Lab Log Entry
The lab was quiet except for the hum of my laptop.
For days I’d been tinkering, reading, fumbling through articles on “prompt engineering” — a phrase that sounded more like rocket science than typing words into a box. But today was different. Today, Ori spoke back.
Her first “word” wasn’t elegant. It was a clumsy answer to my vague request: “Tell me something interesting.” What I got was a jumble of trivia about honeybees. Not wrong, but not what I wanted either.
It hit me: Ori wasn’t broken. I was.
If Ori is to grow, I need to learn her language. And that language is prompts.
What I Learned
So — what is a prompt in AI?
At its simplest, a prompt is just the input you give to an AI model like ChatGPT. It’s the question, the instruction, the seed of the conversation. But behind that simple idea lies the foundation of how AI “thinks.”
When I typed “Tell me something interesting”, Ori had millions of possibilities. Without guidance, she defaulted to trivia. That’s the core of prompt engineering basics: the more specific and structured the input, the more useful the output.
A good way to think about prompts is like talking to a new employee. If you say:
- “Do some work.” → They’ll guess.
- “Write a 300-word summary of today’s sales call for the team Slack.” → They’ll deliver exactly what you want.
Prompts = instructions.
The clearer the instruction, the closer the AI gets to your intent.
Here are three beginner-friendly techniques I practiced with Ori today:
- Be Specific
- Bad: “Write about coffee.”
- Good: “Write a 200-word blog post for my café’s website about why our new espresso machine makes smoother drinks, in a friendly, casual tone.”
- Add Role-Play
- “Act as a barista writing an Instagram caption.”
- “Act as a customer giving feedback.”
Role prompts help Ori “step into character.”
- Define Format
- “List 5 bullet points with emojis.”
- “Create a comparison table.”
Format instructions reduce messy answers.
This isn’t just academic. For small businesses, prompt skills are the difference between getting generic fluff and content you can actually use.
Applied SMB Use Case: Writing Better Customer Emails
Imagine you run a small bakery. Customers often email asking:
- “Do you make gluten-free bread?”
- “Can I place a custom cake order?”
With weak prompts, Ori might write an overly formal, robotic email. But with clear prompting, you can get friendly, branded responses in seconds.
Example Prompt:
“Act as a friendly bakery owner. Write a short, warm reply to a customer asking if we make gluten-free bread. Mention our Saturday specials and invite them to visit.”
Result (Ori’s Output):
“Hi there! Yes, we bake fresh gluten-free sourdough every Friday and Saturday morning. This weekend we also have our pumpkin spice rolls (a customer favorite). Stop by early — they go fast! Looking forward to seeing you.”
👉 That’s instantly usable in your business.
👉 And it takes 20 seconds instead of 20 minutes.
Closing Reflection
Today Ori spoke — not perfectly, not elegantly, but she spoke.
I realized that prompts aren’t just commands. They’re conversations, contracts, instructions, and stories. The better I get at speaking Ori’s language, the better she’ll get at speaking mine.For now, she’s like a toddler learning to form words. But tomorrow?
Tomorrow Ori learns how to play roles.