Last evening, I witnessed the CEO of a mid-sized enterprise in southern Jiangsu open an AI chat interface and inquire, "I'm considering laying off 30% of the workforce in the East China region next quarter, and I also plan to replace the stone lions at the company's headquarters entrance. Are there any feng shui considerations I should take into account?"

After posing her query, she flashed me a cryptic smile and remarked, "One can never be too cautious these days."

It's hard to imagine that this CEO, who appears so composed in interview programs and public events, repeatedly consults AI before making significant decisions—and asks some rather unconventional questions. (She has granted permission for her to be mentioned in this article.)

From my observations, numerous corporate leaders, both domestically and internationally, engage in similar practices.

Of course, such actions are never disclosed in reports or news, nor are they brought to the attention of the board of directors. But isn't that the nature of the world? Public statements, information, and appearances often diverge from private behaviors, cognitions, and practices.

In reality, it's understandable why decision-makers resort to this. Everyone harbors personal concerns and hidden anxieties. Before the advent of AI, figuring out how to factor these into decisions was a challenge, and they could only seek advice from experts or deliberate in private. Now, with AI, it's akin to having an additional confidential advisor who doesn't carry the risk of leaking secrets, unlike certain historical figures. Naturally, they lean heavily on AI and consult it on everything.

In the presence of AI, many executive decision-makers let their guard down for the first time.

However, I hold a counterintuitive perspective on this phenomenon: privately consulting AI poses greater risks than formally integrating AI into the company's decision-making process.

Why? Because when a company formally adopts AI, there are logs, approvals, and team reviews in place. When consulting AI privately, there's no such oversight—just an individual with secrets and a malleable chat interface.

From my understanding, cognitive psychology suggests that after making a risky decision, people instinctively seek external justifications to alleviate their anxiety. So, if you consult AI before making a decision and heed its advice, once the decision goes awry, your brain will unconsciously assign part of the blame to AI. This way, the sting of decision-making errors is dulled—shifting blame to AI feels more justifiable than shifting it to a person, and AI's perceived authority makes people think, "Even AI concurred..."

Consequently, decision-makers' ability to learn from their mistakes is subtly undermined.

What's worse, you might not even be aware of this happening.

At the same time, I wonder if you've noticed something: when querying AI, it's challenging to remain objective; biases often creep in. This creates a suggestibility—AI tends to prioritize finding arguments and explanations that align with the user's intent and then casually mentions a few risks.

If it's about mundane life matters, it's not a significant issue. But when it involves the company's market value, employees' livelihoods, or business operations, the repercussions could be magnified infinitely. Moreover, AI doesn't truly comprehend your company. Even if you provide extensive background information, it might not be aware that your CFO is having an affair with the company's seal administrator, that the East China region manager has lost faith in you, or that your major client is embroiled in internal conflict...

Of course, decision-makers might not be privy to these hidden dynamics either, nor can they necessarily grasp them accurately. What I'm trying to convey is that sometimes such nuances are pivotal for decision-makers to make informed choices, and how to piece them together and interpret them is best left to the human mind, not AI, because you have to take responsibility! Not to mention that inputting such information could lead to leaks.

AI's responses, grounded in public data and general logic, may appear sound on the surface but are actually just a facade lacking depth. Is it truly prudent to rely on this for making responsible decisions?

Furthermore, based on my experience, when executives consult AI privately, it's often during times of intense pressure. Late at night, alone with the moon and a glass of wine, facing the AI chat interface, anxiety can spiral out of control.

Under this anxiety, AI's allure lies in its rapid responses. You pose a question, and it can churn out ten pros and cons analyses and five solution sets within seconds... The human brain can't process this deluge of information under such duress. So, you become overwhelmed with information, start asking follow-up questions, force yourself to think after querying, and the more you ask, the more anxious you become, and the more anxious you are, the more you want to ask...

So, is there a remedy? Actually, it boils down to the fundamental question of "how to use AI scientifically."

For instance, after asking, "I want to do A, what do you think?" immediately follow up with another question: "If my decision is wrong, what are the most likely reasons? Please enumerate the three most probable reasons and provide specific indicators for when each reason manifests; also, what are the most dire consequences?"

Then, don't immediately glance at AI's response. First, ponder your own answer, jot it down, and read it aloud in front of AI.

The act of writing compels your brain to slow down and deliberate. After thinking for yourself, then examine AI's response, which might be more reasonable.

At the same time, be sure to consult multiple AIs, not just one product. The principle of gaining insight from diverse perspectives remains timeless.

In the AI era, a decision-maker's true acumen lies in fully considering AI's advice but daring to chart a different course and being willing to shoulder the consequences of that choice.