How to Deslopify Your AI
Think before prompting, keep your point of view, and review AI work to fit the audience and stakes before sending it.
- Brain-dump your view, confidence, audience, and desired outcome before opening an AI tool.
- Ask for varied directions, choose review checkpoints early, and challenge drafts with distinct reviewer perspectives.
- Read everything, cut it to the audience’s needs, and don’t send work you wouldn’t put your name on.
Practical guidance is clear in the notes; the full discussion adds workflow examples and context. · Best for: Professionals and managers using AI to create work others must read or act on.
The speakers explain why AI-generated work tends toward generic, overlong output, then offer ways to preserve human judgment and reduce the burden on readers. Their central principle: use AI to expand and shape your work, not to replace your thinking.
What makes work feel like slop2:50
- Slop is low-quality, mass-produced content that shifts the effort of thinking and filtering from the sender to the recipient.
- Generic prompts tend to produce typical answers; human preference for familiar, long responses reinforces that tendency.
- People may forward output without checking it because it looks plausible, they are under pressure, or they lack relevant craft knowledge.
A test for AI-assisted work10:36
- Ask whether you can defend every sentence and whether you read the whole piece before sending it.
- Check for a contribution only you could make, such as a point of view, decision, or real data.
- Judge whether the format saves the reader time, and whether you would put your name on the result.
Keep human judgment in the workflow13:35
- Choose which parts are yours and which AI can handle based on the task’s stakes and purpose.
- One speaker uses AI for broad research but keeps the point of view and narrative structure under human control.
- For a long companion essay, the speaker provides a raw spoken transcript and then spends 5–10 hours revising the AI draft.
Before and during AI work26:49
- Think first: take a walk, read, or dictate a messy brain dump of your ideas and concerns.
- Tell the tool how confident you are and how strongly you want it to challenge your view; ask for counterarguments.
- Define the audience and the desired outcome, then request several approaches and compare them.
- Set review checkpoints, such as the outline and first section, before asking for a complete draft.
Review, personalize, and set team norms30:39
- Ask AI to review from distinct perspectives, such as a skeptic, domain expert, target reader, or busy recipient.
- Respect the audience’s reading budget: shorten the work and choose a format they will use.
- Teach AI your preferences using examples of work you like and dislike, but refresh those preferences to avoid repetitive output.
- Leaders can name low-effort work and request a concise, considered version; label AI-assisted drafts when that context matters.
Where the value shifts38:23
- As basic AI production becomes common, differentiation comes from taste, intention, and original thinking.
- Use AI to explore more options and improve quality, while retaining a human fingerprint in the final work.
Step by step
- Before prompting, record your own view, concerns, confidence level, audience, and desired outcome.
- Tell the AI what you believe and how firmly to challenge you; request a counterview when useful.
- Ask for multiple directions that differ in angle, structure, or approach; compare them and choose what serves the task.
- Set human review checkpoints in advance, especially for the outline and early draft, before the work becomes polished.
- Use AI reviewers with distinct roles to find gaps, objections, and audience issues; keep only useful feedback.
- Read the full output, revise it to match your voice and the audience’s time, and send it only when you can stand behind it.
Tools & resources mentioned
- ChatGPT, Claude, and Gemini — compared in a generic LinkedIn-post experiment to show how models can produce similar results; personal context improved the drafts.
- AI writing detectors — discussed as a rough way to spot common structural patterns, not as reliable measures of quality or originality.
- AI humanizers — discussed as tools that alter style, but not substitutes for adding a distinct point of view.
- Figma AI-writing rules — cited as an example of labeling AI-written material so readers have context.
- Companion page on the AI Daily Brief website — described as containing practical materials and slides for the episode.
Notes are AI-generated from the video's transcript and may miss nuance.