Samuel Harček runs a social media agency and coaches Slovak personal brands on turning content into paying clients. His Reels normally land between 2,000 and 7,000 views. One landed at 45,169 with 1,316 comments, about nine times his own average. He says finding the videos worth studying in each client's niche takes him one prompt and saves around five hours a week. This is what he does with the five hours: the anatomy of the outlier, and the loop that produces the next one.

The part nobody optimises

Samuel Harček is 25, based in Žilina, and coaches Slovak and Czech entrepreneurs on turning content into paying clients. He posts three to five Reels a week in Slovak. His bio makes one promise: he helps personal brands get attention and paying clients.

Ask most people what slows content down and they say writing. For someone who has been doing it for years, it never is.

I run a social media agency, which means I'm creating content for multiple clients at the same time, different niches, different voices, different audiences. The hardest part used to be ideation. Coming up with actual ideas that had a chance of performing, for every client, every week, without burning out.

Samuel Harček

Note which word he reaches for. Not slow. Not expensive. Burning out, which is what a standing weekly obligation across a client book actually feels like, and a different problem from needing a faster draft.

Vamos changed that. I can find viral videos in each client's niche with one prompt which alone saves me around 5 hours a week.

Samuel Harček

Five hours a week, on the finding step only, in his own estimate.

What he found, and what it did

In late March he published a 60-second Reel about a Claude update. It did 45,169 views against a baseline of 2,000 to 7,000, roughly nine times his own average, with 1,316 comments and 291 shares.

The comment count is the interesting number, and we will come back to why.

First, how the video is built. Six beats, and every one is doing a specific job.

Sixty seconds, beat by beat

0:00, two triggers in one sentence. He opens face to camera: “Here's what almost everyone missed in the latest Claude update, and the reason I'll never use Canva again.” That single line runs an information gap and a tool-abandonment claim at the same time. Either one would hold a viewer. Together they compound, and they fire before a second of demo has played. The same words appear as on-screen text, so the hook lands with the sound off.

0:07, the credibility bridge. One sentence before the demo: most people thought this was only for simple diagrams, but he found a trick. It reframes what the viewer already believes and pre-empts the “I knew that” dismissal. Short enough not to lose anyone, specific enough to earn the next ten seconds.

0:17, proof instead of description. Cut to screen recording. He types a prompt, and seven finished carousel slides come back. He says the number out loud. Countable output feels real in a way that “it generates carousels” does not. The viewer is watching the thing work, not hearing about it.

0:29, the objection killed on camera. He asks for a revision in plain language and the design changes. This is aimed squarely at the person thinking but what if the first output is wrong. It shows that fixing it needs no design skill, only a sentence.

0:41, the time anchor. Under a minute, against hours in Canva. Not an abstract efficiency claim: a task the viewer has done, priced in their own time. This is where a demo becomes a calculation about the viewer's own week.

0:49, the trigger, immediately. No summary. No thanks for watching. A card: want the exact process, comment CLAUDE. Then it ends.

Why the keyword was the right keyword

The trigger word is the name of the tool the whole video is about. By second 49 it is already sitting in the viewer's head. Typing it costs nothing: no recall, no decision, no scrolling back to check the spelling.

1,316 people typed it.

Each of those is a person who has raised a hand, and each one enters a DM conversation. But the comments do a second job at the same time: a post with that much engagement relative to its views gets pushed to people who do not follow him. The lead capture and the distribution are the same mechanism. Most creators pick a clever keyword and lose both.

One more thing the data shows, and it matters. His outliers are all AI-tool and Instagram-mechanics content. His lifestyle and personal-story posts sit at or below baseline. He is not guessing which lane pays. He can see it.

The loop, and the prompts that run it

One workspace per client. Each one holds its own My Brand: the offer, the audience, the voice and the Rules that client would never break. That is what stops one client's research turning up in another client's draft, and it is why the voice is standing context instead of an instruction retyped into every prompt.

I set the brand voice and tone per client, and the AI writes a full script that's about 90% done straight away.

Samuel Harček

That is the first draft. He still edits it:

I still make edits, and I love that it leaves room for the human touch rather than just spitting out generic output. It doesn't generate the basic AI fluff that most tools do.

Samuel Harček

Find the outliers, per niche. One prompt does exactly this: "Find the outliers in my client's niche from the last two weeks. Write the 10 scripts I'll pitch on Thursday." That is the one prompt standing in for the five hours.

Take the shape apart before you reuse it. The prompt: "Break down this account's last 30 videos. Tell me what to use and what to skip. Then write my version of their best one." It separates the shape of the video from the person in it. Their story is theirs. The shape is the part that travels.

Put the weekly pass on a schedule. The prompt: "Every Monday before my client call, check their account and their niche. Tell me what changed and write the three posts they should make next." As a Routine it runs per workspace and the answer is waiting before the call.

Export what you pitch. The doc carries the words, the shots and the on-screen text, and the tools on any selection sharpen a hook or tighten an opening without leaving it.

What this is worth being careful about

The five hours a week is Samuel's own estimate, for the research step alone, in his own workflow. We did not measure it.

The 45,169 view Reel is a genuine outlier at nine times his own norm, and outliers are outliers. His median Reel still does 2,000 to 7,000. Publishing the structure of the one that broke out is useful precisely because you can see what it did differently, not because the number repeats.

And this holds nothing back deliberately: the hook shape, the trigger mechanism and the prompts are all above. The scarce thing was never the method. It was the hours it took to find which video was worth taking apart.

Overall it's shortened the entire process, from finding inspiration, to ideation, to scripting and refining, in a way that nothing else has.

Samuel Harček

Samuel's customer story: Samuel Harček.