> ## Content Index
> Fetch the complete content index at: https://techessentials.in/llms.txt
> Use this file to discover other available public pages before exploring further.

# AI Surprised the Experts. Still Think Only You Can Decide?
- URL: https://techessentials.in/ai-surprised-the-experts-still-think-only-you-can-decide/
- Published: 2026-11-05T13:44:02.000Z
- Updated: 2026-11-05T13:44:02.000Z
- Author: Sanidhay Kumar
- Tags: AI, Strategy

Some very smart people have spent their careers working on Linux.

You may never have used the name, but you use what it powers. Android phones. Cloud servers. Much of the machinery behind the internet.

The project [began in 1991](https://www.kernel.org/doc/html/v6.3/process/1.Intro.html?ref=techessentials.in). Generations of engineers have built it, reviewed it and improved it. Security specialists study how someone might break it. The code is open for people to inspect.

Then AI found something that surprised the experts.

On 4 November, OpenAI [published research on an AI system that investigated security weaknesses in Linux](https://openai.com/index/a-thousand-fewer-paths-to-root/?ref=techessentials.in). It found weaknesses and built working demonstrations of how they could be used to gain control of a computer from an ordinary user account.

One detail stayed with me.

OpenAI examined a sample of the results and identified ten exploitation techniques it believed were new. It showed those techniques to outside experts who specialise in breaking into Linux systems. The techniques were unfamiliar to them too.

These people know their field deeply. AI still contributed something they had not seen.

Now think about the purchase order waiting for your signature.

## Why every decision reaches you

Your purchase manager has compared the suppliers. The rates are in the sheet. The delivery dates are confirmed. Still, the order comes to you.

Your salesperson has spoken to the customer. The customer wants a discount. The salesperson messages you before replying.

A new project comes in. Your team gathers the requirements, then waits for you to decide the price.

You have reasons for all three.

You know that the cheapest supplier becomes expensive when delivery slips. You know which customer asks for a discount on every order and which one will actually walk away. You know when a small first project could lead to a much bigger account.

You built this judgement over years. You paid for some of it through mistakes. When someone says, "Just delegate," you remember the last time you did.

So you keep deciding. Even on holiday. Even at dinner. Even when you have hired people specifically to take work off your plate.

The work moves until it needs your judgement. Then it waits.

## Who else could make the decision

"Nobody understands this business the way I do."

That may be true.

But you use it to reach a second conclusion: "Nobody else can make these decisions properly."

How recently have you tested that?

You may have tested it with a junior employee three years ago. They missed something important. You took the decision back. Since then, every quotation has come through you.

Now you have another candidate to test.

The Linux research makes me take that possibility seriously. Deep expertise still left room for AI to find something unfamiliar. It makes me wonder how much room there is in the decisions an owner makes every day.

Perhaps AI will miss a detail you consider obvious. Perhaps it will ask a useful question before recommending a price. Perhaps it will spot a cost you have become used to absorbing.

You will learn which by giving it the information and examining its work.

## Let it examine a quotation

Take a quotation you prepared recently. One that required some thought.

Give AI the information you had before deciding the price: the customer's enquiry, the scope of the project, your costs, the delivery commitment and the customer's payment and order history. Explain [anything important that lives only in your head](Your Best Employee's Knowledge Lives in Their Head. That Is Your Biggest Risk.). If winning this job matters because it could open a new market, say so.

Keep your actual quotation aside for the first attempt. Ask:

> Recommend a price and payment terms for this project. Explain what you considered and where the margin could come under pressure. Ask me for any missing information you need before recommending a number.

Then compare its recommendation with yours.

Did it understand why you would offer this customer a better rate? Did it account for the work involved? Did it ask about something you had forgotten to check?

Suppose you priced a project assuming two site visits. AI asks how many visits similar projects actually needed. You check the last few jobs. Most needed four.

That would be a useful contribution to your pricing, even if you disagreed with its final number.

Try several different quotations. Include the awkward ones. Each time, examine where its reasoning holds up, where it needs more context and whether it adds anything to yours.

OpenAI used a specialised research system. Whether the AI you use can handle your quotations is something your own cases will tell you.

## Your company keeps waiting

If it starts making sound recommendations consistently, use that evidence. Let your team prepare quotations with it. Decide which cases they can handle and which still need you.

Your experience helps you recognise a sensible recommendation, explain a missed detail and identify an exception. Over time, more of that experience can become available to the people doing the work.

That is how an afternoon without quotation approvals becomes possible.

The experts who examined those unfamiliar Linux techniques had spent years earning their expertise. There was still something new for them to see.

You have spent years earning yours too.

Before you sign the next purchase order, ask yourself when "only I can decide this" last had to prove itself.