Treat Your First Prompt as a Draft, Not a Final Answer
One of the most useful mindset shifts in working with AI tools: your first prompt doesn't need to be perfect, because the conversation itself is the tool — not any single message in it.
Why this matters
People sometimes spend several minutes carefully wording one "perfect" prompt, when two or three quick exchanges would have gotten them there faster. Trying to anticipate every requirement upfront is slower and less reliable than sending a reasonable first attempt and refining based on what actually comes back.
What a normal iteration looks like
- Send a reasonable first prompt — not perfect, just a genuine attempt.
- Read the response and identify specifically what's off: too long, wrong tone, missing a detail, wrong format.
- Say exactly that: "shorter," "less formal," "add a line about the refund policy," "make this a table instead."
- Repeat until it's actually usable.
This usually takes less total time than agonizing over the first message, and it produces a better result, because each round is informed by something concrete rather than a guess.
Give feedback the AI can act on
Vague feedback produces vague changes, for the same reason vague initial prompts do. "This isn't quite right" gives the model almost nothing to work with. "This is too formal for a message to a coworker — make it sound like how I'd actually talk" gives it a specific direction.
When to restart instead of iterating
Iteration works well when the response is broadly on track but needs adjustment. If a response is fundamentally off-base — wrong topic, wrong audience, misunderstood the task entirely — it's often faster to rewrite the original prompt with the missing context added, rather than trying to steer a response that started in the wrong place.
Carrying context forward
Within one conversation, the AI can see everything said so far — you don't need to repeat earlier context in each follow-up message, just the new instruction. That's part of what makes iterating fast: each message can be short, because it's building on everything already established.
Key takeaway: a back-and-forth conversation that lands on a good result is normal and expected — it's not a sign you did the first prompt wrong.