Case study · own platform, with a partner · Austria · only4sex.at · 18+
An Austrian directory of venues in a regulated adult market, run by Meep Tech and a partner. It started as an old-school site where every listing was typed by hand. Now scrapers, a duplicate engine and a review queue keep it current — and a person only handles what the system isn’t sure about.
Venues change their line-up all the time on their own websites. Typing those changes in by hand doesn’t scale past a handful of venues, and the same person often appears at several venues under slightly different names — so a hand-made directory is both out of date and full of duplicates.
Four stages, each placed on the Autonomy Ladder. The system is at L2 overall: it runs on its own, and people review only the matches it can’t decide.
Every venue and every profile was typed into the site by hand, and went out of date as soon as the venue changed its own page.
67 venue-specific scrapers, each reading one venue’s own profiles page. They run every Monday and Friday, or on demand, and bring in 100+ new profiles per run.
A personAdds a scraper when a new venue joins.
The same person appeared several times — once per venue, with slightly different names and photos.
Profiles whose photos match by hash merge automatically, on import and again in a nightly job — about 40 of the 100+ new profiles in each run. Anything uncertain goes to a review queue instead of being guessed.
A personClears the review queue: Merge or Keep both.
No consistent rule for which sources could be imported.
Sources that put their own age verification or similar gate in front of profiles are skipped by design. The policy lives in code, not in someone’s memory.
A personNothing.
No overview of what had changed since the last update.
After every run the admin receives one digest email listing the new, unchecked profiles. AI tools can read the live data through a read-only MCP server to answer questions about it — they can check, never change.
A personReads the digest.
Search Console, Analytics and Microsoft Clarity heat maps showed where visitors looked and where they gave up. That led to mini galleries inside previews, a prompt offering alternatives when someone scrolls past a profile, and an endless feed of similar profiles in place of a fixed list.
A script that filled in missing venue data picked up favicons and tiny icons as venue logos, and mistook single-profile pages for venue pages. Those fields were cleared by hand and locked, and the script now only fills fields that are empty. An AI-based scraper for the same job filled a good share of the gaps but made too many mistakes to trust, so it isn’t used. Bulk AI-written profile text was considered and deliberately paused.
Automate what can be checked, and send everything else to a person.
Next step
Book a free 30-minute Autonomy Audit. We put your recurring processes on the ladder — no deck, no homework.
