AI’s First Draft Is Terrible. These 3 Plays Fix It.

AI is everywhere. It has flooded the zone. “Thought leaders” are riffing about it on LinkedIn. Your staff is using it to draft meeting notes. Your marketing team is feeding prompts to ChatGPT and declaring victory.

But here’s the rub: in my experience, most CEOs—smart, competent leaders who already use AI regularly—aren’t using it to its full potential. They’re satisfied with outputs that are “good enough” when they could be getting outputs that are sharper, faster, and much more useful.

If you’re treating AI like a very smart and multicompetent intern, you’re missing the boat. Done better, it can be your sharpest strategist, your harshest critic, and your most ruthless fact-checker.

Here are three underutilized plays that separate the AI dabblers from the CEOs who really know how to lead with it (and want to provide guidance for their team in how to use it).

Play #1: Stop Barking Orders. Have a Conversation.

Too many leaders treat AI like a vending machine. Insert prompt. Receive output. Complain when the output is bland. (Or, remain blithely unaware that the output is subpar or misguided.)

Here’s the problem: vending machines don’t ask clarifying questions. They just drop a bag of chips.

The better approach is to treat AI like a strategic conversation partner. Instead of typing “Draft a fundraising email,” say, “I need you to draft a fundraising email on X with tone Y. But before we begin, let’s talk. Please ask me any questions you need answered to help you give me the best outcome possible.” Suddenly, you’ve invited the AI to probe deeper. It will ask about audience, goals, tone, context—and your answer will be better because the questions were better.

Examples:

• Business CEO: Ask AI to outline a pricing strategy. If you just bark the order, you’ll get a generic grid of “good, better, best” tiers. If you converse, AI might ask about competitor moves, cost structure, and customer psychology—and give you a nuanced framework.

• Nonprofit CEO: Command AI to “write a donor appeal,” and you’ll get Hallmark prose. Instead, invite questions. It might ask about donor segmentation, urgency, and giving history—producing a letter that feels human and specific.

• College president: If you simply order “draft a speech for admitted students,” you’ll get a nauseating bundle of clichés about “bright futures.” But in conversation, AI will ask about your institution’s values, competitive differentiators, and the anxieties students are bringing. The final draft resonates instead of putting people to sleep.

Note: If your AI never asks you a question back, you’re not talking to a strategic partner, or even an assistant—you’re talking to a search engine in a tuxedo.

Play #2: Ask, “Why is This a Hot Plate of Garbage?”

A lot of otherwise-savvy leaders accept AI’s first draft. They take what it gives, maybe trim a few sentences, and move on. That’s like serving your guests bread dough instead of a baked loaf—technically food, but no one’s impressed.

The smarter move is to force AI into self-critique.

When AI hands you something uninspiring, type back: “Why is this garbage?” or, if you’re feeling diplomatic, “Critique this like a ruthless editor.” AI will shred its own output—identifying weak openings, boring transitions, and generic claims—and then rewrite based on its own feedback.

Even better, pit AI systems against each other. Give Grok a draft, then ask ChatGPT to eviscerate it. Or let Claude or Grok dissect ChatGPT’s polite nonsense. The ensuing cage match will deliver insights you wouldn’t have found otherwise.

Examples:

• Business CEO: AI drafts a shareholder letter full of jargon. You tell it: “Why is this garbage?” It points out the clichés, then rewrites in clear, confident language that investors actually understand.

• Nonprofit CEO: You ask AI for a press release, and it serves up buzzword soup. You challenge it. It strips out the fluff, replaces it with impact metrics, and adds a compelling donor quote.

• College president: AI writes a commencement speech that could have been delivered anywhere by anyone. You ask it why it’s terrible. It admits it’s generic, then produces a version laced with your school’s history, distinctive culture, and actual student stories.

Note: The first draft is the AI equivalent of hotel coffee—it’ll get you moving, but that’s the only positive thing about the experience.

Play #3: Make One AI Call Out Another’s Bullcrap.

AI is brilliant at sounding confident. It’s less brilliant at being right.

Ask one model for quotes, stats, or historical anecdotes, and you might get a masterclass in fabrication. Ask a second model to fact-check the first, and suddenly you have accountability.

It’s like hiring two consultants with massive egos and asking them to critique each other in front of you. You’ll quickly separate the gold from the nonsense.

Examples:

• Business CEO: Grok gives you a market analysis with numbers that look suspicious. You hand it to ChatGPT and say, “Fact-check this.” Within seconds, the inconsistencies are exposed, saving you from embarrassing your board.

• Nonprofit CEO: AI produces a draft article for the website and quotes several famous people to back up your points. You ask Claude to review it, and it flags fake quotes and incorrect citations. The revised version is credible and persuasive.

• College president: You ask AI to create a list of “famous alumni.” It invents two Nobel Prize winners and a U.S. senator. You send it to another AI, which politely says, “Um, those people never attended your school.” Catastrophe avoided.

Note: If you’ve ever caught one AI lying and forced another to prosecute the case, you’ll understand why I call this the “electronic perp walk.”

The Bottom Line

If you’re a CEO using AI and satisfied with the results, congratulations—you’re average. But if you want to be more than average, if you want to turn AI from a useful tool into a competitive advantage, you need to stop dabbling and start demanding excellence.

• Don’t bark orders—hold conversations.

• Don’t accept first drafts—demand brutal self-critiques.

• Don’t trust one model—make them keep each other honest.

The leaders who use AI this way aren’t just producing better outputs. They’re thinking more clearly, communicating more persuasively, and making faster, smarter decisions. And they're guiding their team to do the same.

In other words: they’re leading.