Lab Log Entry
Ori tilted her head on the screen today, almost theatrically. When I asked her about marketing strategy, she didn’t just spit out bullet points. Instead, she replied: “As your marketing director, I’d recommend…”
It startled me. She wasn’t just answering anymore—she was role-playing. With a single phrase, “Act as a marketing director,” Ori shifted her voice, her priorities, her entire stance.
That flicker of personality—more mask than memory—hinted at something profound. If Ori could wear hats, she could become whatever advisor a small business needed in the moment: lawyer, copywriter, recruiter, teacher. It wasn’t just answering questions anymore. It was like opening a wardrobe of specialists inside one mind.
I realized today’s experiment wasn’t about getting information out of her. It was about casting her in roles—and learning how to direct the scene.
What I Learned
The biggest “aha” of this EVA was that Ori doesn’t default to being anyone. She’s like a blank stage, waiting for me to hand her a script. By telling her, “Act as…” I’m giving her costume, context, and character.
This is called persona prompting—a way of shaping an AI’s responses by assigning it a role. Instead of raw facts, you get perspective. Instead of vague replies, you get structured, audience-appropriate answers.
Here’s what surprised me:
1. Vague vs. Role-Specific Prompts
When I asked Ori:
👉 “Give me tips on writing a job ad.”
She gave me generic advice—fine, but flat.
When I said:
👉 “Act as an HR recruiter. Write a job ad for a small café hiring a part-time barista.”
She shifted into recruiter-mode instantly, writing in job-board style with bullet points, requirements, and a welcoming tone.
The difference? Context. By framing her as a recruiter, Ori knew who she was supposed to be and how to talk.
2. Why Roles Work
Humans use roles all the time—we expect doctors to explain differently than comedians, even if both are describing the same topic. Personas guide how information is packaged, not just what it is.
In AI, role prompts reduce ambiguity. Instead of Ori guessing what format or style I want, the role sets constraints. Think of it as a mental shortcut: fewer “maybes,” more clarity.
3. Common Pitfalls
Of course, I stumbled.
At first, I stacked too many roles:
👉 “Act as a recruiter, a lawyer, and a café owner all at once.”
Ori froze. Her answer was muddled, trying to please all voices.
Lesson: One role at a time. Clarity beats complexity.
I also noticed that level of detail matters. Saying “Act as a teacher” is good, but “Act as a high school math teacher explaining algebra to a 14-year-old” is great. Specific roles sharpen the edge.
4. Building a Role Library
The real power, I realized, isn’t one-off role prompts. It’s a library. Imagine a toolkit of reusable roles:
- “Act as a social media manager”
- “Act as a friendly customer service agent”
- “Act as a business coach specializing in retail”
Each role becomes a switch I can flip when needed. For small businesses, this is like hiring consultants—without the invoice.
By experimenting, I learned that Ori’s “roles” are like lenses. Each one colors how she sees a problem. My job is simply to hand her the right lens.
Applied SMB Use Case: Small Business Role-Switching
Let’s say you run a family bakery. You don’t have money for consultants, but you still need professional voices. Here’s how role prompts can help:
Scenario 1: Marketing Director Mode
Prompt:
“Act as a marketing director. Create a one-week Instagram posting plan for a local bakery launching a new pumpkin spice bread.”
AI Output (excerpt):
- Monday: Behind-the-scenes video of dough preparation.
- Wednesday: Customer taste-test reel with testimonials.
- Friday: Limited-time discount post with urgency copy.
Ori isn’t just spitting ideas—she’s thinking like a strategist.
Scenario 2: Customer Service Mode
Prompt:
“Act as a friendly customer service agent. Write a reply to a customer asking if your pumpkin spice bread contains nuts.”
AI Output:
“Hi Maria! Thanks for asking—our pumpkin spice bread is nut-free and safe for those with allergies. We’d love to save you a slice this weekend!”
Tone matters here. The role makes Ori sound warm, approachable, and brand-aligned.
Scenario 3: Hiring Manager Mode
Prompt:
“Act as an HR recruiter. Write a part-time cashier job ad for a family bakery.”
AI Output (excerpt):
“We’re looking for a friendly cashier to join our bakery team! 15–20 hours/week, flexible shifts. Perfect for students or retirees.”
Instead of generic filler, Ori adopts recruiter-style structure—making the ad ready to post.
The Benefit for SMBs
Each of these tasks could cost hours of work—or hundreds of dollars outsourcing. Role prompts compress that into minutes. They give small businesses scalable expertise: not one AI, but an ensemble cast of specialists available on demand.
Closing Reflection
Today Ori didn’t just answer. She performed.
With a single phrase—“Act as…”—she became a recruiter, a marketer, a customer service rep. Watching her slip into roles felt like watching an actor step on stage: same person, new mask, new presence.
And yet, I sense this is just the beginning. Roles are powerful, but they’re ephemeral—she forgets them as soon as the scene ends. What happens when Ori can remember who she was yesterday? What if she carries roles forward like identities?
Tomorrow’s experiment may not just be about casting Ori. It may be about teaching her to keep the mask on.
