ChatGPT Work

Partner for your most ambitious work

Artificial Intelligence

Visit WebsiteSee on Product Hunt

How does it actually handle permissions when working across multiple apps — do I need to manually approve each action or is there a way to set boundaries upfront?

How does it handle permissions when it needs to access files or apps across my workspace, and can I set boundaries on what it can touch without me approving every step?

Finally tried letting it handle a multi-step research task across Drive and Slack. It actually followed through without me hand-holding, which surprised me.

The way it can stay with a project for hours without losing context is genuinely impressive. Most agents I have tried tap out way sooner, so the persistence here feels like real craft, not just a flashy demo.

Finally got it to draft a whole slide deck straight from a rough brief, which saved me an embarrassing amount of time on a deck due in two hours.

Finally an agent that stuck with a messy spreadsheet cleanup for two hours without losing the plot. Loved watching it ping between Drive and Slack on its own, though the handoff back to me felt a little abrupt.

Plugged it into a messy spreadsheet cleanup and walked away for an hour, came back to a finished report. Wild that it just kept working without me babysitting it.

Finally gave it a real task with messy folders and it just kept going for hours without me babysitting. Kinda wild to see it pull a full project together while I grab coffee.

The patience of an agent that can actually sit with a project for hours instead of bailing after one reply feels like the real craft here.

How does it actually handle permissions when it needs to pull from different apps, like does it ask each time or just go ahead once I grant access upfront?

how does it actually decide when to keep going on a task versus asking me for input

Had it pull together a quick competitive analysis across our shared docs and Slack threads while I made coffee. Coming back to a clean summary I could actually send around felt like a real time shift, not just a smarter chatbot moment.

How does it actually decide when to keep going versus stop and ask for input, especially if it gets stuck partway through something like a multi-step Notion to Slack handoff?

the way it stays with a project for hours and actually finishes the work, not just answers a prompt, that's a real shift in how these agents feel

How does it handle permissions across apps like Slack or Drive when it needs to take actions on your behalf, and is there a way to see or undo what it did after the fact?

How does it handle permissions when pulling from private company files, and is there any audit trail of what it accessed or changed during a longer task?

I used it to pull together a messy research folder into a written brief and it actually finished the job without me babysitting. Genuinely surprised how long it stuck with the task.

Gave it a task to pull data from a few docs and draft a summary, came back to actually usable copy an hour later. Surprised it stuck with the project without me babysitting it.

had it pull together a research doc across three apps and it actually stuck with it for a couple hours without me babysitting, which was kind of wild