The share of French companies with more than ten employees using at least one artificial intelligence technology has increased from about 6% in 2023 to 18% in 2025, according to analyses from INSEE. This tripling in two years raises a concrete question: which technological levers produce a measurable effect on digital transformation, and which remain at the level of discourse?
Digital divide between SMEs and large groups: the gaps that hinder digital transformation
The adoption of AI illustrates a structural imbalance. Approximately 58% of large companies with 250 employees or more use at least one AI component. For structures with 10 to 49 employees, this rate drops to about 15%.
This divide is not solely due to budget. Large organizations have dedicated teams (CIO, chief data officer, innovation unit) capable of managing a project from start to finish. SMEs, on the other hand, often accumulate roles: the leader arbitrates alone between an aging ERP and an automation tool whose return on investment remains unclear.
| Criterion | Large companies (250+ employees) | SMEs (10-49 employees) |
|---|---|---|
| AI Adoption (2025) | ~58% | ~15% |
| Dedicated transformation team | Frequent (CIO, CDO) | Rare, often just the leader |
| API integration capacity | High (existing architectures) | Limited (heterogeneous systems) |
| Access to public funding | Controlled process | Underutilized due to lack of knowledge |
The table shows that digital transformation is not just about choosing a tool. Technical integration capacity and the availability of an internal pilot weigh as much as the budget. Players like Digital Innovators specifically support structures that do not have a CIO to structure their technological roadmap.
AI Act and GDPR: the European regulatory framework as a driver of digital maturity
The European regulation on AI (AI Act, Regulation EU 2024/1689) came into effect on August 1, 2024, with a gradual implementation until 2027-2028. Far from being a mere obstacle, this framework requires companies to document their AI usage, assess risks, and train their teams.
This constraint produces a structuring effect. A company that maps its algorithms to comply with the AI Act simultaneously identifies its redundancies, data blind spots, and undocumented processes. Regulatory compliance becomes an involuntary digital maturity audit.
Concrete obligations for technological innovation projects
The AI Act classifies AI systems by risk level. High-risk uses (automated recruitment, credit scoring, assisted medical diagnosis) require complete technical documentation, human oversight, and compliance assessment before market release.
- Prohibited practices (social scoring, subliminal manipulation) have been effective since February 2025, excluding certain use cases from the design phase
- High-risk AI systems must be compliant by 2027, requiring companies to map their tools now
- The GDPR remains the foundation for any processing of personal data feeding an AI model, with penalties reaching up to 4% of global turnover
In contrast, low-risk systems (chatbots, content generation) only require a transparency obligation: to inform the user that they are interacting with an AI. This gradation allows SMEs to integrate automation tools without a disproportionate compliance burden.
Technological innovation and digital transformation: three levers with measurable impact
Rather than a catalog of technologies, three levers produce observable results when deployed methodically.
Automation of repetitive processes
Invoicing, customer follow-up, monthly reporting: these tasks consume considerable time in SMEs. Automation through RPA (Robotic Process Automation) tools or workflows integrated into an ERP/CRM frees up time for higher value-added activities. Automation does not replace jobs; it shifts the effort towards analysis and decision-making.
API integration to break down silos
An isolated tool produces isolated data. API-based architectures allow connecting a CRM to an invoicing tool, an ERP to a logistics platform, a chatbot to an internal knowledge base. The value of a digital tool depends on its ability to communicate with others.
Large companies have mastered this principle for a long time. For smaller structures, low-code integration platforms reduce the technical barrier by allowing the creation of connectors without custom development.
Artificial intelligence applied to customer relations
Next-generation chatbots, powered by language models, handle first-level requests with a rapidly improving resolution rate. The gain is measured in reduced response time and extended availability (evenings, weekends).
AI applied to predictive marketing also allows for segmenting a customer base with a granularity that is inaccessible manually. However, these uses fall under the transparency obligation of the AI Act: the customer must know that an algorithm is involved in personalizing their journey.
Measuring the return on investment of a digital transformation project
A technological project without tracking indicators remains an act of faith. Three metrics provide a reliable reading of progress:
- The internal adoption rate: a tool deployed but ignored by teams has no impact, regardless of its technical sophistication
- The processing time before/after automation: measuring the concrete reduction on a targeted process (invoicing, customer onboarding, reporting)
- The cost of non-compliance avoided: with the AI Act and GDPR, every undocumented AI project exposes the company to significant financial penalties
A digital transformation project that does not measure its results cannot prove its value to management. This requirement for proof remains the best safeguard against technology investments without a future.
The tripling of AI adoption in France between 2023 and 2025 shows that the movement is underway. The difference between companies that derive real benefits from it and those that accumulate software licenses lies in three factors: an identified internal pilot, regulatory compliance integrated from the start, and result indicators defined before deployment.