Edition 115.

I interrupt our regular news-byte format to share 5 things “the top 0.1%” of users do differently with AI.

But first, a little context.

I just power-walked back to my hotel 🏃💨

A conference I’m attending wrapped less than an hour ago.

These notes are freshly scribbled (and my heart rate is still up).

Yesterday, AI strategist Alicia Lyttle (known as “The Queen of AI” 👑) demo’d how she uses ChatGPT, Claude and her personal AI agent, Maximus, to:

• Build presentations
• Automate email responses
• Stress-test sales materials
• Map her entire team org chart, and
• Sharpen her AI prompts across multiple LLMs

She also said she hasn’t personally sent an email in over *100* days (if I remember correctly) 🤯

You *had* to be in the room for the full picture.

But these 5 takeaways were too useful to hold.

👑 5 reasons why Alicia’s AI might be smarter than yours:

1. _____-mode activated

Before you ask for the output, tell the model *who* it needs to be.

Examples? Professor mode. Sales mode. Skeptical client mode.

The mode can change the questions it asks, the objections it catches and the copy it generates.

Set the mode.

2. Put a price in the prompt

(This is the tip I want to test most.)

“Write a follow-up email” is a little womp-womp.

Try:

“Write the follow-up email a $10,000-an-hour sales consultant would send.”

Define the value and caliber of thinking you desire.

3. Give it references

AI takes direction better with references. So, show it what you think “good” is.

Earlier this year, I art directed a major brand campaign using references for lighting, texture, posing and wardrobe before generating the models.

The same principle applies to words, decks and strategy.

Inspiration is fair. Theft and imitation are not.

Cite the source when relevant. Use references responsibly.

4. Turn the intelligence up

Unless you want a highly agreeable, everything-you-say-is-amazing dialogue with your AI, turn up the Intelligence.

If your tool offers intelligence or reasoning settings, choose the highest level available.

I use High in ChatGPT.

More thinking power. Less flip-flopping, yes-man.

5. Lock the face

Alicia’s phrase: “face lock.”

When editing an AI-generated person, or your own uploaded image, instruct the model to “face lock.”

In other words: preserve the face, identity and defining features.

That way, you can change everything else without you or your subject morphing into a whole new person.

(Or your surprise cousin.)

I haven’t tested this one yet. But it would’ve come in handy for a recent client campaign.

I’m testing these AI prompts *this* week.

Hit reply if you do, too!

And tell me what changed.

More brand bytes next Sunday at 5!