ai for small business: 6 ideas
AI Agents: It’s just a buzz word
AI agents stand out because everyone gets jammed with potential ideas but a lot of them are not executable.
To me, this “AI Agents” wave reminds me of when WordPress started exploding: many tools emerged because people couldn’t code.
People were paying for modal popups… remember those times?
Same trap these days if you don’t understand what you’re being sold.
At its core, it’s just a couple of calls to ChatGPT. That’s it.
Here are 6 ideas that i’ve executed and adopted in my life.
ai for small business: 6 examples that works
1) Voice recording to blog
Actually, this blog was generated by a script that takes my voice, transcribes it into text, then passes it through stages of prompts to finally produce an initial draft that I start with. Then it generates a banner for it.
The trick to making this work is to have the shape of your blog in place.
It’s not the same thing to write a tech blog or a story. The same applies to the reader—it’s not trivial to write in a way that readers enjoy, and that requires basic knowledge first.
The way to go about this is to simply have one script per style with the prompts directly in the code. If you describe to the LLM how to write the blog with the keywords and the transcription, it does it pretty well.
The issue is in not enforcing the system to write the post.
If you worked with a copywriter on a successful blog, you know that the first step is to create a basic template so that they can be onboarded faster.
Here it’s the same thing—you need to onboard your AI agent like you would with an employee, except that it gets it in one iteration.
2) Goal Setting with Personal OKR Tools
The biggest issue I see in medium/small businesses is internal organization of current efforts.
And the bigger the team gets, the worse it gets. Some people only follow goals in their work, very few have the discipline to follow a plan for their dreams.
Personally, I do have an OKR system to keep me accountable.
Before AI, I used to spend one day a week reflecting and projecting for the next week.
Reordering, updating, reformatting text, making sure it makes sense to me.
And if I’m in a manic phase with a lot of ideas, it’s worse—I just lose track of my goals and the cognitive load goes through the roof.
My system is:
- Define the 25-year vision once
- Make the 90-day plan that gets you closer to that vision
- Create the weekly plan to start shredding that current 90-day goal
- Assess progress daily and report/eliminate tasks that are not valuable
Then every 90 days, I look at every week and make sure I’m eliminating inefficiencies.
All of this is done on paper following a couple of open-ended questions for me to reflect.
Now the current setup is a Notion database that has the 90-day goals and my weekly tasks that have to be done. There is an automation with AI that can categorize the data in respect to the 90-day goals, advise and propose a prioritized plan with a justification of its decisions.
This has been a major improvement in my project momentum because thanks to this, I’m still aligned with my vision.
3) Icebreaker in cold emailing
This is the perfect case of simplification I had.
I had a CSV with many contacts and URLs to get context from websites. The automation was simple: for each line, create a message using the data.
It failed miserably.
Here’s how I fixed it:
- Get some initial calls to understand how to hook attention and understand customer problems
Then I realized that because I was dealing mainly with warm leads, I just had to create three personal message templates depending on the lead source. Instead of writing whole new texts, I could simply insert the right template based on these variables.
Then I used AI to make sure the text made sense.
In this case, AI can be used to evaluate if a lead is from XYZ source given some context: you could let AI decide if the lead matches the ICP and then stitch the pre-written icebreaker into the email.
4) Videos to Notes
Super simple automation: transform text from a YouTube video and save it in your second brain / note app.
It might sound stupid, but if you’re a content creator and you save videos for inspiration, you can end up with a list of many links and notes that are just unusable to search later.
With more systems around it, you can easily ask AI to only extract topics among a certain set of videos to prepare brainstorming for your next topics.
5) Decision maker coach
Sometimes I just get stuck in life because I’m afraid to make a decision or can’t commit for an unknown reason.
LLMs are great at providing information back in an unbiased way.
For this, I follow a 12-step question chat that adapts to the topic to evaluate pros and cons, as well as project worst-case scenarios.
At the end, it writes a report that explains what’s happening in my mind with respect to that decision.
6) Meetings to Video Content
Every two Fridays I have an online call with friends. So many ideas and fun things are said there. We are writing an automation to chop these into short-form content.
Then it just puts both people’s faces on top of each other and adds the transcript.
Here AI is used to transcribe the text at the right location and find interesting bits, then just overlay the text together.
The thing here is that we’re able to reinvest time into content without extra effort.
The hack: Ai & habits
You probably have habits or stuff that you do that can be turned into content. If you simplify the process, add code and AI to leverage your flair at scale, you can really multiply your efforts.
In my case, I’m good at talking to people in meetings. So speaking in a mic is easy for me and made it the core of content creation.
This also aligns with other meetings with friends where I can convert that into shorts and YouTube videos.
But system first—if you never made a short or a YouTube video, you will get lost.
Conclusion: What do you need?
Everyone is trying to build a one-for-all platform—the next “sell the shovels” episode of the internet.
Focus the problem: Make your systems time efficient
- Write your current business on paper, step by step
- Look at each step and list the tasks
- Prioritize the tasks that can be automated
- Add AI to either: judge, report, suggest next steps, draft, etc.
- Test ASAP
Here are 3 things to consider when experimenting with agents:
1) System over AI
A well-orchestrated system that can save and report data has more value than just giving everything to an LLM.
LLMs are supposed to stitch text, judge a text, or get information from an unstructured data source (i.e., just an audio recording or image)
But you can’t ask AI: “please organize my life…”
Create the system, automate it, use AI to round edges to get from 70 to 95% efficiency.
2) If you can code, it should not take more than a day
Using AI can make you feel like a wizard. However, excitement and magic do not drive profits.
These proof-of-concept scripts should be developed during “fun days” and tested quickly.
3) Keep it local
Don’t start shipping code on a server.
Just create a simple script and test you idea in 4h.
You can also learn https://jupyter.org/ to run code.
It’s all system and strategies to solve a problem.
Focus that first.