When AI Writes the First Draft, Who Does the Rest?

# When AI Writes the First Draft, Who Does the Rest? Luxury Activist | The Notebook LA (Luxury Activist) 6 Nov 2026 https://luxuryactivist.com/carnets/friday-column-2-when-ai-writes-the-first-draft/ ## In brief AI gives writers and designers more ideas in less time. Someone still has to decide what’s worth keeping and check that it works. In my first draft, I wrote that artificial intelligence simply amplified the talent of whoever used it. I liked the idea enough to put it in twice. Then I asked an AI to help me revise the column. It came back with a study that gave me a reason to rethink the opening. Anil Doshi and Oliver Hauser asked people to write short stories. Some were given story ideas generated by AI. Readers rated their stories as more creative, especially when the writers had scored lower on the researchers’ creativity test. Read together, though, the stories written with AI help had more in common. You could end up with better stories and a stronger sense that you’d read this one before. ## Assisted intelligence In 1955, John McCarthy and three colleagues used the phrase “artificial intelligence” in a proposal for a workshop at Dartmouth. They started with an ambitious idea. If abilities such as learning and using language could be described precisely enough, a machine could simulate them. When I revise a piece with one of these tools, I tend to think of it as assisted intelligence. I can ask for a lead, read the study it points me to and make up my own mind. I’d have a much harder time doing that in quantum physics. A few Greek letters in an equation and I’m already impressed. I could use the tool to learn the basics. To put forward a new idea, I’d also need to know how to test it. ## An apartment for a birdwatcher In a design study, Linus Tan and Max Luhrs imagined an apartment for a birdwatcher. They used Midjourney to explore shapes and spaces. One of the first images, prompted by Melbourne’s street grid and skyscrapers, was too abstract to give them a direction. A later image suggested viewing platforms at different heights. They took that idea into the design. Their paper shows both the images and the architectural proposal they arrived at. In between, they had to work out how the spaces would fit together in three dimensions. I’d probably still be picking my favourite image by the time they started drawing a section. ## The time we forget to count A team can use AI’s speed to explore more ideas. It can also see that speed as a reason to put fewer people on a project. The first version comes quickly either way. If I were doing the sums, I’d want to know how long it takes to check the work, who does the checking and whether that person can still send it back. Then there’s the electricity. The International Energy Agency finds that energy use per AI task is falling as more people use the tools and more demanding applications, such as video generation, take off. In its central projection, electricity consumption across all data centres worldwide rises from 485 TWh in 2025 to about 950 TWh in 2030. Consumption at AI-focused data centres would triple over the same period. One more version might be covered by my subscription. For a service people use every day, those extra versions add up. I could have timed the AI’s first response to my draft. The number would have been flattering. I’d also have to count the time I spent wondering whether I still agreed with myself. ## Sources & references - Anil R. Doshi and Oliver P. Hauser, experiment on writing short stories. Science Advances, 2024: https://doi.org/10.1126/sciadv.adn5290 - John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon, proposal for the summer of 1956. Original document, 1955: https://www-formal.stanford.edu/jmc/history/dartmouth.pdf - Linus Tan and Max Luhrs, architectural design study using Midjourney. The Design Journal, 2024: https://doi.org/10.1080/14606925.2024.2353479 - International Energy Agency, global projections for data centre electricity consumption. Key Questions on Energy and AI, 2026: https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary