Meetup recap: four automations, one funnel - live with Growgo
No slides, nothing pre-rendered. At the first Growgo x TheGrowthLabs meetup we chained four live automations into one competitive-intelligence funnel - and put a price tag on every piece. Here's what happened.
Picture a webinar where nothing is pre-rendered: no deck, no screenshots of results from last week. Just a live Claude session, a brand name typed in, and sixty minutes to prove the thing works in front of everyone watching.
That’s the format we committed to for the meetup series we run with Growgo, the Polish AI training company run by Jan Rudziński - a co-CEO who has trained over 7,000 people in AI, with a background running brands at Carlsberg and advisory teams at Deloitte. The first session ran on June 17. Here’s the recap - and the full recording (in Polish):
Prefer YouTube directly? Watch meetup 1 here.
The opener: six levels of working with AI
We started with a framework we keep coming back to - six levels of working with AI, from chatting with a model all the way to full autonomy, each with a rough output multiplier attached. The honest part: we operate at levels 3 to 5, most companies we meet are still at level 1, and level 6 belongs to OpenAI, Anthropic, and Google for now.
That framework deserves more than a slide-length mention, so it gets its own write-up on this blog. The rest of the evening was about proving levels 3 to 5 live.
Automation 1: Facebook Ad Library teardown
We fed a raw Claude session nothing but Żabka’s Meta Ad Library URL. No context loaded, no warm-up.
It scraped 86 ads through Apify, clustered near-identical creatives into groups, read every image with vision, scored each cluster, and used the UTM parameters buried in the ads’ links to place each one in the funnel - prospecting versus retargeting. On top of that: estimated reach and demographic splits per creative.
The moment that convinced the room: it recognized Krzysztof Ibisz in a creative, verified his Żabka campaign on its own, and rated the ad 5/5. It even caught a He-Man creative and connected it to the film currently in cinemas - nostalgia equity for a protein product, its words, high confidence.
The economics are the real headline. Scraping costs $0.075 per thousand ads searched. The report you just read about: roughly 10 cents. A full sweep of 15 brands: $1.13. Competitive ad intelligence used to be an agency retainer; now it’s pocket change plus an hour of setup.
Automation 2: LinkedIn research
Next we built what we call a local Sales Navigator - LinkedIn research through the anysite connector, driven in plain language.
One query returned around 50 marketing and sales decision-makers - team leads, heads, directors, CMOs - at Polish companies with a digital product. Then, dictated live on the call, a second query pulled Żabka’s whole team into one document. The same setup enriches profiles with recent posts, skills, and contact details, and can follow a person beyond LinkedIn - find them on Instagram or X, check if they run a newsletter. Up to a thousand profiles per query.
Price check: a dedicated LinkedIn scraping tool runs $120 a month. This stack - Claude Pro at $20 plus the connector at $30 - is $50 a month, and it does more.
Automations 3 and 4: opinion radar for Instagram and TikTok
Social listening, but for expressed opinions rather than keywords. The test phrase: “AI zabierze nam pracę” - AI will take our jobs.
The radar searches Instagram and TikTok for people saying that in a full sentence, returns the author, follower count, engagement, and a relevance score, filters to the last 21 days, and emails the report. One surfaced post made the room go quiet: companies that laid people off to replace them with AI are now quietly hiring them back, because fixing AI’s mistakes cost more than the savings.
One practical note from the demo: TikTok is great for lifestyle topics, weaker for substantive ones. Instagram carried the signal here.
The finale: a live integration, from zero
To close, we did the thing you’re not supposed to do in a live demo - built something with no pre-made skill behind it. We added a brand-new social-data connector to Claude from scratch: paste the MCP URL, authorize the account, and query Instagram with one plain sentence. Five minutes, no code.
That was the point of the whole evening, compressed: an agent is only as strong as the tools you plug into it. We don’t build from scratch - we plug into ready frameworks.
The bill
The full funnel shown that evening - competitor ad intel, lead research, social listening - runs on roughly $50 a month of tooling. The social listening piece was still on its free tier in month two.
At the end we announced something bigger is coming - a full course built around this way of working. More on that soon. If you want an invite to the next meetup or the recording of this one, get in touch.