AI Experiments that Actually Worked for Me in 2025

I’m still a true believer that AI will be our helpful assistant within work and life. And I always like experimenting and trying new stuff. So, this article is dedicated to everything I found genuinely helpful this year. ✨

One of the most important lessons this year was learning to ask AI to refine or generate the prompt before creating assistants, agents, or chats. It sounds simple, but it really works. 😀

ChatGPT Projects 

I like staying organized and when I was using ChatGPT I felt frustrated because I couldn’t highlight the important chats for me. Eventually, I found a workaround: structuring everything into projects.

I structured everything I used and provided dedicated instructions for each project. It helps me keep history in different categories.

For example: 

• Blog – an editor that helps polish and review my texts.
• Professional path – a basic mentor that knows my story and encourages me when I’m stuck or not confident.
• Health – not replacing doctors, of course, but perfect for choosing shampoos 😀 or finding alternatives for skincare products.

I also have a “Home assistant” project that helps me cook with ingredients I already have or create a weekly menu.

Perplexity Explorer

Perplexity surprised me with its topic-exploration features. What I found most valuable is the Tasks feature – scheduled assignments. I set up two tasks:

Gather weekly trends, experiments, and tools related to AI in marketing 

Here is an example of instructions: 

And the format of news – clean, structured, plus you can instantly turn the results into a table: 

This was genuinely useful. Receiving weekly updates that highlight why a tool or experiment matters for marketing is a gem. It saves me a lot of time processing huge amounts of information.

Job search (failed)

The second experiment was not so successful, but maybe I should just refine the instructions more carefully. 

I asked Perplexity to find newly published remote jobs in Europe daily, but the results weren’t great: repeated listings, closed positions, and lots of information I didn’t need (like salary trends or CV advice). 

I experimented for a week, changing instructions, but it still wasn’t helpful.

Gemini Gem Bot

Next was creating Gemini Gem Bots. 

I started with a practical assistant – a Career Consultant. I added my experience, expectations, and principles for career growth, and asked the bot to act as my advisor.

Each time I want to apply for a new position, I simply ask if it’s a good fit. 

What I like about Gem Bot is that it correlates my experience with the role, highlights strengths and growth areas, and warns about restrictions (for instance, if the position is limited to Germany only).

Gemini App 

I also played with Google AI Studio and created an app – Career Path – which can help craft individual development plans. Not sure if it will work on your side since I wasn’t planning to publish it, but here it is if you want to try.

Upload the resume and app reads it:

Customize your learning style, write down areas you want to grow in, and that’s it – a personalized growth roadmap is created.

Then AI generates individual recommendations:

Comet Assistant

I saw several experiments from marketers who analyzed websites and generated improvement suggestions using Comet Assistant.

I found a particularly useful editor-helper approach: I asked the assistant to act as an editor directly inside my document, without losing my style. It can either make real-time edits or show suggestions. I prefer suggestions so I can review and accept them.

Works like magic 😀 

And with a refined prompt, it gets even better. 

NotebookLM

This is one of the top tools for me. You can upload files and do anything: create mind maps, infographics, podcasts – a truly helpful resource for learning, marketing analysis, or simplifying complex information.

Nano Banana Animation

I enjoyed playing with Nano Banana too, but I’m a bit tired of the overwhelming amount of generated videos and images everywhere, so I didn’t use it as much.

Sum up

AI is clearly here for the long run, and ignoring it only puts you at a disadvantage. I explored experiments from different tools but always focused on the ones that are practical specifically for me.

I also want to mention one of the most valuable courses I took this year: AI for Business from Google (I’m taking part in the Ukrainian version, but hopefully there’s an equivalent “AI for Small Business” course). 

It’s incredibly practical, structured, and genuinely useful – one of the best I’ve taken in a while.

What about you? What was your most useful case of AI?