The recent report from MIT that 95% of AI pilot projects fail has caused some concern.
Contrary to what some may think, it's not because of the capabilities of the models themselves, but about the implementation.
Here are some of the reasons for the poor performance:
1. Companies are trying to adapt AI to their current processes and procedures. The problem with this bureaucratic approach is that AI does not care about office politics, and it is more important to adapt your processes to its strengths.
2. Build vs. buy. Many companies are trying to build their own AI tools, rather than partner with an experienced vendor. Most companies don't have AI expertise on staff, and the budget to do AI from scratch. Also, any DIY attempts will probably involve open models, and these are not yet as powerful as proprietary LLMs.
3. Most of the current processes involve sales and marketing. But AI can make some of its most productive gains if applied to back-end processes.
The ultimate key, just like any digital transformation, is to practice Strategic Simplicity®, and focus on a few, critical functions and bottlenecks. Pick one at a time, as you first build your AI skills. As the expert in Strategic Simplicity®, this is how I advise my clients.
© 2025 Praveen Puri
