Agent Training Center
Train your marketing agent with brand knowledge, voice examples, sales rules, and content preferences before it acts.
What the agent currently knows
Where the agent learned this
Teach the agent how you sound
We almost killed this product 3 times. Here's what changed when we stopped guessing and started shipping the rough version…
If your content calendar is held together by Notion and willpower, book a 20-min walkthrough. I'll show you the loop we run.
1/ Most founders treat content like a chore. Compounders treat it like a product surface. Here's the weekly operating loop ↓
Fair pushback. The honest answer is: it won't replace strategy. It replaces the 14 hours/week of execution around it.
- Write like a founder, not a corporate brand
- Be direct and useful
- Avoid hype words
- Use specific examples and numbers
- Prefer short, punchy paragraphs
- Make CTAs feel conversational
- Avoid generic AI buzzwords
Rules are derived from your highest-scoring voice samples and applied to every draft before it ever reaches the approval queue.
Show the agent what to copy — and what to avoid
Why it works: Concrete, scroll-stopping, and earns the click.
Why it works: Takes a stance instead of hedging. Builds authority fast.
Why it works: Reader can apply it the same day. Saves and shares go up.
Why it works: Feels like a peer suggestion, not a sales close.
Why it works: Real timeline, real numbers, real mistakes — high trust signal.
Why it fails: Says nothing, attributes nothing, teaches nothing.
Why it fails: ‘10x your growth' with no evidence trips the claims library.
Why it fails: Reads like a press release. Loses the founder edge.
Why it fails: ‘Leverage AI-powered synergy' fails the voice model.
Why it fails: ‘Comment YES if you agree' damages long-term trust.
What the agent is allowed to say
- Saves time on content planning
- Helps maintain brand consistency
- Supports human approval workflows
- Improves content planning discipline
- ROI improvement
- Revenue impact
- Growth guarantees
- Competitive comparisons
- “10x guaranteed ROI”
- Medical / legal / financial advice
- Claims about competitors by name
- Fake testimonials
- Manipulative urgency
The agent uses this library before generating, scheduling, or replying.
How the agent responds to common pushback
Every draft is scored against your voice samples — anything <85% gets rewritten or sent back.
The agent runs execution. Strategy, taste and final calls stay with your team.
Default mode is Supervised. Nothing publishes externally without your approval.
Five autonomy levels, claims library, and per-channel approval gates. You decide what's auto vs reviewed.
Risky comments (angry, legal, competitor) auto-escalate to a human with full thread context.
Week 1 is calibration. Compounding traction usually starts around week 3–4 of consistent posting.
How the agent handles incoming conversations
- Positive comments
- Simple questions
- Thank-you replies
- Clarifications about features
- Angry comments
- Refund / pricing complaints
- Legal / medical / financial topics
- Competitor comparisons
- Sensitive personal issues
- Public accusations
- Threats
- Harassment
- Legal complaints
- PR risk
- Platform policy concerns