When Not to Use AI: I Chose Human Voices on Purpose (Simply Explained)

The Feature I Could Have Added in Five Minutes (And Didn't)

I'm building a product right now. It's a memory book. Real printed pages, the kind that sit on your coffee table. Each page has a little scan code on it. Point your phone at it, and you hear a real person's voice and see short videos of the people who matter to whoever owns the book.

There was an obvious feature I could have added. AI that reads each page out loud in a human-sounding voice. It would have cost almost nothing. A fraction of a penny per page. And here's the kicker: I had already built the part that makes it work. The switch was right there.

I shipped the product without it. On purpose.

This isn't an anti-AI rant. I've built more than 15 AI systems for my own businesses and for clients. I run 29 automated tasks every day. I've saved over 3,000 hours a year by handing boring work to machines. I love this stuff.

But the real skill isn't knowing how to add AI. It's knowing when to leave it out. Almost nobody selling AI talks about that, because they don't get paid to say no.

Figure Out What People Are Actually Buying

Let me explain the memory book a different way.

You could describe it as "a book with audio." That's the wrong description. And the wrong description leads to the wrong decision every time.

What people actually pay for is to hear one specific person's real voice.

A grandparent reading a bedtime story. The way someone laughs before they finish a sentence. The accent. The little pause. The slightly wrong way they say a family nickname. Those aren't just sounds you can swap out. They are the whole point.

Now picture the AI version. It reads the same words in a clean, smooth voice. It might even sound beautiful. And it would ruin everything.

Because the customer isn't buying narration. They're buying proof that this real person sounded like this. A fake voice, even a perfect one, breaks that. It swaps the one thing that can't be replaced for a polished imitation.

That's the simplest test I know for when to skip AI. Ask what the customer is really paying for. Not what the product does. What it's for. Once you answer that honestly, some "obvious" features turn out to be quietly poisonous.

Why Cheaper and Faster Isn't Always Better

Let me argue against myself for a second, because the numbers are real.

The AI voice costs almost nothing. Nobody has to sit a relative down with a microphone or fumble with uploading a file. It works instantly across thousands of books with zero effort. On a spreadsheet, AI wins this in a landslide.

And for most decisions, cheaper-and-faster is exactly right. I make that call all the time. My product creation process used to take 3 to 4 hours per item. Now it takes 20 minutes. My pricing system automatically sets prices on more than 564 products. I don't get sentimental about saving time.

But here's the trap. A fake voice that sounds 95% like grandma isn't 95% as good. It's worse than nothing.

It lands in that creepy zone where "almost her" reads as "not her, and trying to trick me." That last 5% isn't a small gap you fix later. That 5% is the entire product.

This is the whole human-versus-AI debate boiled down to one example. AI wins when the value is speed, information, or volume. It loses badly when the value is a specific human being.

The person who automates everything destroys value in exactly the spots where being real is the point. They can't see it, because their only measuring stick is cost and speed. They're measuring the wrong thing.

Build the Switch, Then Leave It Off

Here's the part that I think actually shows good judgment.

I didn't skip the AI voice because I couldn't build it. I built the whole system so it doesn't care where the sound comes from. A real recording, an uploaded file, an AI voice down the road. To the system, it's all just audio that plays on a page.

Adding the AI voice later would take me a few hours. Most of the work is already done.

So this was never "we can't." It was "we chose not to." That difference is everything.

A product missing a feature because the team hit a wall has a gap. A product missing a feature because the builder deliberately turned off something he already built has a point of view.

I did leave the door open for the right reasons. Someone who has lost their voice and wants it restored. People who can't see and need the words read aloud. A family that wants to recreate a late loved one's voice, with permission. Those are real and they matter.

But every one of those is a choice the customer makes on purpose, with full understanding. Never the silent default. The default has to protect the one thing the product exists to deliver.

A Simple Test for When to Skip AI

You can turn this into a quick gut-check. Here's what I actually run through.

One. Is the thing you're automating the reason people buy? If yes, be careful. Automate the wrapping paper, not the gift inside.

Two. Does "almost perfect" feel creepy instead of impressive? For a loved one's voice, a financial advisor's judgment, a doctor's bedside manner, 95% isn't a win. It's a betrayal.

Three. Would the customer feel cheated if they found out it was AI? If discovering the AI behind the curtain would make someone feel tricked instead of delighted, don't cross that line.

Four. Is the money you save worth the value you destroy? Saving a fraction of a penny on a product that no longer does its one job is a terrible trade.

Now the other side, because I'm not an AI skeptic. Look at where AI clearly wins in my own work. Creating products. Setting prices across 564 items. Drafting content for 313 blog articles. Optimizing images. In every one of those, AI does the typing and the grunt work, and it's brilliant at it.

That's the line. AI does the typing. It doesn't get to be the soul of the product. Automate the labor. Keep the judgment and the human stuff human.

Why Saying No Is the Real Skill

If you run a business, you're being pitched AI for everything right now by people whose only move is "add more AI."

Distrust that reflex. Not because AI is bad, but because someone whose whole job is "put AI here" has no reason to ever tell you no. Their bill grows as the project grows. They're optimizing for their outcome, not yours.

The person who tells you where not to use AI is the one worth hiring.

I've shipped 29 automated tasks. I've cut manual work by 42%. I've pushed revenue per employee up 38%. I am not afraid of this stuff. I run it at full scale every single day.

And I still kill AI features on purpose when they'd make the product worse. The memory book is one example. There are plenty of others where the honest answer to "should we automate this" was a flat no.

That's the same judgment I bring to a client. Sometimes the right answer is no, and a good advisor says it out loud, early, before you've spent six figures finding out the hard way.

Anyone can say yes to more AI. Saying no requires actually understanding the product well enough to know what would break.

Want to explore what AI could do for your business?

Book a free 30-minute strategy call. No pitch deck, no sales team, just a real conversation about your operations and where AI fits.

Book a Discovery Call