I came across this video recently, and it nails the problem with so much of the AI hype right now. At the end of the day, most of it is just fancy autocomplete. The video makes the point from a using AI in coding perspective, posing that every extra line of code you add is another liability. The same is true for AI in general. If it’s not used responsibly, within its scope and limits, it quickly becomes more of a burden than a benefit.
Pair that with a new MIT report showing that 95% of enterprise AI pilots fail to deliver real results, and you start to see a pattern. Companies are spending billions on tools that look clever but rarely help the people doing the actual work.
- AI doesn’t replace context. If you don’t involve staff in design and testing, you’ll never bridge the gap between theory and practice.
- AI doesn’t fix bad processes. Automating the wrong workflow just makes bad outputs faster.
- AI isn’t free. Every model, every line of code, every human you try to replace carries cost, risk, and maintenance.
It’s not that AI has no place, I use it often and it’s extremely useful. But adoption only works when the tool is tied to a real problem and shaped by the people who actually use it. Too often, I see AI pitched as a magic fix by people who barely understand the process they’re inserting it into.
A large language model can’t spin a BIM model out of thin air. You still need source inputs like drawings or point clouds. You still need codified rules, templates, and settings. That’s not AI, that’s automation.
And the data backs this up. Sure, MIT’s State of AI in Business 2025 report found that 95% of enterprise AI projects fail to deliver measurable value. But under the surface, a shadow AI economy is thriving. Employees using tools like ChatGPT and Copilot or running self-hosted models at home are solving real problems with far more effectiveness than the official initiatives.
In other words, the real innovation is happening bottom-up, while big-ticket rollouts fail because they’re disconnected from day-to-day workflows.
Don’t chase AI for AI’s sake. Fix the basics, involve the people who know the work, and then use AI where it truly amplifies value. Otherwise, the hype cycle will keep spinning, failure rates will stay at 90–95%, and the people on the ground will keep outpacing corporate “innovation.”
No Comments