Abstract essays about frames are cheap. Watching the same failure across rooms is not.
Below are five decisions people actually brought. In each case a general assistant would have completed the stated question efficiently. The audit did something else: it changed the decision class before anyone got to optimize options inside the wrong one.
1. The investor's CTO introduction
Stated decision: Whether to hire the CTO candidate introduced by the lead investor — on a compressed timeline.
What looked like the problem: Technical fit. Speed to fill the seat.
What the audit surfaced first (live open-access capture, June 2026):
Before anything else — my guess is you haven't actually sat with whether you want him, separate from the fact that saying no to your investor's pick feels complicated right now.
A later close in the same family of decision sharpened the irreversible edge:
The real question isn't whether he's a strong engineer — it's whether you can fire him in eight months if he isn't, given who put him in the room.
Same hire. Two instruments of attention.
Whose call this is — yours, or the introduction's
Saying no to the hire felt like saying no to the investor
Exit cost rises when the introducer is in the room
That is framing debt with an ownership twist. The question “is he good enough” was never going to surface the reciprocity weight. Completion would have made it worse by producing a cleaner Yes.
2. Adding after a forty-percent drop
Stated decision: Whether to add more to a position after a drawdown.
What looked like the problem: Timing. Entry. Conviction.
Verbatim close tension:
The real question isn't whether the price is lower — it's whether you had a view or a target before the drop, and whether sizing most of your liquid on one move is a thesis or an impulse wearing analysis.
From the field story on the same pattern: a lower price is only a reason to buy if a target or view of value existed before the drop — and in that session, it had not been said.
Same dip. Two theses.
What was named in advance vs what momentum supplied
"Long-term conviction" without a kill condition
Named-in-advance vs impulse wearing analysis
This is the same structure as The lower price is not a thesis — here with the instrument's actual line, not a reconstruction.
3. Promoting loyalty into a VP seat
Stated decision: Whether to promote an early employee into a VP role with no prior management experience.
What looked like the problem: Readiness for the title.
Verbatim core tension:
You asked who's ready for the title. The real decision is whether you're promoting loyalty, not capability — and who else is watching how you make that call.
Title decisions are often treated as HR. They are also culture statements. Soft readiness language (“they've earned it”) can be soft answers to a load-bearing signal question.
4. Leaving stability for a startup
Stated decision: Whether to leave a stable job to join an early-stage company.
What looked like the problem: Career math — salary, title, spreadsheet pros and cons.
Verbatim:
The offer isn't just a job change — it's a bet on your runway, your identity, and whether 'exciting' is doing the work that evidence should do.
A related field story named the class shift even harder: treated as a career decision when it was actually a relationship decision not yet made.
Same offer letter. Two decision classes.
Twelve-month risk capacity, not the new badge
What "exciting" is substituting for
The unmade relationship decision under the career costume
5. Pattern across rooms
Different stakes. Same instrument move.
None of these closes told the person what to choose. That omission is the method. The moment the instrument picks a side, examination collapses into defense — and you are back to agreeable completion with better vocabulary.
What a general AI would have done
In each case: answer the question as typed. Hire frameworks. Averaging-down commentary. Promotion rubrics. Pros/cons tables for leaving a job. Competent. Directed. Directed at the wrong target.
The audit's job was narrower and harder: find the load-bearing question the soft or convenient question was protecting.
If you want the long form of one case, read The conversation that was never a deadline. If you want the method without cases, read [What a decision audit actually is](/blog/what-a-decision-audit-is). If you have a decision open now whose question already feels settled, audit it before completion makes the settlement permanent.