I Planned to Stop at a Café to Work, but the Quotation Was Done While I Talked to ChatGPT
How I prepared a quotation by talking to ChatGPT Voice on the way home, why I pick ChatGPT for voice conversations, and what you still need to check yourself.

That day I had a quotation to send. The plan was to stop at a café on the way home, open my laptop and finish it there.
In the end, I didn't need to stop. On the way, I "called" ChatGPT, explained the quotation while we talked, and it prepared the document right away, including edits to the parts I asked to change mid-conversation. By the time I got home, all that was left was to review it, make a few edits and send it.
I use ChatGPT like this fairly often for everyday work, with the GPT-6.1 Sol model. In this article I explain why this works now, why I pick ChatGPT for talking by voice, and what you still need to check yourself.
Why this works now
Two changes in the last few weeks made this way of working practical. According to the ChatGPT release notes:
- September 23, 2026: Voice arrived in Work mode on web and mobile. You can ask ChatGPT to create documents, presentations and spreadsheets, use connected apps or work in a browser, just by speaking. If you end the call before a task is done, it can continue in text.
- September 29, 2026: GPT-6.1 Sol started rolling out in ChatGPT Work and Codex. OpenAI says it improves on GPT-6 Sol for professional work, coding and computer use.
The "task can continue after the call ends" part is, to me, the most important: you don't have to stay on the line waiting for the document to be finished.
My workflow
- 1Talk on the wayExplain the quotation through Voice
- 2ChatGPT draftsWrites the document, revises mid-conversation
- 3Review at homeCheck figures, names and terms
- 4SendOnce everything is right
While we talk, I treat it like an assistant I've called: I explain what goes in the quotation, and if something needs changing, I ask for it right then, like a normal conversation.
Why ChatGPT, not Claude
This is personal experience, not a measured test. For talking by voice, I find ChatGPT the most fluent and the best communicator. It really feels like calling an assistant. Claude, to me, is very tactical, its answers go straight to the point. But by voice, I feel it takes longer to process what I say, so the conversation doesn't flow as well.
Looking at each one's documentation, there's an explanation for that difference in feel:
| ChatGPT Voice (Live) | Claude voice mode | |
|---|---|---|
| Turn-taking | Can listen and speak at the same time, so interrupting feels natural | Listens continuously and responds to natural pauses, or hold a button to talk |
| Conversation model | A dedicated voice model, GPT-Live-1 or GPT-Live-1 mini depending on plan | The same models as text chat (Opus, Sonnet, Haiku), in their fastest version |
| Languages other than English | Language can be set in Voice settings | Still in beta, per Claude's help page |
Claude keeps improving too. In July 2026, Claude's voice mode was updated to use Opus, Sonnet and Haiku, where before it was Haiku only, and to support Indonesian. So this comparison may change within a few months.
What you still need to check yourself

The review at home isn't a formality. Things I always check:
- Figures. Prices, quantities, tax and totals. OpenAI's own help page reminds you that ChatGPT can make mistakes and important information should be checked, especially anything about dates and times.
- Client names and details. Mishearing one letter of a company name is embarrassing on an official document.
- The data you say out loud. Whatever you say by voice is saved in the chat too. I covered which data is safe to bring to ChatGPT in Want to Summarize Your Company's Financial Data in ChatGPT? Is It Dangerous?.
- Actions that leave the chat. If ChatGPT wants to do something through a connected app, like sending an email, you approve it on screen. According to OpenAI, spoken approval isn't supported. I think that's good, because sending a document to a client should be a decision you make while looking at the result, as I discussed in At What Point Should Humans Still Check AI and Automation Results.
One more thing, about safety. If you use it on the road, especially while driving, set the app up before you leave and don't operate your phone while the vehicle is moving. That's also what OpenAI advises on its ChatGPT Voice help page.
Plans and usage limits
Voice in Work mode needs both Voice and Work access, and Work tasks still count as regular Work usage. Daily Live limits in Chat, per OpenAI's help page:
| Plan | Live limit per 24 hours |
|---|---|
| Go | 3 hours with GPT-Live-1 mini |
| Plus | 3 hours with GPT-Live-1 |
| Pro (US$100/month) | 15 hours with GPT-Live-1 |
| Pro (US$200/month) | Unlimited |
OpenAI is rolling GPT-6.1 Sol out gradually, starting with Pro users and then Plus, Business, Enterprise and Edu. So if it hasn't appeared in your account yet, it probably just hasn't reached you.
Closing
What changed for me isn't just speed, but when the work can start. Travel time that used to just pass can now go into preparing a draft, and desk time is left for what actually needs my attention: checking and deciding.
If you'd like routine work in your team, like quotations or reports, to be prepared faster with AI while someone still checks it before it goes out, tell me what you need.
Is your team still doing something by hand?
Tell me about it, and let's find the fix together.
More notes

Can AI Learn Your Sales Style From WhatsApp Chat History?
Yes, and you usually don't need to train your own model. How to export WhatsApp chats, turn them into a style guide, a ready-to-use prompt, and the customer data risks.
Read the note
Gemini vs ChatGPT for Office Work, Which One Fits Better?
Gemini and ChatGPT can both be used for work, but their strengths differ. A plan price comparison in Rupiah, which suits teams on Google Workspace, and when ChatGPT makes more sense.
Read the note