AI access, costs and lock-in are already a business risk
In brief
According to research firm Tercio, AI has created in a few months the dependencies that cloud computing took ten years to build. The report confirms access cuts, overspent budgets and a large-model market held by a few vendors. A digital-resilience barometer notes that in 2026 several leading models had access limited or suspended by US administrative decision.
Source : LeMagIT - « IA : la dépendance en accéléré (étude) »
What it changes for an SME
If you rely on one or two AI providers, an access cut or poorly tracked token billing can stop a workflow or blow the budget. According to Tercio, plans to switch models often cover a technical outage, rarely a provider shutting the service. Dataiku reports that only 21% of CIOs have full real-time visibility of their agent costs by team or use case. A 2025 MIT study estimated that 95% of organisations had not obtained measurable benefits from generative AI, despite 30 to 40 billion dollars invested.
What you can do
- List the processes that would stop if your AI provider cut access.
- Ask for AI costs broken down by use case and set a spending cap.
- Test a live switch to a second model, including an open-weight one.
Let's talk about the time you could save.
In 20 minutes, we find where AI could save you time or money. Free, no commitment.