Case Study

Aleana GPT

A custom AI that thinks, speaks, and advises in my voice.

Role

Prompt Engineer AI Designer Knowledge Curator

Built with

Custom GPT (ChatGPT) Prompt Engineering Knowledge Files

Type

Personal AI · Always On

Live

↗ Open Aleana GPT

What Is It

A digital extension of me available 24/7.

Aleana GPT

A custom GPT trained on my knowledge, experience, and communication style. It acts as a digital extension of me answering questions about my work, giving career and pricing advice, and engaging with users in the same tone I would use in a real conversation.

Anyone who finds me online through TikTok, LinkedIn, or my projects can talk to Aleana GPT and get the same advice, answers, and perspective they'd get from messaging me directly. Minus the wait.

The Problem

Building in public creates demand that doesn't scale.

The more visible you become, the more people reach out with the same questions:

How do I price my VA services?

The most common question I get especially from people who found Rei or my TikTok content. Everyone wants a quick, personalized answer rather than being pointed to a generic article.

Can I see your resume / how did you structure it?

The same reason I built the Resume Builder. But even after the tool existed, people wanted to talk through it "should I include this section?" "is this bullet point good?" conversational guidance, not just a form.

How did you get started / what would you do in my situation?

Career and pivot questions from students and early-career professionals who feel more comfortable asking a person than reading a blog post. They want someone who's been through it.

Answering the same questions manually across DMs, comments, and Discord doesn't scale and it means the people who ask at 2am never get a response at all.

The Solution

My knowledge. My tone. Always on.

A custom GPT that acts as a 24/7 version of me answering questions in my tone, with my knowledge, using the context I've given it through carefully crafted system prompts and uploaded knowledge files.

A

Aleana GPT

Online

U

I'm a social media VA charging $5/hr on Upwork. My client wants to extend but I feel underpaid. What would you do?

A

Honestly? $5/hr for social media work on Upwork is below market especially once Upwork takes their cut.

Before you do anything, calculate your actual floor rate: your monthly PHP expenses ÷ billable hours, grossed up for Upwork's 20% fee. That number might surprise you.

If it's a good client worth keeping, I'd raise to $7.50–$8/hr with a 30-day notice, framed around the value you've already delivered not around "I need more."

→ Try Rei to get your exact number

U

What if they say no?

A

Then you've learned your client's ceiling. That's useful information. A client who won't pay fair market rate as you grow is a client you'll outgrow anyway better to know now than a year from now.

What It Can Do

Six things it handles without me.

Responds as me

Trained on my tone, phrasing, and perspective. It sounds like I wrote it direct, warm, practical.

VA rate & career advice

Knows the PH VA market, platform fees, niche benchmarks, and how to frame a rate conversation.

Resume guidance

Helps users think through what to include, how to write bullets, and when to use which format.

Points to my tools

Knows about Rei and the Resume Builder links users there when their question is better answered with a tool.

Answers about my work

Can explain how I built my projects, why I made certain decisions, and what my background is.

Always available

Responds instantly at any hour no DM queue, no "I'll get back to you." Immediate, personalized answers.

How It Was Built

Prompt engineering is the whole product.

Layer	What I Used	Purpose
Platform	`ChatGPT Custom GPT`	No-code AI deployment, public URL
Personality	`System prompt`	Defines tone, values, communication style
Knowledge	`Uploaded files`	My projects, rate data, career context
Guardrails	`Prompt instructions`	Keeps responses on-brand, avoids hallucination

The Hard Parts

Making it sound like me, not like a generic AI

The default GPT voice is helpful but generic. Getting it to match my tone required iterating on the system prompt extensively defining not just what to say but how: short sentences when making a point, direct recommendations instead of hedging, warmth without being performatively enthusiastic. I tested it by asking questions I'd actually been asked in DMs and refining until the answer felt like something I'd write myself.

Giving it accurate knowledge without hallucination

LLMs confidently make up specific numbers (rates, fees, stats). I uploaded structured knowledge files with the real figures from Rei's niche benchmarks, platform fee breakdowns, and market data and instructed it to cite these rather than generate from base model knowledge. For anything outside the knowledge files, it's instructed to say so rather than guess.

Deciding what's in scope and what isn't

A personal AI can creep into trying to answer everything. I deliberately scoped it: VA rates, career guidance, my projects, and questions about me. Outside that, it redirects not because it can't answer, but because staying focused keeps it genuinely useful rather than being a general-purpose chatbot wearing my name.

Results

One link. Infinite conversations.

24/7 Available without any queue or delay

1 Single link shareable everywhere I appear online

∞ Conversations handled without adding to my DMs

"The best use of an AI isn't to replace a person it's to let a person be in more places at once."

What I Learned

What worked, what I'd change.

What Worked Well

Scoping it tightly to my actual domain made a huge difference. A focused GPT that's excellent at VA career advice is more useful than a broad one that's mediocre at everything.

What I'd Do Differently

I'd add a short intro message that sets expectations what it knows, what it doesn't, and when to use Rei instead. First-time users sometimes don't know what to ask.

Unexpected insight: Building this forced me to articulate my own perspective clearly enough to teach it to a machine. That process writing out my values, my reasoning, my advice patterns made me a better communicator even outside the GPT.