augments human capabilities in R&D, enabling the calculation of complex scenarios, the formulation of molecules, or the evaluation of natural ecosystems, where computing power is truly needed.
Frugal AI makes technology accessible to SMEs, associations, and large groups concerned about their ecological transition, embodying a sustainable vision: doing better with less, without sacrificing performance.
Towards collective action: the “Frugal AI For Good” initiative
To turn this awareness into concrete action, it is no longer enough to write or theorize. We must equip ourselves with collective tools to measure and guide our transition to responsible AI.
That is why I am proud to announce the upcoming launch of the “Frugal AI For Good” platform.
This unique initiative aims to create an open global ecosystem:
- Open Source and Free: It will be accessible to all, with no barriers to entry.
- Cooperative: A place to share use cases, models, and feedback.
- Auditable: The platform will offer free tools to assess the environmental and social impact of AI projects.
To conclude
Based on rigorous scientific indicators, including planetary boundaries and Kate Raworth’s Donut economic framework, we will finally be able to distinguish real growth drivers from costly gadgets. AI is no longer an option, but it must be chosen wisely. Let’s join the movement for AI that serves people, businesses, and the planet.
If you are interested, you can learn more about Patrick Noel’s approach and become an ambassador for the cooperative platform “Frugal AI For Good” .
