Generative artificial intelligence, embodied by tools like ChatGPT, has quickly captured the attention of the general public and businesses. Since its launch in November 2022, ChatGPT has experienced explosive adoption, reaching 400 million weekly users to date. This technology promises to transform various sectors by improving efficiency and creativity. However, despite this enthusiasm, questions remain about its ability to create value at scale for organizations.
Eurogroup Consulting and La Javaness conducted a study between November 2024 and January 2025 on the use of’Generative AI with 46 organizations, including large international groups, mid-cap companies, public bodies and the sectors covered, which were diverse, including industry, energy, services, transport, the public sector as well as banking and financial services.
A study revealing that the adoption of generative AI is still largely experimental. Although all organizations surveyed show a strong interest in this technology, more than 95 % of projects are still in a pilot stage or are in the process of scaling. Investments, sometimes in the millions of euros, are hampered by technological, organizational, cultural, and human obstacles. The majority of initiatives were launched less than a year ago, and organizations are struggling to move from experimentation to large-scale adoption.
A definite craze but still experimental adoption
Generative AI is generating massive interest in organizations, but its integration is still in the exploratory phase. All companies and public organizations participating in our benchmark expressed strong interest in this technology, yet the majority of initiatives remain recent—often launched less than a year ago.
In our sample, more than 95% of projects are still in the pilot stage or are in the process of scaling up. Despite significant investments – sometimes of several million euros – the challenges hindering large-scale deployment go far beyond the budgetary issue alone. Organizations are facing technological, organizational, cultural, and human obstacles, making full and lasting adoption still uncertain.
Differentiated strategies for governing generative AI tools
The framework for accessing and guidance on using generative AI tools varies from one organization to another. The sample studied highlights three strategies for framing generative AI tools.
First, access to consumer tools must be regulated under conditions: many organizations authorize the use of consumer tools like ChatGPT or Copilot, but within a strict framework. Usage charters specify best practices and the risks related to data confidentiality and security.
Another main strategy: accompanying users to train them in the uses of generative AI tools: some organizations combine access to tools with training programs to help users make the most of these technologies while minimizing risks. These training sessions cover topics such as understanding AI, prompt engineering best practices, and responsible resource usage.
Last but not least: developing secure internal tools: To master their data, some organizations develop generative AI solutions in a restricted and secure environment. These platforms serve office and business uses, with specific training for users.
Despite productivity gains at the individual level, value creation at the organizational level remains uncertain.
Generative AI tools enable significant productivity gains at the individual level. Regular users report saving a few hours per week. However, these gains do not always translate into value creation at the organizational level. Leaders perceive generative AI as a strategic lever, but structuring effective integration remains a challenge. Sectors such as IT, marketing, risk and fraud management, and legal professions show more advanced adoption, integrating generative AI into their daily practices. Specific use cases have already demonstrated how AI can automate certain tasks, such as code auto-completion (the most widespread), incident management, code documentation, or log analysis, thereby reducing response times and increasing team efficiency.
The Four Major Challenges for Deploying Generative AI
The sample organizations widely share four fundamental challenges to consider for the large-scale deployment of generative AI solutions. The goal is to ensure both effective and sustainable adoption, as well as a balanced strategy combining innovation, performance, and responsibility, allowing to fully leverage the opportunities of this technology while controlling its costs and impacts.
- Profitability and risk-taking, finding the right balance The cost of implementing generative AI solutions is high, and immediate profitability is rarely guaranteed. Organizations must balance cost, performance, and return on investment, adopting a pragmatic approach to maximize long-term gains.
- Data: moving beyond data governance Data quality, access, and security are crucial. Generative AI models require massive volumes of data, which are often scattered and of uneven quality. Effective data governance is essential to ensure relevant and regulation-compliant results.
- Technological dependence and digital sovereignty Dependence on large technology companies, primarily American, poses security and data control risks. Less than a quarter of organizations favor the use of a local sovereign ecosystem, despite initiatives to promote sovereign AI.
- IEnvironmental impact, an underestimated challenge Generative AI is energy-intensive, with a significant carbon footprint. Controlling this impact requires model optimization, the use of ecological infrastructure, and responsible AI usage.
Generative AI impresses by the speed of its spread, its adoption on an individual scale and the widespread productivity gains it enables. While its use is still experimental at scale, its deployment in organizations should see accelerated growth, even faster than that of traditional AI. However, to maximize its impact and achieve performance gains at scale, it is essential that organizations target specific business uses, where the added value is greatest, such as IT, customer and user service, marketing or financial analysis. But for these gains to translate into tangible return on investment, it will be necessary to carry out business transformations, precisely measure the performance of solutions and adjust their use according to the results observed. Furthermore, the challenges of digital sovereignty and technological dependence cannot be ignored. The question is no longer whether it will prevail, but how organizations can adopt it sustainably and strategically.
