How to Use AI to Find Flaws in Business Documents

I intended to save time by writing product documents with AI.
I ended up losing out instead.

At first, I did not realize what was happening. I thought AI could replace the hard work of thinking and writing drafts from scratch.
I was actually sacrificing quality for fake speed.

Why did this happen?

Imagine this scenario. You ask a total stranger who does not understand your company context to create your core strategy.
On paper, the sentences look neat. The vocabulary sounds highly intelligent.
But when you read it closely, the document is empty.
It has no soul. It does not solve any real user problems.

The real power of AI is not in writing documents from a blank page.
Its true power emerges when you use it as your harshest critic.
AI is exceptional at finding flaws in business documents you have already written with your own brain.

As a product manager, your job is not just to tidy up a feature list.
Your job is to ensure no user scenario is missed.
And the human brain always has blind spots.
This is especially true when you have stared at a screen and thought about one single problem for days on end.
Eventually, every piece of logic will make perfect sense in your own head.

Efficiency is an old habit

Before I get into the technical steps, I want to clarify one thing about efficiency.

Making companies lean is an old habit of mine. It started long before AI became a trend.

If you look at my profile page, there is a figure highlighting more than $4 million in savings per year.
That number is not some AI magic trick. It is the result of years of hard work before ChatGPT even existed.
I drove these savings across companies like Tokopedia, Hijra, and Flip.

Where did that money come from? It came from various cost saving initiatives across different departments.
It came from internal process improvements, operational cost optimization, cutting redundant server traffic, and making much more precise product decisions.
Customer ticket automation was just one of them, not the only source. We looked at everything from infrastructure billing to third party API usage to find where the company was leaking cash.

Back then, getting leverage at that scale required a specific setup. You needed a strategic position. You needed the backing of dozens of engineers. You needed expensive tracking systems.
Not everyone had the luxury to execute the efficiency ideas they found.
Because executing those ideas was very expensive.

But now, the situation has completely flipped.
AI is the sharpest tool available to continue that exact same habit.
Most importantly, that massive leverage can now be taught and used by everyone on your team.
Nobody has to wait in line to ask the engineering team for their time.

The practical way to find document flaws

Here is the real workflow I use to get actual value out of these tools.

Never ask AI to write your first draft.
That is the most common trap people fall into.

You must sit down, think, and write the product requirements yourself.
Use your own words. Provide the real context of the problem you are solving.
Only after your draft is done should you feed it into an AI model.

But do not use lazy prompts like asking it to fix your grammar.
That is a massive waste of AI potential.
The correct prompt positions the AI as your most critical debate opponent.

Give it a very specific role.
Try something like this: "You are the most meticulous red team and QA tester. Your job now is to find flaws in this business document. Find holes in the logic. Look for extreme edge cases I have not considered. Attack all the weak assumptions in my writing."

The results will make you pause.
The AI will tear your document apart without mercy.

It might find cases you completely missed: "What happens if the user closes the application exactly when the payment is being processed by the partner bank?"
Or it might say: "Your document assumes all phone numbers start with a local country code, but what about expatriates using foreign numbers to register?"

The AI will give you a long list of potential flaws in the field.
Your job as a human is to review and choose. You decide which flaws are fatal and absolutely must be closed before the project enters the development phase.

Maker bias and the need for a third party

Finding flaws using AI is highly effective for one simple reason.
AI has no maker bias.

As humans, we naturally feel pride in our own work.
We tend to protect the ideas we have built with great effort.
Because of this, we often find it incredibly hard to see the weaknesses in our own creations.

A machine does not care about your feelings. It has no emotion or empathy for your writing.
It just looks for inconsistent data patterns and broken logic.
And that is exactly what you need before a document is handed over to hundreds of people to execute.

Maker bias often causes long project delays.
How many times have you seen an engineering team stop right in the middle of development? They stop because they found a technical condition the document writer completely missed.

Suddenly, a project that was supposed to take two weeks stretches into a month. The engineers have to wait for the product manager to clarify the requirements. The product manager has to go back to the drawing board to figure out a solution.

Every time work stops because a business document is incomplete, the company bleeds money.
Technical time is very expensive. A developer sitting idle waiting for logic clarification is a massive waste of resources.

It is much better to feel slapped by AI at the draft stage than to watch your product break when used by real users.
Fixing logic in a document is free.
Fixing broken code in a production environment carries a very high price tag.

Putting AI in the right place

My core principle on using tools remains consistent.
Put the strongest tool in the place where mistakes are the most expensive.

The biggest mistake a product maker can make is not a typo. The biggest mistake is designing flawed system logic.

If you get used to critically analyzing your own documents before anyone else reads them, your overall quality will skyrocket.
People will see you as a sharp thinker who anticipates every possibility.
Therefore, you must use this technology sharply.

Stop wasting your AI quota just to make your sentences sound pretty.
Start using the machine to rigorously test your logical thinking.

Which do you prefer? Looking smart when presenting an initial draft?
Or launching a product that is truly rock solid when it finally hits the market?

For those who want to dive deeper into this kind of thinking framework, you can gather with like minded people at AI Circle. We dissect practical, sensible ways to use AI for daily work.
But if you are a decision maker and feel your entire team needs this training, we can discuss it through my corporate training program. Help your team become critical thinkers, not just automated typists.