The Prompt Problem
You've been told to "learn prompt engineering." Spend hours crafting the perfect instruction. Tweak, revise, iterate.
Here's the truth: the best prompt is no prompt at all.
Just talk to your AI assistant like you'd talk to a colleague.
Why Voice Wins
Claire Vo recently pointed out something obvious in retrospect: people talk more freely than they type.
When you type, you pressure yourself to be concise, clear, perfect. You edit as you go. You leave out context because "it's probably not relevant."
When you talk? You ramble. You include tangents. You mention that thing your customer said last week. You explain the background story. You give the AI way more context — not because you're trying to, but because speech is how humans naturally communicate.
And AI needs context. More context = better results.
The Real-World Impact
Think about the last time you delegated a task to a human assistant. Did you write them a perfectly structured brief? Or did you walk over to their desk and say:
"Hey, can you handle this thing? So the customer called yesterday and they're upset about X, but actually the real issue is Y, and I think we should do Z but I'm not sure — can you look into it and draft something?"
That rambling 30-second explanation? That's what your AI needs. And it's way easier to speak it than type it.
Practical Voice-First Workflows
1. Morning Briefing
Instead of typing a to-do list into your AI, record a 2-minute voice note while making coffee:
"Okay, today I need to follow up with the Johnson account — they emailed about the proposal but I haven't responded. Also need to check on that vendor quote, I think it was for the office supplies. Oh, and draft a post for LinkedIn about the new service we're launching, keep it casual, mention how we're helping local businesses save time..."
Your AI transcribes this, extracts action items, and starts drafting. You just talked. No formatting, no bullet points, no "prompt structure."
2. Customer Response Drafts
Customer sent a complex email? Don't type out a response from scratch. Talk it out:
"Reply to this — thank them for reaching out, apologize for the delay, explain that we were waiting on the vendor quote which just came in, and propose Tuesday at 2pm for a call. Keep it friendly but professional, they're a bit frustrated so acknowledge that."
AI drafts the email. You review, tweak, send. Done in 60 seconds.
3. Content Brainstorming
Need blog post ideas? Content calendar? Marketing angles? Talk through it:
"I want to write about AI automation for contractors, but not the generic stuff. What would actually help them? Probably estimating — they hate doing quotes. And maybe scheduling, because no-shows kill them. And follow-ups, because they're bad at that. What are some angles that would resonate?"
AI gives you 10 angles. You pick one. Write the post. Done.
Platform Support
You don't need special tools. Voice is everywhere:
- OpenAI ChatGPT: Tap the microphone, talk naturally
- Google Gemini: Voice input in the mobile app
- Claude: Voice notes via mobile, or use your phone's dictation
- Apple Notes: Dictate, then copy/paste into AI
- Windows/Mac: Built-in dictation (Win+H on Windows, Fn+Fn on Mac)
The platform doesn't matter. The workflow does.
The Counterintuitive Part
Here's what feels weird at first: you'll get better results when you're less structured.
Typed prompt:
"Write a customer email. Tone: professional. Include: apology for delay, vendor quote received, propose Tuesday 2pm call."
Spoken prompt:
"Hey, so this customer emailed us yesterday and they're pretty frustrated because we haven't gotten back to them. Honestly we were waiting on this vendor quote and it just came in this morning — I totally forgot to follow up. Can you draft something that thanks them for reaching out, apologizes without making excuses, mentions the quote is ready, and suggests Tuesday at 2pm for a call? They're a bit annoyed so we should acknowledge that but not dwell on it. Keep it friendly but professional."
Which one gives the AI more to work with? Which one sounds like a real human? Which one is easier to produce?
When Voice Doesn't Work
Voice isn't perfect for everything:
- Code generation: You'll still want to type specifics (file names, function signatures, exact requirements)
- Data-heavy tasks: "Pull Q3 revenue from row 47" is faster to type
- Sensitive info: Maybe don't dictate banking details out loud in a coffee shop
But for 80% of knowledge work — emails, drafts, brainstorming, planning, research — voice is faster and more natural.
The Takeaway
Stop treating AI like a search engine that needs perfect queries.
Start treating it like a colleague who needs context.
And the easiest way to give context? Just talk.
What's your experience with voice vs. typing for AI? Drop a comment if you've tried this — does it work for you, or am I missing something?
Top comments (0)