France remains on the European podium for declared AI maturity and in the basement for impact measurement. This gap is not a temporary paradox: it is the signature of an investment cycle driven by the fear of obsolescence rather than capital allocation. ActuIA already documented this pattern in 2024, with the French industry strategically investing in AI for marginal ROI, then noting that only 10% of companies perceived a very significant financial impact. Two years later, the SME base shifts from 15% to 55%, but 80% of organizations remain blind to returns. At this stage, the budgetary risk is no longer inaction, it is the silent renewal of pilots without thresholds. Investment committees that arbitrate their 2026 AI budgets this summer would be wise to refuse any pilot renewal beyond twelve months without a euro figure of the targeted value and without an automatic stop clause at 18 months. Budgetary pressure will achieve what method has not.
49% of French organizations plateau at the pilot stage, are on pause, or have not started, and over 80% measure no tangible financial impact from their AI - while France ranks second in Europe for declared AI maturity, at 36%, according to the Enterprise AI Adoption Impact Index published by Infor on April 22, 2026 on a YouGov panel of 1,000 decision-makers surveyed between late March and early April in five sectors and four countries. France leads the UK (27%) and the US (21%) but remains behind Germany (38%). For the 2026 AI budgets under review, the lack of ROI measurement is no longer a methodological gap: it's a budgetary risk.
Unmeasured ROI, a Symptom Before Being a Cause
The financial diagnosis converges among sources. Bpifrance Conseil and Siparex, in their April 2026 white paper published on the francenum.gouv.fr portal, recall a 2025 McKinsey study indicating that over 80% of organizations that invested in generative AI saw no tangible financial impact, due to a lack of prioritization of the right topics and real adoption by users. On the declared barriers side, the Infor study places security, sovereignty, and compliance first at 36%, ahead of lack of internal talent at 25% and unclear ROI at 23%. The declared barriers, however, intersect with technical and cultural dimensions that ROI measurement captures only imperfectly. French prudence, long read as a hindrance, is beginning to align with the AI Act, GDPR, and NIS2 directive, three frameworks whose operational maturity in France is becoming a structural asset.
Dominant Uses and Return Window
The portfolio of deployed use cases remains focused on simple tasks. Several sectoral barometers published in early 2026 document a predominance of summarization, translation, and content generation functions in French deployments, and a much more limited presence of business value uses - sales forecasting or inventory planning. The return delay reflects this concentration. According to the Deloitte AI ROI report published in October 2025, built on 1,800 experts in fourteen countries, the satisfactory return on investment of a typical AI use case materializes over two to four years, significantly more than the seven to twelve months expected from an IT investment, and only 6% of respondents report a return in less than a year. On agent AI, only 10% report significant ROI. On the operational use side, nearly half (49%) of AI-generated results require manual review by a business expert, which weighs on the real productivity calculation. In the manufacturing sector, the signal partially reverses: according to Rockwell Automation, only 2% of French manufacturers identified AI as their best ROI lever in 2025, a low threshold that opens significant sectoral catch-up potential.
A Very French Actor Map
The French enterprise AI market presents a supplier selection anomaly. According to a Spendesk study reported by LeMagIT on March 3, 2026 covering 2,500 European companies, Dust represented 61% of the total AI spending tracked on the Spendesk platform in France, ahead of OpenAI (26%) and Anthropic (11%), while OpenAI dominates elsewhere in Europe. These data are transactional, from a closed perimeter of AI tools tracked on the platform, and not from a declarative survey, so they should be taken with caution. The French publisher claimed €6 million ARR in 2025. However, the finding should be read with the panel's caveat: 1,700 of the 2,500 companies surveyed are French, which may bias European representativeness. On the SME dynamic side, Bpifrance Conseil and Siparex report that the share of SMEs that have started an AI project has increased from 15% to 55% in two years. ActuIA already documented this exploration dynamic by noting that French SMEs and mid-caps remained a territory to explore according to Bpifrance - the shift of the SME base above 50% in two years shifts the center of gravity of French AI adoption.
The gap between declared adoption and measured ROI is measured in years, not months. According to Deloitte (1,800 AI experts, 14 countries, October 2025), the return on investment of a typical AI use case is between two and four years, about two to five times longer than a typical IT investment according to the same Deloitte ranges.