Mistakes to avoid when deploying AI effectively
- 115K
- Jan 16
- 2 min read
AI has become a standard feature on the roadmaps of most companies. Yet, by 2025, the gap between stated ambitions and actual impact remains significant. According to a BCG study , 74% of companies are still failing to generate tangible value from their AI projects, despite increasing investments. The difficulties encountered are rarely technical; they stem primarily from deployment errors.
In most startups, difficulties in deploying AI can be explained by a few recurring errors, often organizational rather than technical:
Lack of clear prioritization
Launching too many AI projects in parallel disperses teams, slows down execution, and prevents high-potential use cases from reaching critical mass.
Decisions are too centralized.
Managing AI solely from tech teams distances projects from business realities and limits their adoption in daily workflows.
Inappropriate indicators
Focusing on immediate ROI leads to underestimating key factors such as adoption, the robustness of solutions, or their ability to evolve with the product.
Too opportunistic an approach
Treating AI as a series of isolated POCs, rather than as a product asset, creates an accumulation of experiments without scaling up.
Scaling AI without multiplying errors
As AI becomes more focused on execution, deployment errors become more costly. The rise of building blocks like agentic AI enhances its potential, but also its complexity.
The most common mistake is adopting these technologies without a clear link to a priority business problem. Without a framework or adoption roadmap, they generate product debt and slow down time-to-market.
On the contrary, the most mature teams approach these technologies as progressive levers, integrated into a clear product strategy aligned with their level of maturity, which allows them to deploy AI effectively .
115K is La Banque Postale's venture capital fund. It invests in fintech, insurtech, cybersecurity, data, and AI startups across the European Union, from Seed to Series B.
