70% of Employees Use Personal AI Tools Daily Amid Enterprise AI Stagnation

A new MIT Project NANDA report titled “State of AI in Business 2025” reveals a striking trend: while formal enterprise AI adoption lags, a widespread “shadow AI economy” is thriving. The study finds that employees at over 90% of companies regularly use personal chatbots like ChatGPT and Claude for daily tasks, often without IT approval. In contrast, only 40% of organizations have officially purchased large language model (LLM) subscriptions [1].

This divergence has led to what the report terms the “GenAI divide.” Despite $30–40 billion in investments in generative AI, only 5% of firms report transformative returns, while 95% say formal AI initiatives have had no measurable impact on their profit and loss statements. Meanwhile, employees are leveraging personal AI tools to automate workflows and increase productivity in real time [1].

The report attributes this divide to the ease of use and flexibility of consumer-grade AI tools. Employees bypass slow-moving enterprise projects by adopting tools that fit their specific workflows, iterate quickly, and require minimal integration. This creates a feedback loop: the more workers rely on personal AI tools, the more they expect enterprise solutions to match or exceed the same level of adaptability [1].

Shadow AI users often interact with LLMs multiple times daily, outpacing the slow adoption of formal AI projects, which are frequently stuck in pilot phases. This trend is reshaping the perception of AI in the workplace. The report notes that 70% of users prefer AI for drafting emails, and 65% rely on it for basic analysis. However, 90% still prefer humans for “mission-critical” tasks [1].

Project NANDA highlights a broader issue: enterprise AI tools often fail to deliver because they lack memory, adaptability, and learning capabilities. These limitations, rather than regulatory or performance issues, are key barriers to successful AI integration [1].

The study also debunks several myths around AI’s impact. It finds that generative AI is not yet transforming business operations or displacing jobs on a significant scale. Additionally, internal AI development projects fail more frequently than externally sourced solutions [1].

As AI adoption moves forward, the report suggests that organizations must recognize and build upon the informal AI usage already happening across their workforce. Those that fail to adapt risk falling behind in an increasingly AI-driven economy [1].

Source:

[1] “State of AI in Business 2025” — https://fortune.com/2025/08/19/shadow-ai-economy-mit-study-genai-divide-llm-chatbots/