For once-in-career decisions, suspend your agent and deliberate fully — automation is for recurring patterns, not strategic pivots

Why This Is a Rule

Decision agents (Only automate decisions that are frequent, stable, and low-stakes — all three must hold or automation introduces risk) are designed for recurring, stable, low-stakes decisions. They work by applying a pattern learned from past instances to the current instance. But strategic pivots — career changes, major investments, relationship-defining conversations, business model shifts — are by definition not instances of a recurring pattern. They're novel situations where the context differs fundamentally from anything the agent was calibrated against.

Applying an agent to a novel strategic decision produces what looks like decisiveness but is actually pattern-matching against the wrong reference class. Your "how to evaluate job offers" agent was calibrated against incremental career moves. Applying it to a career-defining pivot (academia to industry, employee to founder, country relocation) will produce an answer, but the answer treats a discontinuous decision as if it were continuous — weighting salary, commute, and title when the actual decision variables are identity, purpose, and risk tolerance.

The danger is that the agent feels helpful in the moment. It reduces the discomfort of uncertainty by producing a clear recommendation. But the recommendation is based on the wrong model — an incremental-move model applied to a discontinuous pivot.

When This Fires

• When facing a decision that will fundamentally redirect your career, relationships, or life trajectory
• When the stakes are so high that getting it wrong isn't recoverable through iteration
• When the decision context doesn't match the recurring patterns your agent was trained on
• When you notice yourself applying a decision heuristic to a situation that feels qualitatively different from its usual context

Common Failure Mode

Using agents for strategic decisions because deliberation is uncomfortable: "My framework says to evaluate based on X, Y, Z — and the answer is clearly A." The speed and confidence feel good, but they're artifacts of applying a pattern inappropriately. The discomfort of full deliberation for a strategic pivot isn't a bug — it's the signal that the situation demands more processing than a heuristic can provide.

The Protocol

(1) When facing a major decision, ask: "Is this a routine recurrence of a pattern I've seen before, or a qualitatively new situation?" (2) If routine → use your agent. That's what it's for. (3) If novel/strategic → explicitly suspend the agent. Write: "This decision exceeds my agent's design parameters." (4) Return to full deliberation: gather new information, consult diverse perspectives, consider second-order consequences, sit with uncertainty. (5) After the decision, debrief: what factors mattered that the agent wouldn't have considered? Use this to refine the agent's override conditions for future novel situations.