AI Makes Workplace Writing Cheap But Reading Remains Expensive
Reactions from ranked influencers
2 posts
So true. Seeing this happen a lot. @shreyas used to say, there is no writing culture unless the org has the reading culture. But that's broken now. Irony is this tweet will end up being used as a good prompt for agents than humans lol.
One thing that is increasingly driving me crazy: AI made workplace writing cheap, but reading is still expensive! Again and again I see posts that are now twice as long as need be, and more meandering than a walk in the woods. The problem is simply this: the writer thought about what they wanted to say and asked a model to draft it. They did they stop to think: "What do I really want the reader to take away?" Before you ask AI to draft your next email, memo, or Slack update, answer these five questions: 1. Who exactly needs to read this? Is this for your manager? The leadership team? Your whole company? Treat their attention like a budget. If 100 people each spend three minutes reading your note, you have just spent five hours of company time. Make the note crisp! 2. Who has the least context? Write for the least-informed person in your target audience. Your manager may not remember the outcome your team is chasing. Remind them: “Our goal is to launch X by Y date and get Z response.” Your immediate team may know that DNU means “daily new users.” A broader audience probably will not. (AI is especially good at speaking jargon!) 3. What should readers do after reading? Be specific: - Prevent a bad decision: explain what is happening, then state what you do not want people to do. - Get a decision: put DECISION NEEDED in the first line and lay out the options. - Create confidence: lead with the status and the evidence, so readers can stop after the first paragraph. - Change a behavior: name the current behavior, the problem it causes, the replacement behavior, and how it will be enforced. If you cannot name the intended response, AI will happily produce three paragraphs that lead nowhere. 4. What is the minimum context required to get that outcome? Include the goal. Explain how you got here only when it changes the decision. State what you need from the reader. Then stop. If you feel the need to include every piece of history, at least use a TL;DR followed by “Additional context.” AI can help compress the note, but it cannot know which details matter unless you have told it what the reader needs and it will happily write you a tome. 5. Is text the clearest way to make the point? When AI makes text easy, it is tempting to reach for text even when another format would work better. If you are explaining a data insight, include the chart. If you are describing a product decision or bug, include a screenshot or short video. If the message is complex and you are clearer in person, record a short video. AI can draft the sentences, but it's still your responsibility to make sure that the collective five hours that 100 people spent reading them was a good use of time!
One thing that is increasingly driving me crazy: AI made workplace writing cheap, but reading is still expensive! Again and again I see posts that are now twice as long as need be, and more meandering than a walk in the woods. The problem is simply this: the writer thought about what they wanted to say and asked a model to draft it. They did they stop to think: "What do I really want the reader to take away?" Before you ask AI to draft your next email, memo, or Slack update, answer these five questions: 1. Who exactly needs to read this? Is this for your manager? The leadership team? Your whole company? Treat their attention like a budget. If 100 people each spend three minutes reading your note, you have just spent five hours of company time. Make the note crisp! 2. Who has the least context? Write for the least-informed person in your target audience. Your manager may not remember the outcome your team is chasing. Remind them: “Our goal is to launch X by Y date and get Z response.” Your immediate team may know that DNU means “daily new users.” A broader audience probably will not. (AI is especially good at speaking jargon!) 3. What should readers do after reading? Be specific: - Prevent a bad decision: explain what is happening, then state what you do not want people to do. - Get a decision: put DECISION NEEDED in the first line and lay out the options. - Create confidence: lead with the status and the evidence, so readers can stop after the first paragraph. - Change a behavior: name the current behavior, the problem it causes, the replacement behavior, and how it will be enforced. If you cannot name the intended response, AI will happily produce three paragraphs that lead nowhere. 4. What is the minimum context required to get that outcome? Include the goal. Explain how you got here only when it changes the decision. State what you need from the reader. Then stop. If you feel the need to include every piece of history, at least use a TL;DR followed by “Additional context.” AI can help compress the note, but it cannot know which details matter unless you have told it what the reader needs and it will happily write you a tome. 5. Is text the clearest way to make the point? When AI makes text easy, it is tempting to reach for text even when another format would work better. If you are explaining a data insight, include the chart. If you are describing a product decision or bug, include a screenshot or short video. If the message is complex and you are clearer in person, record a short video. AI can draft the sentences, but it's still your responsibility to make sure that the collective five hours that 100 people spent reading them was a good use of time!