Last week someone asked me how they can use AI to grow their business.

I asked what his biggest problem was, what’s the bottleneck.

He couldn't answer… (he actually said “I want to be more efficient“… Like everyone)

That's the problem, and it's everywhere.

Here's how most people adopt AI: they see a tool trending on X or Instagram, they try it, they play with it for an hour, they think "cool," and they move on. Then they see another one next week, and they repeat the process.

Start by deeply understanding your business. Run this quick self-consulting exercise 👇

The following 6 paragraphs state introductory level AI use cases. I actually hope you already knew all of them. If not then you have a low hanging fruit you can apply now.

You write one post or newsletter, then manually turn it into content for other platforms.
Use AI to instantly rewrite it into a Twitter thread, LinkedIn post, or carousel.
Same idea → everywhere, in minutes instead of hours.

You spend 30–60 min making one thumbnail or visual.
Use AI to generate multiple designs from a prompt and pick the best one.
More options, way less time.

You copy-paste data from different tools to make reports.
Use AI to pull and summarize everything automatically.
You stop collecting data and just read insights.

You open tons of tabs to understand competitors or a market.
Use AI to summarize everything into key points in one go.
Hours of reading → a few minutes.

You write cold messages one by one and try to personalize them.
Use AI to generate custom intros based on someone’s profile.
Still personal, just way faster.

You need a dev to build a simple tool or test an idea.
Use AI to generate a basic version or prototype yourself.
You go from idea → something real in a day.

After a few months they've tried 20 tools and nothing has actually changed about how they work. Their output is the same. Their bottlenecks are the same. They just have a longer list of accounts they've signed up for.

We've tested over 500 AI tools in the last two years. Want to know how many we use regularly? Less than 10. The classic LLMs with solid skills/memory/project setups, our IDEs to build software, a couple image and video generators and that’s pretty much it. Here and there one for a specific use case, but on a daily/weekly basis that’s it.

The reason isn't that the other 490 were bad. Some were great. They just didn't solve a problem we actually had.

Most people pick AI tools like they're browsing Netflix. What looks interesting? What's getting buzz? What did that guy on Twitter say was a game-changer?

That's the opposite way of doing it.

Imagine walking into a hardware store with no idea what's broken in your house. You'd just wander the aisles buying whatever looks useful. That's how most people shop for AI tools.

The people I know who are actually getting results? They start with the bottleneck. Not "what AI can do" but "what's slowing me down." The specific task eating their time, the process that feels way too manual, the output that takes 3 hours when it should take 30 minutes.

They get clear on the problem. Then they find the tool. In that order.

Before I test any new AI tool now, I run it through three questions. These three questions have saved us more time than any tool ever has, because they've saved me from wasting hours on tools that don't matter.

Three questions. You can run through them in 30 seconds, and they've killed more bad tool adoptions than I can count.

Most tools don't survive Question 1. The ones that do but fail Question 2 go on a "maybe later" list. The ones that pass all three are the ones I actually install and keep.

This is how we went from 500 tested tools to the 6-7 we actually rely on. The filter did more for my productivity than any individual tool.

Audit your workflows. Be granular. Say you want to improve your Marketing and decide to map what the actual goals are.

Don’t write:
Marketing goal: getting customers —> prompt ChatGPT “How to get more customers?“

Write:
Marketing goal: getting customers —> how do I achieve that? —> cold outreach —> how do I achieve that? —> writing LinkedIn dms —> how do I achieve that? —> crafting a good message —> how do I achieve that? —> research to craft the right message.

NOTE: an arrow doesn’t mean that’s the only possible action to achieve the previous step. Your sheet of paper or whiteboard will end up looking more like a tree than a linear schema. For instance, warm outreach is another way of getting customers. Running ads is another one. Etc.

Then you understand that to “get more customers“, you go over at least 5 different subtasks. All of them can be sprinkled or replaced with AI.

Just reply to let me know any specifics. This newsletter is built on what you actually need.

If this was useful, forward it to that friend. Yep, the one that just popped into your mind.

Stefano (@stefano_martell on X)
Co-Founder, The AI Leverage by Bestapps.ai

PS: Reduce your Claude Code token consumption by 89% with this