If you have ever watched a LinkedIn account get restricted after a tool sent connection requests all night, you already know the lesson. Safe automation is not about hiding what you are doing. It is about steady pacing and simple habits so your activity still looks human while you keep new conversations moving.
Why full blast fails
LinkedIn does not publish one public scorecard for every limit, but patterns still matter. A burst of invites at odd hours, a wall of pending requests that never clear, and the same timing on every action all make sloppy automation easy to spot.
Tools that tune for raw volume often push you straight into that pattern. You might get a short spike in invites sent, then spend weeks winning trust back on the account. I designed Flow AI the other way: fixed caps, predictable hours, and spacing so outreach stays steady instead of noisy.
Limits, timing, and pacing
In Flow AI, lists power Auto-pilot. A list is your campaign. When a list is on, the product runs the outreach steps under rules we bake into the code. You cannot crank them higher than we allow.
Per LinkedIn account, each day allows up to 15 connection requests, up to 80 post likes, and up to 80 profile visits. Campaign activity only runs between 9am and 6pm in that account's local time zone. Actions are spread so the average gap between touches is about 15 minutes. That cuts the "machine gun" feel you get when everything fires at once.
Your dashboard shows usage against those limits so you can see where you stand without guessing.
Warm-up and withdrawals
Brand-new sender accounts do not jump straight to 15 connection requests. Warm-up ramps day by day: day 1 sends 1 request, day 2 sends 2, and so on through day 15, when you reach the full 15 per day. That gradual curve is built in for accounts using the product for the first time.
We also withdraw a connection request if it is still not accepted after 21 days. Too many lingering pending invites is another pattern that looks suspicious, so we clear them instead of letting the pile grow forever.
The sequence has several steps (profile visits, likes, timed waits, then the connection request). You do not tweak every tiny step in the UI. The product runs the same path for each person on the list so behaviour stays consistent.
Human review and scaling safely
Auto-pilot handles the steady rhythm of engagement and connection requests. When you move into actual DMs, Flow AI is built so you review before send. AI can draft a reply from the prospect's profile, conversation history, and your offer context, but the message goes out when you are happy with it, not on a blind timer.
That split matters for safety and for quality. Inboxes reward messages that sound like a person decided to write them, not a script that fired because a step counter hit zero.
When you need more daily connection capacity without piling risk on one profile, add more team accounts. Multiple senders split prospects across teammates with no overlap: two people should not chase the same contact. Total safe capacity grows with the number of active accounts (15 connection requests per account per day, depending on how many lists you run and warm-up state).
If you want to try the same guardrails we ship to customers, you can Try Flow AI free and see limits, timing, and warm-up on your own workspace.