Free First Lesson
What AI Can and Can't Do: A Quick Mental Model
From AI at Work: The Practical Toolkit for Every Professional
Loading interactive lesson preview…
Free sample · No credit card required
From AI at Work: The Practical Toolkit for Every Professional
Free sample · No credit card required
Think of any task you do at work. Two questions tell you whether an LLM is a great fit, a so-so fit, or a bad fit:
1. How language-intensive is it? Does the task involve reading, writing, summarizing, explaining, or transforming text? The more language is the core of the task, the more an LLM can help.
2. How costly is a mistake? If the AI gets it wrong, what happens? A typo in a draft email? No big deal. A wrong number in a financial report? Big deal. The lower the consequence of error, the safer it is to let AI take a swing.
Put these together and you get a simple map of where AI…
Task Fit Quadrant
High-fit tasks (top-left): high language intensity, low consequence of error. Drafting emails, summarizing meeting notes, brainstorming ideas, rewriting a paragraph. Let the AI go first — you review.
Low-fit tasks (bottom-right): low language intensity, high consequence of error. Calculating tax, running a medical diagnosis, controlling machinery. Keep these human.
The gray zone: high language intensity AND high consequence (e.g., drafting a legal contract) or low language intensity AND low consequence (e.g., sorting files by name). Use AI, but keep a human in the loop.
The 80/20 Rule
Most knowledge work is a mix. Find the language-heavy, low-risk slices of a task and delegate those. You keep the judgment; AI does the drafting.