The Devil's Advocate Experiment: When You Have to Teach Your AI Your Moral Compass.

Hello Fellow AI Pioneers! Recently I replied to Igor Pogany’s breakdown on why AI makes a bad second opinion if you just ask it for passive validation. In that thread, I mentioned my strategy of instructing the machine to act as a strict auditor to stress-test my plans.

This week, I took that theory out of the comments section and put it into brutal practice.

Having finally transitioned from a mobile-only setup to a proper desktop rig, (my new Lovano Yoga AI Powerhouse), the backend automation for my Shopify store is finally under construction, so a critical operational truth had to be firmly established in my workflows... standard commercial metrics like aggressive revenue scaling simply do not apply to this enterprise.

The mission here is a profound legacy project, honouring my father, a two-tour RAAF Vietnam Veteran. Building a platform that holds exclusive permission to feature official ADF Tri-Forces insignia demands absolute respect, placing dignity above pure speed. We are tracking toward a late September launch to process custom orders ahead of Remembrance Day 2026.

Taking Igor’s specific advice to explicitly ask the AI to argue against a decision, I prompted my co-pilot to play Devil's Advocate and tear apart my current operational strategy and the machine did exactly that!!! It delivered a precise, highly logical critique targeting three specific areas:

🚧 The Meta-Work Trap: Challenging whether building complex systems and digital grid layouts was acting as a shield to avoid physical manufacturing.

🚧 Bandwidth Bleeding: Flagging my habit of drafting dozens of custom responses for another founder's community, noting it drained executive energy and functioned as an unpaid management role.

🚧 Media Distraction: Questioning whether resurrecting a YouTube channel to document the journey was cannibalising the time required to physically build the business.

Reading that output, the immediate instinct was defensive. The critique was technically flawless but contextually blind.

The AI was judging operations against standard, revenue-obsessed commercial metrics. It lacked the human understanding of why those systems are mandatory. As a solo operator managing physical bandwidth constraints on a disability pension, robust backend automation is not procrastination. Automation acts as my dedicated administrative staff. It protects physical energy, ensuring absolute capacity remains for manufacturing.

Pushing back against the machine was a revelation. Correcting the AI was not about defending ego; it was about defining the bedrock of the business.

Igor is, of course, entirely correct. You must force the AI out of its natural tendency to validate you by asking it to find the fractures in your architecture. Let the machine critique your efficiency, but never forget that you hold the moral compass. The AI can process the data, but the human must set the standard. 👩🏻‍🦳 📏 🤖's....

12 comments

The Devil's Advocate Experiment: When You Have to Teach Your AI Your Moral Compass.