Some requests take a long time to come out because you are trying to organize the idea while you write. Speaking can let the context come out more freely: what happened, what the question is and why a detail matters. Voice, follow-ups and mobility are ways of working with AI worth judging by the routine they help organize.
Tell apart the ways of using voice
With ChatGPT dictation, audio becomes text for you to review. A voice conversation involves back-and-forth with the system. Taking actions depends on the tools and permissions of the environment. Voice is a way of interacting; the ability to act has to be checked separately.
Use speech to recover context
A founder coming out of a meeting with several notes can describe what was discussed and ask for it to be organized into decisions, open questions and next steps. Then it needs checking: a similar-sounding name, a lost "not" or a possibility recorded as a decision all change the meaning. The review pays special attention to commitments, people and deadlines.
It also helps to separate what was said in the meeting from what you thought afterward, so a personal reflection does not show up as a group agreement.
Test one change at a time
Follow-ups. Pick something that needs to be watched again, such as preparing for a meeting. Describe the information needed, the occasion and the alert you expect. Recurring runs depend on configuration: writing an instruction in the chat does not prove the follow-up was turned on.
Voice. Explain out loud a situation that usually takes a long text, and check whether the summary kept the open questions and commitments.
Mobility. Pick up a task on another device, when the environment allows. Notice whether the switch helps or just adds interruptions.
Decide what should reach you
An alert about missing information can be useful; a stream of notifications about details with no consequence only scatters attention. Agree on which events justify an alert and which can wait for the next summary. If the AI has access to files or apps, spell out what becomes a draft and what it is authorized to carry out.
Mobility also needs a purpose: clearing a task while commuting can be convenient; filling every gap in the day may work against the result you want. Thinking about where the time AI frees up should go helps with this choice.
At the end of each test, note what got simpler and what new effort appeared. In AI Thinking by Dani Hayoshi, the best interface is the one that lets you understand, decide and finish the work. See also How to use AI at work beyond asking for text.
Dani Hayoshi is the author of AI Thinking. Her work connects artificial intelligence, education and entrepreneurship. About Dani.