Quality Control in High-Stakes Communications and AI Orchestration

Recently, my favorite baseball team misspelled the name of their newly-signed Japanese import. The star player’s incorrect name was hanging above his locker as he arrived at Spring Training. (Ironically, he showed up early to get some extra reps in before the start of the Cactus League.) Some supporters of human slop might say, “What’s the big deal? It’s not that deep.” Except, he posted a pic of the blunder on social media and rightly so. His two-year contract is worth $34 million; put some respect on the man’s name. Literally!

The best countermeasure is always to have human eyes review drafts. Yet, what do you do when team members are unavailable, you’re moving fast, and have a message to deliver to an audience whose attention is pulled in every direction? How do you stress-test your drafts? Not just for you. For your reader, customer, or partner.

If editors were enough, billion-dollar brands wouldn’t publish their fuck-ups in 2026. AI can write cheaply on its own and quickly, but only humans can make meaningful decisions about taste and take action with real-world consequences. Such results are felt at an emotional and social level, which is why if “two heads are better than one” we say, “three LLMs (at least) are better than one”.

Now, enter: Combat Writing. This methodology works with or without AI. (If privacy is a concern, you can run offline models locally.)

Combat Writing has four stages:

Strategy

• Humans write their piece entirely on their own, finding their own unique structure.

Sparring

• Engage in dialogue with at least 3 people or LLMs regarding content and organization.

Battle

• Go back and forth revising and editing while challenging concepts, wording and strategy.

Champion

• Publish and follow through with action tied directly to the purpose of the piece, track effects.

Synthesizing outputs from different LLMs, we engage in adversarial review, asking: Where does the AI crew agree and disagree on our communications? If conducting a team exercise ask: How do we embrace rather than reject the biases each member contributes to the discussion of this project? Since each model has their own unique training data and humans have intuition, this creates conditions for insights to emerge.

I’ll leave you with our motto that comes up when discussing Combat Writing in person. Why “combat”? The answer is always the same. “Reading is peace. Writing is war.”

[Build quality control infrastructure into your organization’s writing process. Join The Writing Production pilot: https://www.learningproducers.com/services ]