Investors love the line: “execution is everything.” And it’s true, but incomplete. I’ve been talking to myself about the relationship between two different asymmetries: information asymmetry (who knows what) and execution asymmetry (who can actually do what).
AI is shrinking information asymmetry
Execution asymmetry remains in place and perhaps even expanding as execution-oriented people begin using AI. I’m thinking about this split because it helps me think up new business models.
Defining the asymmetries
Information asymmetry: when one party has substantially more or better information than another. Classic examples are a mechanic and a car owner, or a patient and a specialist — you can’t easily verify what’s happening under the hood or inside the body.
Execution asymmetry: when one party can execute a task others cannot, either because of specialized training, practice, resources, or institutional position. A neurosurgeon exemplifies this: access to information doesn’t substitute for the years of training and the practiced skill required to perform surgery.
A weak idea well-executed routinely outperforms a strong idea that never ships. Being able to execute requires many concrete, hard-to-copy things such as connecting with the right people, building culture, managing complexity, and cultivating a unique pattern recognition that scales. This combination of abilities does not transfer instantly with information.
AI changes information asymmetry by dramatically lowering the cost of acquiring knowledge. For many domains, someone can go from ignorance to useful competence relatively easily. That levels the playing field on the informational side:
Customers can hold providers accountable.
Founders can start businesses that previously might not have made economic sense. Now they do because the input cost is low (AI tokens) and there is less or no coordination cost involved (co-workers).
Investors can screen opportunities more efficiently.
Execution is still king
Knowing how to do something and being able to do it at scale remain discrete. In many cases AI augments human operators, making execution quicker or cheaper. Yet, we still need human intuition, know-how and pattern recognition.
Some execution asymmetries are institutional and long-lived. It takes time to build trust whether with humans or objects.
Where execution relies on repetitive cognitive tasks, it appears easy to see how AI-enablement will dictate the future.
The most interesting edges will be people capable of ingesting loads of information AND with superior execution abilities.
