Have the AI outline the deck first, then fix three lines, not twenty pages
You have a long document, a folder of notes and twenty minutes to fill tomorrow. You hand it all over and ask for a deck. What comes back is twenty pages, neatly laid out, about a slightly different subject than the one you meant to present. Stop it before it makes any pages. The first round asks for three things: the one sentence the audience should still have in their head next week, the handful of points that support that sentence, and which of those points your material does not actually back up. That last one is the useful part. The thin points tend to be the ones you were already uneasy about, and you have found them while the fix is still cheap. When the source is a single long document, do a round before even that. Ask only for the main arguments it contains, five or six of them, and read the list against your own memory of the document. What matters is what the list leaves out, and you can only see that before any structure exists to distract you. Ask it to keep a cut list as it goes, everything from your material that did not earn a place. Read that list once. It is a fair picture of what you habitually over-explain, and it changes what you write next time. Only when the skeleton has survived your own editing does it get to build pages. Even then, have it describe what it intends to put on each page in conversation first, and approve that. Generating a deck and correcting it afterwards costs more than the two minutes of reading a plan. Rework drops from an evening to a few minutes, and the argument gets audited on the way. The one thing it cannot do is judge how long you will talk. It will happily tell you the deck runs twenty minutes. Count the pages and decide that yourself.
Comments
Tried this at work? Add what happened.
Sign in to comment