Most of my earlier AI builds were low-risk: a company website, a personal blog, small internal tools. If something breaks, it is a 10-minute fix. If it slips by three days, nothing meaningful collapses.
An ERP is a different animal.
In a manpower business, failures do not show up as glitches. They show up as penalties, customer complaints, or legal disputes — and they break trust with business stakeholders. This is not a surface-area problem you can improvise through.
That risk forced a hard decision: the project brief could not be a planning ritual. It had to function as a defensive weapon for the business.
Standing on Shoulders, Not Reinventing Judgment
I already knew what I wanted to build. That was not the problem. The problem was interrogation.
I needed a system that would challenge my assumptions harder than I could on my own, surfacing edge cases I could not see. AI is good at pattern recall, but without boundaries it also invents confidently.
So I did not brainstorm. I adopted the BMAD process — not for convenience, but for pressure-tested questioning. If a way of turning vague intent into precise scope has already been thought through, rebuilding that thinking from scratch is wasteful. Using what has already been built and tested is not a shortcut. It is discipline.
What the Project Brief Is Actually For
The brief was not written to describe features. It existed to do three things.
First: define non-negotiables — what must be true for the system to be valid. Second: kill ideas early — what is not needed now, even if it sounds good. Third: lock phase boundaries — hard-coding the line between survival-critical and nice-to-have.
I forced the AI to ask relentlessly: what breaks if this is wrong? Who gets affected? What happens when someone violates the rule on purpose?
The brief became the place where ambition is restrained to real-world reality before it leaks into the code.
Phase Discipline as a System, Not Willpower
My biggest risk was not ignorance. It was letting ideas that felt small sneak into Phase 1.
Once something was classified as nice-to-have, it stopped being debatable. Focus was not maintained through motivation. It was enforced by the system.
The project brief does not help me imagine the system. It protects the foundation from my own ambition.
Develop Your Survival List
Audit your current project or operational cadence. Build your Phase 1 survival list.
If a feature or process is not required to solve your immediate problem or fix a critical issue, defer it. Stop debating it.