top of page
Search

95% of GenAI projects fail. And yet, your colleagues are already making progress… on their own.

By 2025, generative artificial intelligence (GenAI) will no longer be a topic of research. It will be a daily reality in businesses. But this adoption will not be a smooth process.


This is demonstrated by the latest MIT study: “The GenAI Divide: State of AI in Business 2025” (NANDA initiative) . Based on 300 AI initiatives , 52 organizations and more than 150 executives interviewed , it provides an unfiltered look at the strengths and weaknesses of AI adoption in business.


What the MIT study reveals


Some key figures:

  • 80% of companies have already tested GenAI tools, with a real impact on individual productivity

  • But only 5% of projects reach the stage of large-scale deployment

  • 50% of budgets are allocated to marketing, while the real return on investment lies hidden in the back office (finance, support, procurement, etc.)


The paradox of “shadow AI”


One of the most striking observations concerns the gap between the official strategy and the reality on the ground.

  • 90% of employees already use AI with their personal accounts (ChatGPT, Claude, Copilot, etc.)

  • However, only 40% of companies report having officially deployed these tools

In short: while initiatives driven by management are stagnating, "out-of-the-box" usage is exploding. This phenomenon, called shadow AI , raises two major issues:

  1. Increased security risks (data leaks, lack of control)

  2. A strategic disconnect: AI is progressing, but outside the company's vision.


Five myths debunked by MIT


The study dispels several misconceptions about AI in business:


  1. “AI will replace most jobs” → Layoffs remain marginal

  2. “AI is already transforming businesses” → Only 5% of workflows are truly integrated

  3. “Large companies are slow” → 90% have already seriously explored purchasing tools

  4. “The obstacle is technological or legal” → The real stumbling block is integration into workflows

  5. “High-performing companies develop their own tools” → Internal projects fail twice as often


What this means for leaders


The study's message is clear:


The question is no longer “should we launch an AI project?”

The real question is: “how do we channel and structure what already exists?”

Without a comprehensive strategy, AI will take hold despite your efforts. With a structured approach, it becomes a lever for collective performance, capable of truly transforming your processes and results.


The SoSharp conviction


At SoSharp , we support leaders and their teams to:

  • Detect existing uses (including shadow AI).

  • To structure a realistic and secure roadmap.

  • Make AI a collective reflex, integrated into workflows and corporate culture.


Our mission: to transform the dispersed potential of shadow AI into widespread adoption, creating sustainable value.


 
 
 

Comments

bottom of page