Almost every AI tool has the same limitation: it only exists when you open it. Close the tab and it forgets you are alive. It will never be the one to say "hey, your flight time just changed" or "you asked me to remind you about this."

A proactive AI assistant flips that. It can send the first message.

This guide explains what "proactive" actually means for an AI assistant, the everyday things it can text you about, how it works under the hood, and how to keep it from becoming noise.

Quick answer: Yes, an AI can text you first. A proactive assistant reaches out for reminders you set, follow-ups it is chasing on your behalf, prices or web pages it is watching, and scheduled check-ins like a morning brief. folk does all of this inside iMessage, Telegram, and WhatsApp, and can even call your phone when a text is not enough.

What does a proactive AI assistant mean?

Most assistants are reactive: you send a message, they reply. A proactive assistant can start the conversation on its own, based on things you asked for or conditions it is watching.

The distinction matters because the most valuable moments are the ones you would otherwise forget. You are not going to open an app to check whether a price dropped. You want the app to tell you.

That only works if the assistant lives somewhere it can reach you, which is why proactive assistants tend to live in your text messages rather than a website.

What kinds of things can an AI text you about?

The useful triggers fall into a few buckets:

• Reminders and routines. "Remind me to call mom on Sunday." "Every morning at 7, text me my schedule and the weather." These fire on schedule, even while you are offline.
• Follow-ups it chases for you. "Follow up with Sara if she has not replied by Friday." When the deadline hits, it checks the thread and nudges you, and it never messages the other person without your say-so.
• Watchers on the web. "Watch this product page and tell me if the price drops below $80." "Ping me if my flight time changes." You give a plain-language condition, and it only messages you when the condition is met.
• Scheduled check-ins. A morning brief with your day and the weather, or an evening review. A gentle rhythm, not a firehose.
• A phone call, when it matters. For the things a buzz will not fix, some assistants can actually ring you.

folk groups all of these under reminders and nudges, and the phone piece under folk calls you.

How does an AI know when to reach out?

There is no magic. Proactive messages come from three honest sources:

1. Things you scheduled. Reminders and routines run on a timer, the same idea as a cron job, and fire at the time you set.
2. Conditions it is watching. A watcher re-checks a page, a price, or a topic on an interval you pick, and messages you only when your condition becomes true.
3. Context it already has. Because a good assistant keeps memory of what matters to you, it can connect a change to something you care about, instead of pinging you about noise.

The key design principle is that a proactive message should always trace back to something you asked for or a rule you set. It should never feel random.

Can an AI actually call me, not just text?

Yes, and it is the right tool for a narrow set of moments: a wake-up call, an urgent reminder, or a nudge you cannot ignore. folk can place a real call to your own verified number, in a voice you pick, and you can call it back and it will pick up knowing who you are.

The important guardrail: it only ever calls your own number. It cannot call your friends, your ex, or a restaurant on your behalf. See folk calls you for how that works.

How do I keep a proactive assistant from spamming me?

This is the real fear, and it is fair. A proactive assistant is only good if it respects your attention. Look for three things:

• A cap on how often it reaches out. Nudges should be rate-limited, not constant.
• One-word control. You should be able to say "stop the nudges" or "snooze until next week" and have it obey immediately.
• Everything opt-in. The only guaranteed messages should be the reminders you explicitly set.

folk is built this way: nudges are capped, easy to snooze or turn off, and you stay in control from any chat. See the reminders and nudges controls.

The bottom line

An AI that only answers when spoken to is a search box. An AI that can text you first, responsibly, is closer to an assistant.

The trick is proactivity with restraint: it reaches out for the things you actually asked about, and it goes quiet the moment you tell it to.

If you want to try it, get started with folk and set one reminder or one watcher. That first unprompted "hey, the price just dropped" is when it clicks. For the bigger picture, see the best AI assistant that lives in your texts.