What area would you NEVER trust to Ai, in your inbox?
by•
I mainly follow the principle of "reversibility".
EG. An archived newsletter or a starred investor email can be always reversed.
If ai is right -> makes my life easier
If ai is wrong -> doesn't hurt me
Meanwhile, I would NEVER let ai auto-send things (especially to other humans) for me.
Because no matter how accurate its tonality is to me,
no matter how well it understands my context,
I just don't... Maybe there's some kinda humanity in pressing "send" yourself!?
What about you? Where would you draw the line? Or just raw-dog it and trust the ai?
152 views
Replies
I draw the line at anything that speaks for me.
Dirac
@ethan_marshall So, anything that other people can see?
I agree on your point but more than that i would'nt even want ai to look at my banking mails.
Dirac
@nischay_kashyap You can manage areas of your inbox and choose what to shut/privatize in Dirac
@peterz_shu I would rather miss an optimization than send the wrong message.
Dirac
@caleb_jonathan I 100% agree with that.
I like the reversibility principle. I would add one more layer: does the AI action change someone else's reality, or only organize your own view?
For me:
- safe: summarize, classify, draft, highlight risk, suggest a next step
- caution: archive, label, prioritize, change CRM/status fields
- human-only: send, commit money, approve, reject, delete, or make a promise on my behalf
That distinction keeps coming up across products. With ReplyFlow, the safest workflow is test-mode or draft replies before connecting a live customer channel. With PayWise, AI can suggest which payment or debt needs attention, but the user should still make the actual money decision. With RaiseReady, AI can expose missing assumptions in a funding plan, but it should not pretend the business is ready before the owner verifies the numbers.
So yes, reversible actions are a good first filter. External consequences are the second filter.
Dirac
@estreamedia Yeah that is another great point!
You're onto something here...
And this is not just applicable for email alone. But most ai-related apps out there should follow this.
Dirac
Some great comments We Got ------->
The reversibility principle is sharp. I might steal it, honestly. It's the cleanest rule I've seen for where to hand off vs hold on.
I draw the same line in ads, and it maps almost exactly. I'll let AI watch the data, flag problems, draft 15 variations, recommend the budget move. All reversible or pre-send. But the actual "spend more money" call and anything going out under my name to a human stays mine.
Same instinct as your send button: the irreversible stuff and the relationship stuff stay manual. Money and trust, basically. Those are the two I won't auto-pilot.
The part you said that I keep chewing on is the "humanity in pressing send yourself." I don't think it's sentimental. When you press send, you re-read it as you. The AI optimized for tone, but you're the only one accountable for what it means. That last human pass is where the judgment lives.
So yeah, not raw-dogging it. AI does the reps, I keep the calls that can't be undone.
Dirac
@nipuntaneja Yeah, most Ai-agents related things should follow this. EG. ai shouldn't just start an ad campaign and post on meta ads. You always want to be the one checking to make sure that there are no errors.
Thanks for your POV :)
memi
I would trust AI to sort, summarize, and prepare. I would not trust it to send anything where tone changes the relationship. The cost of one wrong "sure, sounds good" is weirdly high.
Dirac
@sarveshsea So even if you see that it follows your tone,, you still wouldn't trust ai with tone?
Intersting, thanks for sharing!
For me the line is: seconds of cleanup vs. relationship damage. Archiving or starring = reversible. Sending, auto-CC’ing, or accepting the wrong meeting = not really reversible. Anthropic frames this well around human approval for higher-stakes tool use: https://docs.anthropic.com/claude/docs/tool-use Would you put calendar accepts closer to archiving or sending?
Dirac
@fabriziowexare Yeah, this is developing deeper on the "reversibility" I mentioned. So true.
Me personally? I would put calendar accepts into sending.
I believe in AI governance. No matter how powerful the AI or how intelligent, if the AI can not explain what it did vs what it was allowed to do and prove that it was allowed to do it....why on earth would I turn it loose in the real world?
Dirac
@thomasbcoleman Can you explain more on Ai governance? Seems interesting
@peterz_shu Hey Peter — you asked about AI governance, so here's the quick version, and you're going to recognize it. Governance = can I trust an AI to act on its own: see what it did, why, and undo it. You already built a governance rule into Dirac — "background tasks follow one rule: reversibility." That's exactly the line serious agent builders draw. You just didn't call it that.
What I work on is the layer up: continuity engineering. The premise in plain terms — an agent fails when it loses track of what it's doing faster than it gets smart. Doesn't matter how good the model is if it forgets what it already decided or what it's not allowed to do. For something like Dirac running in the background across threads (and more so once Sentry/Stripe/Linear are in), that "remember the rules and prove you followed them" layer is what keeps trust as it does more.
Easiest way to see it: I make a free, open-source extension, Prompt Accelerator (Chrome/Firefox), that keeps an inspectable record of a chat's real objective, decisions, and what was ruled out — so state doesn't drift. There's a heavier version (EchoGate) built for agents that take real actions: reversibility-plus, where every action is checked against fixed rules and logged so you can replay exactly what happened. That's more Dirac's world long-term.
Not pitching — you're already thinking about it the right way. Happy to send the link or just trade notes on agent-trust. Congrats on the launch.
Same line of thinking here. AI sorting, summarizing, drafting? Fine. But auto-sending, especially to a client or a stakeholder? Never. It's not even about accuracy, it's about accountability. When something goes out with my name on it, I want to have actually read it first, even if the AI nailed the tone 100%.
There's also something real about the act of pressing send yourself. Like a small moment of "yes, I'm choosing to say this." Removing that feels like it removes something that actually matters.