Some Thoughts on the AI Divide

AI is a transformation project, not a software install. Companies that miss this may miss the ROI too.

There’s a great divide in the world of AI right now. On one side is the technology: the large language models, the AI agents, and the mega corporations promoting them. The other side hosts the businesses being sold on the promise of AI. Caught somewhere in the middle are the working people being told they have to use the technology.

The problem is that communication just rolls downhill. The AI evangelists don’t explain how it can provide value to a company—and the companies don’t explain how it provides value to their people/workers. It’s also why AI is so often falling short for so many companies. It’s so often sold to organizations as a straightforward path to efficiency and growth. But the real product is far messier. It should come as no surprise, then, that AI initiatives generate ROI less than 25% of the time.

But AI doesn’t fail because of the technology itself, AI fails because companies are too focused on the technology and not the people using it. In fact, most of the time AI fails at a company is because at least 50% of the people being asked to use it never wanted it in the first place.

I’m not here to debate whether AI has incredible, outsized potential, because I believe that it does. The point I’m trying to make is that it’s no magic bullet, and its success comes down to the rank-and-file people using it every day.

AI works from training, from properly calibrated algorithms, from persistent and sustained corrective use. It’s not at all a plug-and-play technology – it’s something you have to nourish and tinker with. This can be a difficult undertaking for even the largest, most advanced companies. In fact, enterprise companies struggle with it most of all. It’s difficult to perpetually calibrate and adjust such tools when no one is bought in on it in the first place.

I like using the example of a manufacturing company that thought they were doing everything right, and instead inadvertently did the opposite.

Fearing they were falling behind the competition, they built an incredible AI quality-control system for the production line. It was technically exceptional, detecting defects with 95% accuracy, eons better than the old manual method. Yet within six short months, “less than 10% of quality issues were being routed through the AI system” (1). The implementation relied on tech and ignored its people, added extra steps to workflows, and provided little explanation for its decisions. The workers didn’t like or understand it, so they chose to work around it. Another one of the 75% of AI pilots that failed.

AI is not a technology project, it’s a transformation process. Every successful transformation starts with defining a strategy to reach an end state, communicating that strategy to your people, creating a roadmap or plan, and then managing and executing the plan.

I’ve found that if people are involved early, if they’re allowed to give their opinion, if the strategy is explained to them, adoption not only skyrockets, but so does the likelihood of success.

The next thing that increases the probability of achieving ROI, is getting small wins fast. If people can immediately experience success with AI, they’re much more likely to adopt bigger projects.