A deck is an argument, not a summary
The reason generated decks feel thin is rarely the design. It is that the tool has been asked to describe an idea rather than to defend one. A summary lists what the idea is. An argument tells the room why this idea, why now, and why the obvious alternative is worse.
Before you generate anything, write the one sentence you want someone to repeat in the corridor afterwards. If the tool cannot be told that sentence, it will average your idea into something safe, and safe is what loses.
What to feed it
The quality of a generated deck tracks the quality of four inputs far more than it tracks the prompt.
- The tension. Not the category problem — the specific human contradiction the work resolves.
- The proof you already have. One real number, one real quote, one real observation beats a page of assertions.
- The constraint. Budget, channel, legal, whatever it is. Constraints make ideas specific; without them you get wallpaper.
- The room. A deck for a founder is not a deck for a procurement panel, and the argument order changes completely.
The slides it gets wrong
Two slides are consistently weak in generated decks, and they are the two the room remembers.
The idea slide comes back over-explained. A strong idea slide is short enough to read in the time it takes to look up. If your generated version needs three sentences to land, the idea is not finished. That is useful information, not a formatting problem.
The "why us" slide comes back generic, because the tool has no access to the thing that makes you credible. Write it yourself. Every time.
Where to take the pen back
Use generation for the parts that are labour: structure, sequencing, the supporting slides, the version of the argument you need for a different audience. Keep the idea slide, the opening, and the ask.
The test is simple. Read the deck aloud. Anywhere you hear yourself explaining rather than convincing, that section needs a human.