I finally have the words to describe the feeling of being overextended when scaling work with AI: 'Cognitive Debt.'
It’s that creeping sensation of losing the plot. You're moving fast, hitting milestones, but you've stopped building a deep, shared understanding of why the work is moving in a certain direction.
In a recent talk titled 'Understanding is the New Bottleneck' Notion's Geoffrey Litt highlighted this concept (originally coined by researcher Margaret-Anne Storey). He argues that as AI agents write code at unprecedented speeds, the human understanding of that code becomes our biggest vulnerability.
You don't need to be a software engineer to feel this. I don't write code, but I see cognitive debt everywhere.
Whether you are designing a new customer journey, running complex data analysis, or building out a strategic product roadmap, letting AI take the wheel without understanding the engine leaves you stranded. We don't just need to understand AI's outputs to verify that they are correct. We need to understand them so we can actually participate.
When you maintain a rich conceptual understanding of a project, you retain the ability to fluidly pivot, recombine ideas, and figure out the next strategic leap. If you let the AI accumulate all the context, you lose your ability to lead the work.
The fix is surprisingly simple: demand transparency from your tools. Treat your AI like a collaborator and ask it to explain its steps, its logic, and its decisions. Stay in the loop.