We used gunj to support our own founder's LinkedIn engagement for 150 days. Here is what it produced.
Before onboarding outside users, we used gunj to support our own founder's engagement every day, drafting in his voice. This is one account, not a customer cohort, and the numbers come straight from gunj's production data.
Five months of supported engagement
*May and October are partial months (started 25 May, data through 21 Oct).
The account grew alongside the engagement
Follower growth
6,381 to 8,846 followers (+2,465), alongside the founder's own posting.
LinkedIn SSI
Social Selling Index 62.2 to 67.7, driven by the brand and audience pillars.
Invite acceptance
1,164 of 1,907 invites accepted.
ICP-fit leads surfaced
From 10,299 prospects scored against the ideal customer profile: 16 hot now, 591 future fit.
Over 16 work-weeks the founder did not spend
gunj helped draft and deliver 22,242 comments, replies and DMs, each written in the founder's voice. At a conservative two minutes of read-and-write time each, that is about 741 hours. After subtracting the founder's own time approving each draft (about 15 seconds each), the net saving is about 650 hours.
What made it safe to run daily
Golden-hour slot engine
A post's first 60 to 90 minutes decide its reach. The engine watches each creator's posting window and comments inside it instead of hours late.
Draft, critic, regenerate
Every comment is scored by a second model before it posts. Anything under the bar is rewritten, and hard rules strip the tells of AI writing.
Per-account isolation
Every comment, reaction, invite and DM goes out from that creator's own linked LinkedIn account, never a shared one.
Human-paced activity
Activity is paced across the day and kept within LinkedIn's limits, rather than sent in bursts.
Duplicate protection
LinkedIn exposes one post under two ID formats. We check both, so the same post is never commented on twice.
Reciprocal engagement
People who comment on your posts are picked up automatically and engaged back, so the relationship loop runs without tracking.
Queue-and-worker architecture
Work runs through queues and workers, so one stuck job cannot silently drop a day of engagement.
Methodology and caveats
One account. Every figure is from a single founder account, 25 May to 21 Oct 2026, from gunj's production database. It is a dogfooding run, not an external customer cohort. Counts are completed actions only.
Follower growth. The founder also published his own content in this period. Engagement contributed to the growth, but it is not the only cause.
Founder approval. The founder approved the drafts before they went out.
Time saved. The hours figure is an estimate: 22,242 written interactions at 2 minutes of read-and-write time each, minus about 93 hours of founder approval time (15 seconds per draft).
SSI and followers come from LinkedIn's own metrics, captured as periodic snapshots. Your results will vary with your audience, content and starting point.
Want this running on your account?
gunj is invite-only right now. Tell us about your account and we will show you what a daily engagement routine would look like.