Look, you're copy-pasting the same three paragraphs into ChatGPT fifty times a month. Your product brief. Your company bio. Your approved terminology. Every. Single. Time.
Then the AI generates something that's technically right but sounds nothing like you: wrong words, wrong tone, doesn't actually understand your specific angle. So you edit. Then you paste again. Then you edit again.
That workflow is broken. And it's eating your drafts.
Here's the thing though: the fix isn't some fancy new feature. It's just thinking about AI writing in layers. Your persona teaches the AI how you actually sound. Your Knowledge Cards hold the durable facts you shouldn't have to repeat. Your task prompt describes this specific job. Keep those three separate, and suddenly you're not wasting cycles on context you've already given.
Let me show you how.
Why Repeating Context Wastes Drafts
Every time you copy-paste your company description into a prompt, you're starting from zero. The AI has no memory. It doesn't know that last Tuesday you explained your product does X but not Y. It doesn't know your approved way to talk about your audience. It never picks up on the terminology you prefer or the tone your customers actually expect.
So what goes down?
You draft an email to a prospect. You paste in your product description, your company story, your positioning. The AI spits out something technically accurate but uses words you'd never touch. You edit. You trim. You rephrase. Suddenly fifteen minutes gone on a draft that should've taken three.
Then you do it again. And again. Every email, every blog post, every proposal. Same context. Same explanations. Same waste.
The real problem is that most AI tools treat every prompt like a one-off. No system for storing context you use repeatedly. So you become the system. And honestly? Your memory is worse than a database.
ToneClone fixes this by separating what should be learned from what should be stored from what you need to describe right now.
Three Layers: Persona vs Knowledge Card vs This-Task Prompt
Think of AI writing as three completely separate problems.
Layer 1: How you sound (Persona)
Train a persona on 3 to 5 of your actual emails, LinkedIn posts, or documents. ToneClone learns your vocabulary, your sentence length, how you kick off ideas, how you land them. A persona isn't a description. It's a fine-tuned model. It learns patterns. That's literally it.
You train it once. You use it everywhere. You never re-explain it.
Layer 2: Facts that don't change (Knowledge Card)
A Knowledge Card holds the durable stuff: your product description, your company mission, your approved terminology, your bio, your current offer. These are things you write once and attach over and over. They don't shift week to week.
When you draft a sales email, you attach your Knowledge Card. The AI knows your product, your audience, your positioning. You're not pasting. You're not describing. You just attach it and it's there.
Layer 3: What you're doing right now (Task Prompt)
The task prompt is this specific draft only. The customer's name. The topic they raised. The deadline. The goal of this email. Today's date. One-off details that live nowhere else.
Keep it lean. Don't dump the Knowledge Card stuff in here again. Don't re-explain your persona. Just tell the AI what you want from this one thing.
Most people mash all three together into one monster prompt and then wonder why they're exhausted.
What Belongs on a Knowledge Card (And What Doesn't)
A Knowledge Card should hold:
- Your product or service description (1-2 sentences of what it is and who it's for)
- Key differentiators or unique features (short list, not a novel)
- Your company bio or professional bio (who you are, what you do, why you do it)
- Approved terminology or jargon you always reach for
- Your current offer or main call-to-action
- Event brief (if relevant: date, location, what you're speaking about)
- Pricing or package details you mention constantly
A Knowledge Card should NOT hold:
- One-off deadlines or time-sensitive facts (those go in the task prompt)
- Unverified claims or marketing hype you haven't proven (the AI will just repeat it)
- Secrets, passwords, or sensitive info that shouldn't be anywhere
- Entire product dumps or spec sheets (stick to essentials; the AI doesn't need everything)
- Personal details you don't want trained into a model
- Context that changes weekly (put dynamic stuff in the task prompt instead)
Simple rule: if you'd use it in five different drafts without changing it, it belongs on the Knowledge Card. If it's specific to this one draft, it goes in the task prompt.
How to Build One Focused Card and Attach It When Drafting
Start small. Don't try to cram your whole business onto the first card.
Open ToneClone and create a Knowledge Card. Name it something useful: "Product Brief" or "Company Bio" or "Sales Assets". Pick the surface you'll use most, probably the web editor or the browser extension.
Write one section at a time. Product description first. Then your bio. Then approved terminology or differentiators. Keep each section short. The AI doesn't need your full pitch deck. It needs a summary it can reference.
Once you've built the card, save it.
Now when you draft, you've got two options.
In the web editor: create a new draft, select your trained persona, then click "Attach Knowledge Card" and pick the one you just built. The card is now there when the AI generates.
In the browser extension: type or highlight your task prompt, open the ToneClone sidebar, select your persona, select your Knowledge Card, and hit generate. Same deal. Context is there without you typing it.
You're not pasting anything. You're not re-explaining. You're just attaching and writing.
Concrete Workflow: Drafting a Project Update Using a Product Brief Card
Say you're at Acme Design Co., and you need to send a project update to a client.
You've already trained a persona on three of your past emails. You've also built a Knowledge Card called "Client Project Brief" that has: what Acme Design Co. does, who you're serving, the main deliverables for this website redesign project, and the timeline.
You open the browser extension. You navigate to your email draft. You type your task prompt:
"Draft a project update for the Apex Corp website redesign. We're two weeks into the discovery phase. Status: on track. Next milestone is wireframes due March 15. Keep it warm but professional. Two paragraphs."
That's it. You don't paste the project brief. You don't explain what Acme Design Co. does. You just say what you want from this draft.
You select your trained persona from the dropdown. You select "Client Project Brief" as your Knowledge Card. You hit generate.
The AI now knows:
- How you sound (persona)
- What the project is and who you're serving (Knowledge Card)
- What you want to say right now (task prompt)
It generates a draft. You read it. It nails your tone. It gets the project. It uses language that fits the relationship. You send it.
You didn't repeat a single thing.
Verify Time-Sensitive Facts Yourself Before You Send
Here's the real talk: Knowledge Cards are solid for static info, but they're not your brain's replacement.
If you're dropping a deadline, a price, a date, or anything that moves, double-check it yourself before you hit send. Don't let the AI be your fact source for stuff that matters.
The Knowledge Card gives the AI context. You give the AI the specific facts. You're still the editor.
Common Mistakes (And How to Avoid Them)
Mistake 1: One giant card with everything.
Don't dump your entire company handbook into a Knowledge Card. The AI doesn't need it. It gets confused and pulls random details that don't fit the current draft. Build one focused card per context. One for product. One for your personal bio. One for a specific offer. Attach what you need.
Mistake 2: Treating the Knowledge Card as your memory palace.
A Knowledge Card is not a safe place to store secrets, client lists, or sensitive data. It's a tool you attach to drafts. If you wouldn't want an AI to reference it, don't put it on a Knowledge Card.
Mistake 3: Skipping persona training.
Some folks try to use Knowledge Cards alone and skip training a persona. Then they attach the card and the AI still sounds generic. Usually underperforms. Train a persona first. It teaches the AI your voice. Then attach a Knowledge Card. It teaches the AI your facts. You usually need both.
Mistake 4: Dumping task details onto the card.
Your Knowledge Card shouldn't change for every draft. If you keep adding one-off deadlines, customer names, or specific requests to it, you're using it wrong. Those belong in the task prompt. The Knowledge Card is permanent. The task prompt is temporary.
Wrap and Your Next Move
Three layers. Persona teaches sound. Knowledge Card stores facts. Task prompt describes this draft.
Build one Knowledge Card this week. Pick your most-repeated context. Could be your product description. Could be your professional bio. Could be an offer you mention in every sales email. Pull out just the essential info. Don't over-explain. Save it.
Next time you draft, attach it. Skip the copy-paste. Write faster. Sound like yourself.
Build your next card
If you keep pasting the same product brief, bio, or offer into every AI chat, put that material on a Knowledge Card once. Train a persona on your writing, attach the card when you draft, and keep one-off details in the task prompt.
Build a reusable Knowledge Card
Store the context you reuse, then draft in your trained voice across the surfaces you already write in.