❗️Until yesterday, I thought I was a complete noob in AI.
Now I don't think so. I'll try to tell how I got a promotion 😅

A friend invited me as a +1 to an open event of one of the well-known business clubs in St. Petersburg. The club is decent (reviews are really good) - I checked the conditions on the website: to join, you need a personal income of at least 5 million per year, company revenue of at least 30 million, a team on staff, an entry fee of 400k, and an annual subscription. In short: not riffraff, not gypsy info, not experts. In my opinion, the people in the room are business owners who count money and value their time.

The theme of the evening was "AI and neural networks for marketing, sales, and personal brand." Well, that's definitely about me, right? I thought and gladly agreed to go to the meeting.

Their new resident was speaking. Quote: "Marketer, expert in digital, neural networks, and business automation." I won't name names - no advertising, no anti-advertising. As a speaker, very lively, energetic, confident. And the presentation could have been interesting - customer portrait, customer journey map (CJM), delegation to AI, personal brand, marketing plan.

🔧 And then the tools started...

First. From the entire market variety, she recommended only what works without a VPN - Yandex GPT and DeepSeek. Okay, I understand the logic - accessibility is important. But she completely ignored what has happened in the last three months (at least). And quite a lot has happened: Claude Code + the agent system that was rolled out literally yesterday, GPT learned to work with documents at a level that didn't exist before, NotebookLM... well, the list is really wide. All these models either work through a simple browser or through a basic workaround. She didn't say a word about it.

Second. The prompts she showed were from the series "write me a sales text about my product, target audience - women 30-45." This is the level of early 2025. Today, normal work with a model involves roles, context, query chains, iterations. Models have long stopped guessing - they need to be explained. Her approach would produce exactly what everyone saw in feeds a year ago - AI content that no one reads.

Third. She showed generated videos as cool cases. I looked at the screen and thought - this is such crap, like spaghetti and six fingers. Artifacts, blurry background, length of 15 seconds. Today, VEO, Sora, Kling, and Hailuo do this orders of magnitude better. The people in the room seemed not to know this - they were greedily photographing the slides in the moment.

Fourth. The final touch - of course, the presentation was generated by AI, with errors she didn't check. Funny moment: in an example poster for a conditional LLC "PetroStalEnergo," the banner read "PetroStalEnergo" (misspelled). I wanted to shout it out, but I held back and we pretended no one noticed.

And here I want to say something important. The speaker's ideas themselves were correct and logical - customer portrait, customer journey, delegating routine to AI. This all works and is needed. The problem is not the ideas - the problem is the tools she suggested to implement these ideas. Trying to save a measly $100-200 per month on normal products leads to exactly what we see every day - gray, faceless, identical content that everyone has learned to ignore. And if for a personal blog it's just a shame, in business it's critical. Because in business, you build loyalty over years, you convey your values through every point of contact with the client over years. And then, into this reputation built over years, you insert a banner with a typo generated by a free model. For the sake of saving or speed.

This applies not only to content or analysis. The same story happens in products. Every day I see landing pages and SaaS projects made with vibe coding - they are visible from afar. I use them only if there is a direct acute pain for a specific problem and there are no alternatives. In other cases, I skip them as another dead project. And most people do the same - I'm sure.