?‍? While the workplace has welcomed AI integration with open arms, it may not have unfolded as originally planned. A new MIT study shows that employees are passing over their company’s LLM subscriptions for personal generative AI accounts. The result? A boom in the “shadow AI economy,” according to Fortune.

Generative AI pilots are flunking. Despite enterprises flocking to invest in AI, the return on investment is falling far short for most of them. Only 5% of companies are seeing returns and transformation from adopting AI pilot programs. While generative AI is failing on the enterprise front, it’s enjoying a surge on the employee front as they rely on personal AI tools for day-to-day work. The contrast is stark: 40% of companies have invested in official LLM subscriptions as opposed to over 90% of employees frequently using personal AI chatbots.

What are AI pilot programs doing wrong? This time, AI is not to blame. Inherently, the models are performing well, but it’s the way the companies are integrating them that’s flawed. Workers prefer models like ChatGPT and Claude that are easy to use, adaptable, and flexible. As enterprises custom-build these AI programs to suit their work, the pilots prove difficult to use and integrate with rigid interfaces and a lack of persistent memory. Ultimately, the pilots fail to reach production levels and end up costing more than they return.

The companies have got it all wrong. While workers log into their Copilot and ChatGPT accounts, IT and C-suite teams are mistaken to think their GenAI investments are what their employees need. In reality, the report shows that workers actually prefer to do the high-stakes work themselves and delegate simple tasks like drafting emails and looking up summaries to AI agents. As a result, employees become accustomed to their personal AI tools, experiencing better workflow and productivity, and having less patience for their enterprise’s overly complicated AI models.

To top it off, it turns out everyone’s been getting AI wrong. When AI first arrived, prophecies of AI taking over jobs and pushing humans to the side were everywhere. However, the study reveals that, as of now, AI has replaced only a few jobs and failed to make huge transformations in the workplace. In that regard, we can rest assured that human labor is still in demand.

Tags: