Managing LinkedIn account with operational controls
Salesbot helps manage risk with conservative cadence, daily limits, an emergency stop and human review of AI actions. No tool can guarantee how LinkedIn will evaluate account activity.
Controlled spacing between actions
Even when multiple messages arrive at once via MCP or a bulk import, they are not sent in a burst. Every action is scheduled with a conservative cadence and within configured limits.
Randomized jitter between actions
A random offset is inserted between individual messages, connection requests and profile visits. No two actions happen in the same second or at a regular interval.
Allowed working hours
Actions only run during the hours the user configures (typically 9 AM–5 PM in their timezone). Outside that window actions are deferred to the next working day.
Daily limits per user
A hard cap on the number of connection requests, messages and profile visits per day. Once exhausted, the rest is moved to the next day.
Gradual ramp-up
A newly connected LinkedIn account starts at 10 % of the limit and scales gradually (10 → 30 → 60 → 100 %) based on connection age. This helps avoid sudden activity on new accounts.
Fatigue model
We track account "fatigue" (activity_streak_days, daily_fatigue). If the user runs at max for several days in a row, the system slows down on its own.
Cooldowns between action types
Minimum delays between connect → message → follow-up message keep the action sequence controlled and conservative.
Detection of unreachable contacts
3rd-degree / out-of-network leads are not retried indefinitely – after a failure they are paused for 7 days and only then re-checked. This saves actions and supports conservative operation.
Emergency Stop
The user has an "Emergency Stop" button in the UI that instantly halts all scheduled actions across every campaign.
Human-in-the-loop
- AI drafts go through an approval step, giving the user a review option before sending (MessageApprovalDialog).
- Optional auto-approval with a second check (GPT-5-nano verifies the Gemini draft, 1 retry with feedback).
- Every action has a trace in the DB of who/what triggered it (user × MCP × cron).