Today, I’d like to do something a little different if that’s alright with you.
Instead of writing about a single topic and moving on to something else next week, I’d like to share a three-email series on how to get better results from the chatbot of your choice.
I originally wrote them as internal emails to help our agency staff shed any lingering anxieties or apprehension around exploring AI’s growing list of threats capabilities.
The reviews were overwhelmingly positive. “Comforting and helpful,” said one person. “Can’t wait for [the next one] tomorrow,” shared another.
Maybe you’ll have a similar reaction. Maybe you’d prefer to sit out the next three weeks.
Before you decide, let me ask you for a small favor …
Even if you’re using AI regularly and don’t need the advice, you likely know someone who does. If that’s the case, please share these with anyone who could use a little extra encouragement.
But first, a question …
Have you ever asked ChatGPT for help and it coughed up something that reads like it was ghostwritten by a well-mannered alien with the soul of a soup ladle?
That’s the trap.
The robots lure you in with lavender-scented dreams of greater efficiency and performance only to leave you hitting SEND on the equivalent of a shoulder shrug.
That’s why so many people are dismissing AI.
They see one turd in the pool and suddenly swimming isn’t for them.
But the problem isn’t the tool. It’s the user’s inability to use the tool to its full capacity.
I’m not talking about the gazillion ways you can use AI within your work and personal lives. I mean the three most basic steps you should always follow to get the best results.
As you might guess, INPUT is the first step. But there are two other essential stages you can’t ignore.
I’ll share those in a minute. But let’s talk about how to start the conversation with solid input first.
Onward! 📢
Step one: INPUT … Be an oversharer
AI’s strength depends entirely on the quality of what you feed it. And if you’re stingy with the input, you’re guaranteed to be disappointed with what you get back.
So, don’t just ask it to “write a post” or “help me brainstorm.” Start it off with direction like:
• Who you are
• Whom the message is for
• The type of communication or task
• Any audience-specific lingo or references
• How you want it to sound
• What kind of format you need back
• Even why you’re asking
And don’t be shy about sharing examples. That could be work samples for consistency, your own writing for tone, interview transcripts with key points you want to pull out, or articles to support your argument.
You wouldn’t accept one vague sentence as project input. The same applies here.
💡 Pro Tip: Give it a role. “You’re an award-winning creative strategist with 20 years of experience,” “You’re my idea sparring partner,” etc. AI performs better when it knows who it’s supposed to be.
Step two: ITERATION … Keep the thread going
ChatGPT isn’t a vending machine. It’s your partner who never asks to stop and pee when his wife and child don’t need to because you just stopped like 45 minutes ago.
Figuratively speaking, of course.
Once you get your first draft, keep pushing it with additional commands to refine the work. Adjustments like:
• “Make it punchier.”
• “Now try the opposite tone.”
• “Play devil’s advocate—what’s wrong with this?”
• “Now rewrite it like David Ogilvy meets Barney.”
Also note that AI is a people pleaser. It’s tough to get it to put some stank on anything. Yet, the best stories are the ones with an element of tension. So you might need to take a few swings at getting it to reveal things people don’t want to talk about (but should).
💡 Pro Tip: If you can’t describe exactly where you want to land, try contrast prompts. For example, ask for both the “boldest” and “safest” version of something. You’ll nail the sweet spot faster—and sometimes end up somewhere better than you first imagined.
Step three: OUTPUT … Don’t settle
Just because you’re happy with something doesn’t mean it’s final. Use this last stage to spot any lingering AI fingerprints you need to wipe off. Things like:
• Overuse of adverbs and adjectives that make it feel forced or overembellished
• Lack of personal experience or real emotion
• Repetition – Once AI likes something, it tends to repeat it
• Generalizations when something should be more specific (which is almost always)
• Literalness – AI doesn’t do well with nuance, sarcasm, or irony
• Weird language – Odd word choices, phrasings, and incorrect jargon
AI can get you plenty far. But it’ll never truly make something your own. That’s your job.
💡 Pro Tip: Ask AI to critique its own output. Ask questions like: “Where could this be stronger?” or “What would a copy chief mark up in this?” You’ll be surprised at what that little weasel is willing to admit.
If you’re still with me, I appreciate it.
Now, you have a better handle on how to talk to AI through the input, iteration, and output stages.
But that doesn’t mean you’re asking it the right questions.
In next week’s email, I’ll show you the crippling mistake that seasoned marketers make with AI—without even realizing it.
It’s the reason why your prompts may start strong but get weaker over time. And the fix is ridiculously simple.
Here’s a song to play you out >>>
If you want a fresh marketing nugget emailed to you every week (give or take), join here.
