In the last session you set up Dictation on Mac and voice typing on Windows, and taught your agent to speak up when a job is done. That covered your coding agent. This session widens the habit: talking to any AI you use, and having it read answers back, so a workday can run with your hands mostly free.
Most chat apps (ChatGPT, Claude, Grok) have a voice mode in their mobile and desktop apps. Look for a microphone or voice icon near the message box; the spot moves between versions, so check the app's settings if you do not see one. Open it before you head out on a walk.
Talk through a real problem the way you would think it through with a colleague. A business owner might soundboard pricing for a new rush service option: "how much extra should I charge, given the extra time it takes and what a customer would actually pay for speed?" A nonprofit director might soundboard an event: "I'm planning a fall fundraiser, walk me through a twelve-week timeline and what usually goes wrong."
Let it ask questions back and keep talking. When you land somewhere useful, say: "write that up as a short summary I can read later." You come home with a decision made and a document waiting, keyboard untouched.
Recap: on Mac, hit your Dictation shortcut inside Terminal and speak your prompt. On Windows, press Win + H inside PowerShell, speak, then press Win + H again to stop. Either way, talk naturally in full sentences, then press Enter.
Use it here too. Instead of typing a task, speak it: "Look at last week's invoices and tell me which clients still owe money." Same habit as Step 1, aimed at your agent.
Sometimes the answer is long, and reading it is now the slow part, not writing the prompt. Two fixes.
First, ask your agent to speak its own summary. The say command from the last session works here too: "summarize what you found, then run say with that summary." Keep it short; say suits a sentence or two, not a full report.
Second, use the read-aloud feature built into your operating system for anything longer, like a full report or a long email draft. On Mac, look for Speak Selection or Spoken Content under Accessibility in System Settings: select text anywhere and it reads it back. On Windows, look for Narrator, or Read Aloud inside Microsoft Edge. Menu names shift between versions, so search Settings for "read aloud" if you do not see them where expected.
Recap: Claude Code supports a Stop hook, a command that runs automatically when the agent finishes responding. Paste this into ~/.claude/settings.json (create the file if it does not exist):
{
"hooks": {
"Stop": [
{
"hooks": [
{ "type": "command", "command": "say 'Your agent is done'" }
]
}
]
}
}
Restart Claude Code and it announces itself every time a job lands. Combine this with Steps 1 and 2: soundboard the plan, speak the task, then do something else until it calls you back.
Put the pieces together into one loop:
That is a full workday where the keyboard is optional for most of it.
Voice modes send your audio to the app's provider to become text, the same as any other AI conversation. Do not soundboard a customer's or donor's personal details out loud: names, account numbers, health information, donation history. Keep it general ("a client who is behind on payment," not their name and balance), and save specifics for typed work inside your protected tools.
You have soundboarded one real decision, spoken a task into your coding agent without typing it, heard a summary read back to you, and gotten a spoken ping when a job finished. Run that loop once this week and notice how little of your day actually needed a keyboard.
Template included: Members get a printable card with the voice-first daily routine (soundboard, dictate, step away, spoken report), where to find voice mode in chat apps, where to find read-aloud settings on Mac and Windows, and the privacy rule for what not to say out loud. Find it in your member dashboard under Templates.
Short on time? Alliance Optimal will implement this end to end, wired into your stack, and hand you the keys.