The Five Failure Modes

Module 5 · Section 3 of 4

The Five Failure Modes

Now let’s talk about how this system breaks — not because the system is flawed, but because of how people use it. These are the patterns that cause people to abandon the system or publish generic work despite having a style guide.

Failure Mode 1: Lazy Prompting Produces Lazy Output

This is what it looks like: You give AI a vague prompt, accept the first output, apply minimal editing, publish work that sounds like everyone else despite having a style guide.

Real example: “Claude, write a newsletter about AI voice training.”

What this produces: Generic structure you’ve seen a hundred times, obvious points anyone could make, no distinctive perspective, content that could run on any AI-focused publication without changing a word.

This is what works better: “Here’s my outline. Here’s my main argument: most people use AI to sound more professional, which actually destroys their voice. Here’s my personal example of catching myself doing this. Now help me structure this clearly while maintaining the patterns in my style guide. Key points to cover: [specific list]. Controversial stance: AI should make you more efficiently yourself, not more efficiently generic.”

The difference: The second prompt gives AI your thinking to execute, not a blank canvas to fill with generic templates.

The fix process: (1) Always start with your thinking (15 minutes minimum before touching AI), (2) Give AI specific instructions rooted in your argument, not open-ended requests, (3) Reference your style guide explicitly in every prompt, (4) Provide examples of what you want from your own past work, (5) Be prepared to reject and redirect if first output misses the mark.

Remember: The style guide helps, but vague prompts override good guidelines every time.

Failure Mode 2: Skipping the Read-Aloud Test

This is what it looks like: Text looks fine on screen but sounds robotic when spoken. You publish it anyway because it “seems good enough.”

Real example from my editing:

AI output (looked fine on screen): “It is important to note that the utilization of artificial intelligence in content creation necessitates careful consideration of authenticity metrics and voice preservation methodologies.”

After reading aloud (rewritten): “Using AI for writing? You still need to sound like yourself. That’s not negotiable.”

Why this matters: Your voice lives in how things sound when spoken. Visual editing catches technical errors. Audio editing catches voice problems. If you skip the read-aloud test, you miss 50% of voice issues.

The fix: Read EVERYTHING aloud before publishing. Not in your head — actually say the words. Record yourself reading it if that helps. Listen for places you stumble. Notice where you lose energy. Notice sentences that require two breaths. Rewrite until it sounds natural.

No exceptions to this rule. The 5 minutes it takes to read aloud will catch problems that would take 30 minutes to diagnose through visual editing alone.

Failure Mode 3: Treating Style Guide as Static

This is what it looks like: Your voice evolves. Your style guide doesn’t. AI output starts feeling increasingly “off” but you can’t figure out why. You start rewriting 60-70% of content instead of the expected 30-40%.

Signs this is happening: AI output consistently feels wrong even when you reference the guide, you’re rewriting more than half of what AI produces, your natural writing doesn’t match the guide anymore, you’ve developed new patterns or phrases AI doesn’t know about, you’re avoiding certain topics because the guide doesn’t handle them well.

Why this happens: Voice isn’t static. The way you wrote two years ago isn’t the way you write now. Your characteristic phrases evolve. Your sentence rhythms mature. Your tolerance for formality changes. If your guide doesn’t evolve with you, it becomes a constraint instead of a tool.

The fix schedule: Weekly quick review (5 minutes) — scan your latest piece and note anything that needed heavy editing; Monthly minor tweaks (15 minutes) — update forbidden list with new AI-isms you caught, add new characteristic phrases that have emerged; Quarterly major revision (1 hour) — gather new writing samples, run new voice analysis, compare to existing guide, update to reflect current patterns; Yearly complete overhaul (2-3 hours) — rebuild guide from scratch with current writing samples.

Treat the guide as living documentation. It should grow more accurate over time, not more stale.

Failure Mode 4: Over-Relying on AI’s Judgment

This is what it looks like: “AI suggested this phrasing so it must be better.” You edit toward what sounds “good” instead of what sounds like you.

Real example:

My original: “This is hard and sometimes you’ll want to quit.”

AI suggestion: “While this presents challenges, maintaining persistence will yield positive outcomes.”

What I kept: The original. Because that’s how I talk. The AI suggestion is more formal, more polished, more “professional” — and completely wrong for my voice.

The trap: AI optimizes for what sounds good in general. You need to optimize for what sounds like YOU specifically. Those are different targets.

When AI suggests alternatives, it’s trying to improve clarity, polish, formality — all legitimate goals, but not YOUR goals. Your goal is authentic voice, even if that voice is less polished than AI’s suggestions.

Decision framework for evaluating AI suggestions:

1. Does this sound like me? (Most important)
2. Would I say this in conversation? (Voice check)
3. Does this make my point clearer? (Clarity check)
4. Am I editing toward “better” or toward “me”? (Direction check)

If the answer to #4 is “better,” you’re editing wrong. You’re letting AI sand down your rough edges, hedge your strong opinions, formalize your casual tone, remove your personality quirks.

Those rough edges ARE your voice. Protect them.

Failure Mode 5: Forgetting Why You’re Doing This

This is what it looks like: You start using AI to sound more formal, more impressive, more professional — more like everyone else. Your rough edges disappear. Your personality gets sanded down. You become indistinguishable despite having a style guide.

What drives this: The belief that “professional” means formal, polished, corporate-sounding. The fear that casual language makes you seem less expert. The temptation to hedge opinions to avoid controversy. The instinct to add jargon to sound more authoritative. The desire to make everything perfectly polished.

The reality: Those rough edges ARE your voice. Those quirks make you recognizable. That personality is your competitive advantage in a world flooding with generic AI content.

What to preserve (even if it feels “unprofessional”): ADHD tangents if that’s how you think, occasional swearing if that’s how you talk, honest admissions of difficulty instead of fake confidence, grammar rules broken intentionally for rhythm, conversational asides that build connection, rough edges that make it distinctively yours.

The test: If your friends read this without seeing your name, would they recognize it as yours?

If no, keep editing. If yes, you’re done — even if it’s not “perfect.”