We ran a social media agency for 2 years. now we replaced ourselves with ai here's what we learned.

quick context: before publy.ch, my co-founder patrick and i ran a social media agency for two years. wrote thousands of posts, managed accounts for smbs across switzerland, the whole thing.

it was profitable. but two things kept bugging us.

1. clients couldn't afford us long-term

every client loved the output. then 6 months in, they'd ask: "is there a way to do this for less?" we'd lose them to a freelancer who charged half. or worse they'd just stop posting.

the demand for on-brand social content was massive. our pricing model was the problem.

2. we tested every ai tool and none of them worked

when chatgpt blew up, we tried to integrate it into our workflow. then we tried jasper, copy.ai, every "ai social media tool" on the market.

all of them produced the same flat, generic output. our clients would read it and say "this doesn't sound like us." they were right.

the reason: every tool used a single prompt for everyone. no actual voice capture. no understanding of the business. just "write a linkedin post about X."

so we built publy.ch

we took everything that worked at the agency the briefing process, the voice profiles, the content frameworks and built it into software.

the difference: instead of one prompt for everyone, publy.ch builds a per-customer voice profile during onboarding. takes about 5 minutes. then every post generated for that customer pulls from their specific voice doc, not chatgpt's default.

result: output that actually sounds like the customer wrote it.

the numbers so far:

• launched 1 months ago

• bootstrapped, no funding

• main acquisition: cold email + linkedin dms (no ads yet)

what i'd do differently:

1. launched too late. we sat on the agency knowledge for 6 months "perfecting" the product. should've shipped at month 2 with ugly ui and 1 customer.

1. underpriced at start. chf 9/month felt safe. it attracted hobbyists who churned in 30 days. raising to chf 39 actually improved retention.

1. didn't track utm sources from day 1. wasted 3 months not knowing which channels worked. fix this before launch, not after.

1. kept the agency running too long. split focus. should've shut it down at month 3 to fully commit.

what's next:

• shipping team workspaces (multi-user with roles)

• aiming for 1000 customers by end of year

happy to answer anything including stuff i screwed up. ihers, ama away.

— dario, co-founder publy.ch