AI Didn't Save My Business. Here Is What It Actually Did.
There are two conversations happening about AI right now and neither of them is especially useful.
The first is the hype version.
AI is going to replace everything, you need to adopt it immediately, founders who are not building their entire workflow around it are already falling behind.
The second is the backlash version.
AI produces generic, flat content that sounds like everyone else, it is overrated, the real work still requires a human.
Both of those things are partially true. Neither of them will help you figure out where this tool actually fits inside your specific business.
So this is the honest middle.
What AI has genuinely changed about how I work, where I pulled back from it, and what I figured out when I came back with different expectations.
Why I pulled back:
When AI tools became genuinely usable, I used them the way most people do at first. As a generator. Write this email. Draft this caption. Give me an outline for this offer.
And it worked, in a surface level way. The output was fast. Sometimes it was good. But I kept noticing a quality to the writing that I could not quite leave alone.
A certain professional smoothness that felt like it had been assembled from a lot of content rather than lived inside a business. Clean sentences that said the right things but lacked the specific texture of how I actually think.
The more I used it for client-facing writing, the more time I spent editing the flatness out. And at some point I realized I was spending almost as much time fixing the AI output as I would have spent writing something myself. Just with worse raw material to start from.
So I pulled back. Not from AI entirely, but from using it as a first tool for anything that needed to sound like me.
What changed when I came back:
When I came back to AI, the thing that was different was what I was asking of it.
The first time around, I was asking it to think for me. To decide what to say, how to position things, what angle to take. And that is exactly the part of the work that AI cannot do well.
Not because it is not capable in a technical sense, but because it has no context. It does not know my clients. It does not know the conversation I have been having with my audience for the past year. It does not know what I noticed on a client call last week that is sitting with me and needs to come out somewhere.
AI can execute. What it can't do is decide what is worth executing. And the moment I stopped asking it to make that decision for me, the whole relationship with the tool changed.
Where AI actually earns its place
I want to be specific here because vague endorsements of AI are not useful to anyone.
The first place AI genuinely saves me time is research.
When I need to get oriented on something quickly, AI gives me a structured starting point that would have taken thirty minutes to assemble from scratch. Not a replacement for going deeper, but a fast first layer.
The second is getting unstuck.
When I know what I want to say but cannot find the opening, talking it through with an AI tool is genuinely useful. Not because it gives me the answer, but because articulating the thing I'm trying to say, even to a tool, helps me find it. It's thinking out loud with something that reflects it back in a different structure.
The third is functional writing.
Things that need to exist and need to be accurate but do not need my voice. Confirmation copy. Process documentation. Structural emails that are informational rather than relational. AI handles those well and I do not feel like I'm losing anything by letting it.
The fourth surprised me: stress testing strategy.
I will share an idea and ask AI to push back on it. I will describe a positioning direction and ask what I am missing. It does not always catch something I had not already sensed, but sometimes it names something I had been circling without landing on. As a thinking partner for testing ideas I am already holding, it is genuinely useful.
Those four things.
Research, getting unstuck, functional writing, and stress testing.
That is where AI earns its place in my workflow. Everything else stays mine.
What AI can't do, no matter how good your prompts are
This is the part that matters most and the part that gets the least attention in most AI conversations.
AI works on the surface of whatever you show it.
It does not see underneath. It does not know the history of your business, the context that makes certain things land with your specific audience, or the structural issues that are causing the problems you are asking it to solve.
If your onboarding is creating friction, AI will help you write a cleaner version of a sequence but if the flow is jammed, new copy solves the wrong problem.
If your sales page is not converting, AI will rewrite the copy without noticing that the positioning has drifted from what your actual clients are currently buying.
If your emails are not landing, AI will suggest new subject lines without recognizing that your audience has outgrown the kind of content you are sending them.
In every case, AI produces a better version of what you gave it. Which is only useful if what you gave it was already pointed in the right direction.
The founders who get into trouble with AI are not the ones using it badly. They are the ones using it to move faster inside strategies and systems that were already not quite working.
AI accelerates what is already there. If what is already there is solid, that acceleration is genuinely valuable. If what is already there has gaps, AI helps you produce more content with those same gaps built in.
The brand voice problem nobody is really naming
There is something happening quietly inside a lot of established brands right now that I think is worth saying directly.
When AI does too much of your content writing, your voice starts to drift.
Not suddenly and not obviously. But over time there is a sameness that settles in. A certain evenness of tone. A tendency toward complete, well-structured sentences that feel professional but lack the specific texture of how you actually think and talk.
Your audience knows your voice. Not always consciously, but they have been reading your emails and seeing your content long enough to have an internal sense of how you sound. When that starts to shift, even subtly, engagement changes. Not dramatically. Just slightly. And slightly is enough to matter over time.
The solution is not to stop using AI for content. It is to stay in the thinking seat. You decide what to say, what angle to take, what is true about where your audience is right now. If you then use AI to help with execution, you edit the output back into your voice rather than publishing what it produces.
The difference between AI as ghostwriter and AI as first draft is the difference between outsourcing your thinking and outsourcing your typing. One costs you your brand. The other just saves you time.
The one question worth asking
Rather than a framework or a list of approved use cases, I want to leave you with a single question to ask yourself every time you reach for an AI tool.
Am I using this to move faster on something I have already thought through? Or am I using this to avoid thinking through something?
If it is the first, AI will probably serve you well. You have done the work that matters and you are using the tool to execute it more efficiently. That is a genuinely good use of a genuinely useful tool.
If it is the second, pause. Because what you are about to produce will be faster but it will not be better. And in a lot of cases it will create more work later, when you have to go back and address the thing you were avoiding in the first place.
AI is not going anywhere. The pressure to use it more and build your whole workflow around it is not going anywhere. But your thinking, your voice, and your judgment about what your clients actually need are yours. No tool replaces those things.
Use the parts that help. Leave the parts that cost you. That is honestly the whole strategy.
🎧 Listen to Ep. 177: AI Didn't Save My Business. But It Did Change How I Work.
