As large organizations roll out AI tools, the eagerness to adopt new technology needs to be balanced with the risk of using AI to do the wrong things better.
Instead of helping us understand underlying conditions, AI makes it easy to build one-off solutions that treat symptoms and ignore the feedback and learning needed for improvement.
Easier Symptoms, Harder Causes
Imagine a senior manager who’s always busy in back-to-back meetings. They are constantly context-switching, preparing status reports, and putting out fires. They are a bottleneck, causing teams to wait for their decisions.
A reasonable response might be to treat this as a time-management problem: “Let’s use AI to help this manager more quickly write emails, reports, and build presentations.”
This looks like a win because it relieves the manager’s immediate pains. But because AI has made dealing with the symptoms so much faster and easier, nobody thinks to stop to ask why the manager was so overloaded in the first place. By making the symptoms easier to manage, AI pushes the hidden, more complex root causes further out of sight. We skip problem exploration and apply a hastily generated solution to an issue we don’t fully understand.
Lowering the Cost of Bad Ideas
I’ve found that organizations naturally prefer to adopt additive solutions, adding new processes and systems to layer on top of older ones that no longer work. Because AI brings the cost of generating outputs down to virtually nothing, it’s more feasible than ever to throw an AI solution at surface-level problems, all while leaving the root causes unaddressed.
AI will gladly do what you tell it. It will generate entire presentation decks to tell any story stakeholders want to hear, but unless prompted, it will not fix a broken process. At some point, someone needs to step back and ask: “Why do we need hundreds of slide decks in the first place?”
To ensure AI is actually helping, we have to treat it as a tool within a larger and more important continuous improvement process by asking two questions:
- What’s the real problem? To create solutions with meaningful impact, you need to look beyond surface-level issues and ask why the system is behaving this way in the first place. If we can’t identify the root cause, AI will help us build the wrong thing faster.
- How will we know if we are successful? What do we expect to happen after we implement an AI-generated solution? Does the solution relieve a system bottleneck, or are we just optimizing non-value-adding work?
Rewarding a Strong Continuous Improvement Process
When anyone can use AI to create polished solutions for whatever challenge you give it, it’s almost more important to have the ability to decide what not to do.
But if solutions are rewarded regardless of their value, that is exactly what people will deliver. In organizations that reward busyness, activity, and outputs, AI will accelerate velocity in the wrong direction. In cases like these, it becomes a tool for sustaining existing bureaucracy—layering on new solutions, meetings, and processes while ignoring the real problems.
To change this, we have to stop treating solutions as the goal. New solutions are easier than ever to generate thanks to AI, but they are still just another kind of output. Instead, we need to emphasize a strong, continuous-improvement process and culture that values every step of the cycle: identifying the right problem to solve, implementing a thoughtful experiment, gathering feedback, and learning from the results.
