As businesses and public administrations ramp up their experimentation, measuring the return on investment (ROI) of artificial intelligence solutions is becoming a major strategic challenge for digital departments. Beyond simple technical optimisation, the integration of generative and agentic AI technologies calls into question organisations’ capacity for innovation and organisational transformation.
Cigref is today publishing a Memo from its dedicated task force, offering a pragmatic and realistic assessment framework to support decision-makers in this paradigm shift.
Ce rapport est disponible en français (French version).
A technological typology at the service of strategy
To assess value, we must first distinguish between the three technological families that are redefining business processes:
- Traditional AI: focused on accuracy and the massive exploitation of data for predictive maintenance or optimisation.
- Generative AI: adopted for the automation of daily tasks such as document synthesis or information retrieval.
- Agentic AI: a higher level aimed at imitating human reasoning to make decisions and execute actions autonomously.
Horizontal AI vs vertical AI: two approaches to value creation
The report highlights the importance of differentiating approaches according to use:
- Horizontal AI is aimed at all employees. While it saves time, its real value depends on the reallocation of that time to higher value-added tasks or collective intelligence.
- Vertical AI is integrated into core business processes (finance, supply chain, legal). Its value lies in deep co-creation with business lines to generate disruptive innovations.
Defining an evaluation framework adapted to IT transformation
One of the major conclusions of our work is that traditional accounting methods are unsuitable for transformative AI. Justifying a generative AI project as one would for a simple server migration is a dead end.
Cost management: between visibility and hidden costs
The ROI calculation must include a comprehensive view of expenses:
- Specific, easily identifiable costs: engineering, cloud infrastructure, licences, energy consumption and inference.
- Hidden transformation costs: these can represent 30 to 40% of total costs. They include risk management, compliance (particularly with the AI Act), security and, above all, change management.
The ‘portfolio management’ approach
The ROI calculation must include a comprehensive view of expenses:
- Foundations: intelligent and structured management of data, a coherent AI technology stack interconnected with the existing infrastructure, and software enabling agile consumption of functionalities. Here, success is measured by the time-to-market of future use cases rather than immediate financial gain.
- Horizontal Initiatives: measured by the adoption rate and quality of new employee activities.
- Vertical Initiatives: evaluated via Value Driver Trees, which link strategic objectives (e.g., increasing margins) to specific operational levers (e.g., reducing forecasting errors).
The pillars of sustainable and sovereign performance
The successful integration of AI relies on robust digital governance. This is not just a regulatory constraint; it guarantees responsible adoption and controlled security.
Industrialisation via the AI Factory
The transition from the experimental stage (PoC) to sustainable impact is illustrated by the example of setting up a Centre of Excellence or AI Factory. This model provides a unified technical platform, guarantees ethics and disseminates skills across the board.
Human capital and skills: the strategic lever
In a changing environment, skills become a strategic lever for stability. The challenge for CIOs is to anticipate HR needs 10 years down the line, focusing on upskilling so that employees remain in control of the tool. Productivity should not be an end in itself, but a lever for sustainable competitiveness.
Conclusion: towards a strategic vision of AI
Artificial intelligence cannot be treated as a simple cost optimisation tool. It is an investment in the organisation’s capacity for innovation. While bottom-up approaches can help to acculturate teams, only top-down projects, aligned with the overall strategy, can have a structural impact on long-term performance.
To learn more about these methodologies and discover detailed performance indicators (operational, financial and organisational), we invite you to read our full Memo.