I don’t know how many of you have used ChatGPT voice mode, but this is something which is really interesting and which I found very productive to use AI after using it very much. I was actually typing like a madman. No matter the time, venue, mood of the day. I type but finally that changed with Voice mode.
If you want to watch the video, you can over here -
Watch the accompanying video on YouTube
//video place holder here
ChatGPT voice mode is something which was clubbed together by the ChatGPT, where they ensured that chat states or the queries and everything are there. Work mode is where we actually have the ability to provide the files to it, like plot work. There is something distinct over here: the voice mode which can be used for both but work mode via voice can access your files to get things done.
The voice mode is something which did change the whole perspective because, unless we type what we actually want, this is something where we actually have a conversation with the AI. This is something more like we talk in real time. It’s not that we talk, wait, and then get back the result, but it’s actually doing the work behind the scenes in getting us results in real time.
The Test
What I have done is I have pointed it to a read-only personal SFDX repo. I mentioned that we’ll have to use the metadata files, config and not change anything unless specificied. That’s it. Voice mode is not dictation. You don’t dictate; you talk, really talk, in a flow that will actually get things done, not copy, then type, then paste. It’s in that mode: the conversation, you interrupt, and steer and adjust with what you need. This is something which you can just stop composing a clean prompt, and you can just say it. Of course, if there is a mistake, you can change it again.
Now, this is the state of my org. There are actually 50 flows, and exactly only one carries documentation and testing tools with different names like test, test bro, bro, rs test_*. There is only one document out of all the flows that’s almost like real life, right? What’s the first move that I had done? The first move is I asked it to open the folder, read-only, and made it understand that it needs to understand what the particular flow called annual revenue generating doc covers and what the flow does in the org, like the XML file and what it actually does. See if there are any differences, what logic is present in the flow but the document didn’t mention, and if there is anything uncertain or unsure that needs to be flagged.
The Part which is interesting
This is where it was interesting because it did sweep across the flows and did understand what my conventions are, but before that, it stopped and showed me the exact element in the flow XML where it showed the discrepancy. Sometimes documents do lie, and metadata is the actual truth. This is something which is much needed for admins, analysts or any persona. This is something which actually changes the way we work, because in a real project, there will be multiple flows connected together, like flows and subflows. You have to pinpoint the way it has to work by giving it context so that it works and understands before it goes sideways. If you feel like it is going sideways, especially with voice mode, you can just stop, and then it gets done.
The flexibility it offered
Now, how did it run while I was working? I was able to pin it across the windows, only a few windows. You can read another tab and then check the doc, and you can talk to it the whole time whenever you move through the windows, tabs, or any application. It is more of an orchestration, and then you come back and then output. Mute is the one which will actually make things better because you can mute the voice mode and tell things out loud yourself. It doesn’t need every word that is not performing a prompt but is thinking beside it and then deciding what to do.
The output
Finally, what has surfaced? What are the junk names that are present in the proof of concept or not ready for production yet, which was only done and everything? It did help me rename the flows and write the description for each element, and then create all the documents, which are perfect. What I felt is that it let me speak freely, with no question in my mind. I speak as I have it in my mind. This is something that needs to be ensured in any project time rather than spending much time when you have a coworker kind of persona with AI. You can actually talk like how Iron Man talks with Jarvis. This is something which did change everything, the way I view work with AI, and you can try it too. This is how the actual voice mode did work. Let me know your thoughts if you have tried it. Until then, bye. See you next time.
If this appeared in your inbox and not a subscribe yet, hit that Subscribe now button and get latest insights I work at the intersection of Salesforce and AI
