Do these statistics suggest we are hurtling to an AI cliff? No. Instead, they confirm that it’s time to rethink what AI means for business.

Lessons from history are helpful here. When electricity arrived in the late 1800s, factories did the obvious thing: they swapped gas lamps for lightbulbs. The result was brighter, safer workplaces. But the true revolution came later, when factories reorganized around electric motors. Production lines were redesigned and whole industries changed. The lightbulb was the headline, but the re-engineered factory was the real story.

The AI revolution is unfolding in a similar way today. Chatbots are our light bulbs—useful, visible, but shallow. The real transformation will come only when companies change how they work.

Take the recent story about a financial institution that gave its employees a generative AI assistant to draft emails, summarize documents, and do basic analysis. Wall Street rivals rushed to follow. These tools are handy. They shave minutes off routine tasks. But they don’t make a company more competitive. They don’t transform how a business operates.

No wonder results have been underwhelming. In fact, a June McKinsey survey found that 80% of companies report no meaningful bottom-line impact from AI. Businesses see the promise but not the payoff. Why? Because they’re still stuck at the light-bulb stage.

Every technology follows adoption stages. With AI, we’re moving through three. The first was panic: “Organize our data and get me some AI, so we’re not left out.” The second stage – where we are now – is about how we use AI to engage and interact with information: “Give me a chatbot so I can ask questions, perform routine tasks and explore for insights,” The third stage – what’s still to come –  is the real revolution: “Give me enterprise-grade generative AI that will do complex work, integrate seamlessly with my systems, deliver results, and reshape how we operate.” Most companies are still stuck in stages one and two. Few have reached the third. Few are realizing the full potential of AI.

Three takeaways

I spend a lot of time working with companies making the transition from stage two to three. I have started to see similarities among those having early success and those yet to find meaningful ROI.

First, boring is beautiful when it comes to enterprise AI. Find the most mundane tasks, the ones that must get done for the business to operate, but no one wants to do. Eliminate the ones you don’t need and automate the ones you need but can be handled by AI. You’ll see immediate improvement in productivity and give your team more time to innovate, supercharging the business.

Second, it’s critical to define the use cases that matter most to business operations. AI shouldn’t just produce reports more quickly; it should change how deals are sourced, and decisions are made. It shouldn’t just answer questions; it should restructure how procurement is done, and supply chains are managed. In other words, AI allows us to reimagine the routine tasks and underlying systems the drive our most essential operations. It’s the modern version of redesigning the factory floor.  

Third, it’s time to redefine our metrics for success. When organizations don’t have well-defined use cases, they struggle to identify how they quantify success. Most organizations are looking for productivity gains or cost savings, but AI changes how value is created. By focusing on specific use cases, hard and soft KPIs become far easier to define.

We are at an inflection point, but history is clear. The lightbulb dazzled workers in the 1890s, but it was the companies that used electricity to rethink how they operate that unlocked the real potential of this transformative technology. History is repeating itself. It’s a lesson all business leaders should remember. 

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.