4 Infusing Voice & Crafting Tailored Instructions
Module 4: Infusing Voice & Crafting Tailored Instructions
From the course: Design Your Personal AI Brain – Conspire With Yourself: AI Systems for Personal Growth
Welcome to Module 4!
By now, you’ve chosen your first use case and built a meaningful knowledge base. Fantastic!
Now it’s time to make your system truly feel like you.
This module is all about adding your voice, tone, and personality to the system — and when appropriate, crafting short instructions that help your AI act in the right role or mindset.
Why This Step Matters
A great AI brain doesn’t just respond with relevant facts — it sounds like you, reflects your values, and works in a way that aligns with your thinking style.
Without this step, even the best resources can feel impersonal or off-brand.
With it, your system becomes a true extension of your mind.
You Don’t Always Need Instructions
Let’s start with something important:
You don’t have to write detailed instructions for every system.
In fact, many of the best-performing personal AI setups rely mainly on:
-
High-quality context
-
Clear examples of your tone or work
-
Prompting in the moment
Overly rigid instructions can actually reduce flexibility or cause AI to behave in odd ways.
So, only use tailored instructions if you really need to guide the system’s role, tone, or formatting style.
When Instructions Are Helpful
There are definitely times when custom instructions make a difference. For example, when you want your AI to:
-
Take on a specific role (e.g. “act as an experienced learning designer”)
-
Use a consistent tone (e.g. “be warm, clear, and slightly witty”)
-
Deliver responses in a specific format (e.g. markdown slides or bullet-point summaries)
In these cases, instructions help your AI set the right frame before generating responses.
Examples of Effective Custom Instructions
Here are some examples of well-phrased, short instructions:
-
🎓 “Act as an engaging online educator. Focus on clarity, motivation, and practical takeaways.”
-
💼 “Take the tone of a thoughtful senior career coach. Offer concise and supportive advice.”
-
✍️ “You are a ghostwriter trained in my Substack style. Write in a warm, reflective, slightly poetic tone.”
Notice how these are:
-
Short
-
Clear
-
Aligned to the purpose
-
Free of jargon or over-engineering
Infuse Your Voice Through Examples
One of the most powerful ways to teach your system how you sound is to share actual examples of your writing.
Look for:
-
Emails or messages you’re proud of
-
Social posts or essays that reflect your tone
-
Favourite metaphors or phrases you often use
-
Feedback you’ve given that shows your style
Upload 2–3 examples, and label them clearly — for example:
“Professional_Tone_Example” or “Reflective_Style_Sample”.
These give your AI something concrete to pattern-match against — often better than trying to describe your voice in abstract terms.
Optional: Create a System Card
If you’re using tools like Claude or ChatGPT, you can go a step further by creating a System Card.
This is like a mini profile or spec sheet for your AI.
It includes:
-
🧠 Who the AI is (its role + purpose)
-
✍️ How it should speak
-
📘 What reference material it can draw from
-
⚠️ What to avoid (e.g. “don’t use overly corporate language”)
This helps ensure consistency — especially useful if you want to reuse the same system across multiple projects or share it with a collaborator.
Your Task: Bring Your Voice to Life
Here’s what to do now:
✅ Upload or paste 2–3 short examples of writing or communication that reflect your tone
✅ Decide whether you need tailored instructions — and if so, write just 1–2 short, role-based lines
✅ Test your system with simple prompts like:
“Summarise this article in my style”
“Draft a reply in the tone of my writing samples”
Make note of what works — and don’t worry about getting it perfect the first time. This is a living system.
Up Next…
In Module 5, we’ll test your system with real tasks and show you how to iterate and save what works.
This is where your AI brain starts becoming a true collaborator.
Let’s dive in!
