Give two people the same AI and the same afternoon, and watch what happens.
The first opens the tool and types “write me a marketing plan.” Back comes a page of competent, generic advice: know your audience, leverage social media, measure your results. They skim it, recognise it could have been written about any company selling anything, sigh, and close the tab. AI, they conclude, is overrated.
The second tells the tool who it is (“you are a marketing strategist who works with independent coffee roasters”), who they are, what their shop sells, who their customers are, what they have tried, and their budget. They paste in three campaigns they admire. They ask for a one-page plan with a specific structure. When the draft returns, they push: “the second idea is too expensive, replace it; make the tone warmer; cut it by half.” Ten minutes later they have something genuinely useful, tailored, and theirs.
Same tool. Opposite outcomes. The difference was a skill every manager already knows. The first person used the AI. The second delegated to it.
What is the core reframe?
Most people are disappointed by AI because they treat it like a search box or a vending machine. The cure, in the Wharton professor Ethan Mollick’s three words, is: delegate, don’t abdicate.1 The value is not in the typing. It is in the managing. You are not operating a tool, you are directing a worker, and the quality of what you get back depends almost entirely on the quality of your direction, exactly as it would with a person.
What kind of worker are you managing?
Picture it accurately, because you cannot manage well what you have mis-imagined. The machine is fast, tireless, widely read, and eager to help. It is also relentlessly literal, has no memory of you between sessions, and knows nothing about your company or constraints unless you tell it. Most importantly, when it reaches the edge of its knowledge it does not stop and admit ignorance. It produces a confident, fluent, wrong answer in the same tone it used for the right ones.2 Manage it like a brilliant new hire on their first morning: assume they know nothing about your world, give them context and examples, and check their work before it goes out under your name.
What separates the people who win from the people who give up?
Researchers describe three styles. Centaurs divide work deliberately, keeping some tasks and handing distinct others to the machine. Cyborgs weave the two together sentence by sentence. Self-automators hand over an entire task and disengage, barely looking at what comes back.3 The self-automators are the cautionary tale. They learn the least, supervise nothing, and produce the worst work, because handing over a whole job and looking away is not delegation. It is abdication with extra steps.
How do you actually do it?
Give the model a role, supply your context, show an example of what good looks like, and specify the deliverable, its format, and its length, because the model cannot hit a target you have not defined. Then treat the first output as a draft, not an answer, and iterate. The people who get nothing from AI and the people who get a great deal are usually not using different tools. They are exercising different amounts of management. Be a centaur. Decide, on purpose, what you keep and what you give away, and stay the editor-in-chief of everything that carries your name.
Frequently asked questions
What does delegate, don’t abdicate mean for AI?
It means directing AI like a capable new hire, giving it context, examples, and clear instructions, then checking its work, rather than handing over a whole task and disengaging. The phrase is from Ethan Mollick, and abdication produces the worst results.
Why do people get bad results from AI?
Because they treat it like a search box, typing short contextless requests and judging the generic output that falls out. The value is in the managing, not the typing: the quality of what you get back tracks the quality of your direction.
How do you get better results from AI?
Tell it who it is and who you are, supply context and examples, specify the format and length you want, then push back on the draft to refine it. Manage it the way you would manage a brilliant new hire on their first morning.
About the author
Tom Goodwin is the author of Don’t Work Harder, a book about taking the time AI gives back as time rather than more work. He is a co-founder of GAMEPLAN and writes on productivity, technology, and the economics of the working week.
Footnotes
Footnotes
-
Ethan Mollick, Co-Intelligence: Living and Working with AI (Portfolio/Penguin, 2024): “delegate, don’t abdicate,” with the human as editor-in-chief. ↩
-
On the “jagged frontier”, AI producing confident, fluent output even where it is wrong, see Fabrizio Dell’Acqua et al., “Navigating the Jagged Technological Frontier” (Harvard Business School, 2023). ↩
-
The centaur, cyborg, and self-automator distinction is discussed in Mollick’s work and associated research; passive self-automators tend to learn least and produce weaker results. ↩