Pickerbird’s three access tiers — Closed, Ask-me-first, and Open — decide what your AI can touch. You set them per item, and the app enforces them, so the AI cannot override them.
Most apps that say you’re “in control” mean it the way a sign means it. Please don’t walk on the grass. The sign asks. Nothing stops you.
Pickerbird’s three tiers are not a sign. They’re a lock.
What are the three tiers?
Every note, every photo, every conversation you have carries one of three settings, and you set it:
Closed. Your AI can’t see it. Not “won’t” — can’t. The content stays sealed to the assistant, and no clever phrasing, no follow-up question, no update on our end quietly changes that. Closed is where a new note or photo starts, before you decide otherwise.
Ask-me-first. Your AI can tell the item exists and can ask to use it — but it has to check with you, in the moment, every time. You say yes, it proceeds. You say nothing, it doesn’t. The permission closes again when you’re done.
Open. Your AI can use it freely, because you’ve decided this is shared ground — your grocery list, your travel notes, the things you’d happily hand over.
Can the AI be talked into ignoring my rules?
No — because the rule was never the assistant’s to interpret.
The important part isn’t the three labels. It’s the word enforced. On most AI products, the boundary between what the assistant should and shouldn’t touch lives inside the assistant’s own judgment — an instruction it’s been given, which a stray sentence or a crafted input can talk it out of. That’s the grass sign. Pickerbird puts the boundary outside the assistant, in the app itself. The assistant doesn’t decide whether to respect a Closed item, any more than you decide whether gravity applies today. The route to that content simply isn’t there.
Two things follow, and they’re worth saying plainly.
Does Closed hide my own content from me?
No. Closed locks the AI out, never you. You can always see your own content — all of it, including anything the assistant can’t.
How do I know what the AI actually used?
You look. This is control you can check, not control you have to believe. When the assistant does something, you can see what it reached for and what it skipped. “Used your notes, skipped your messages” isn’t a reassurance we type into a policy — it’s what happened, shown to you.
That principle — that a limit is only real if it’s enforced somewhere the AI can’t reach — is the same one behind privacy enforced in code, not policy.
Can I change a tier later?
Yes, whenever you like. Set the tiers once. Change any of them with a tap. From then on, your AI works for your whole life — and only ever touches the parts you’ve opened to it.
I worked through this argument in more detail here: The Enabling Guards of Agentic AI.
Join the waitlist to be among the first to try Pickerbird.
FAQ
What are Pickerbird's three access tiers?
Closed, Ask-me-first, and Open. Closed means your AI cannot see the item at all. Ask-me-first means it must check with you in the moment, every time, before using it. Open means it can use the item freely. You set the tier per item, and you can change it at any time.
Can the AI be talked into ignoring my rules?
No. The boundary sits outside the assistant, in the app itself, so it is not a matter of the assistant's judgment. There is no route to a Closed item for it to find, which means no phrasing, follow-up question, or crafted input can talk it into one.
Does the Closed tier hide my own content from me?
No. Closed locks the AI out, never you. You can always see all of your own content, including anything the assistant cannot reach.
How do I know which content the AI actually used?
After the assistant does something, you can see what it reached for and what it skipped — for example, that it used your notes and skipped your messages. It is a record of what happened, not a reassurance written into a policy.