So your AI actually knows your business — not just the generic internet. This is the difference between AI that sounds like everyone else and AI that sounds like you.
Start with what you have: your existing product data. Whether you're on Shopify or another platform, extract everything about your products first.
For each product: names, full descriptions, materials and fabric tech, price points, categories, variants (colors/sizes), performance data (bestsellers, seasonal trends), and customer language from reviews.
#!/bin/bash # Shopify product pull — adjust for your platform curl -X GET "https://your-store.myshopify.com/admin/api/2024-01/products.json" -H "X-Shopify-Access-Token: $SHOPIFY_TOKEN" -H "Content-Type: application/json" > /tmp/products-raw.json echo "Pulled $(cat /tmp/products-raw.json | jq '.products | length') products"
Pro tip: Don't skip the metafields — that's where your custom product attributes live (fabric technology, care instructions, fit details).
Raw product data is messy. Before it goes into your knowledge base: clean HTML tags, standardize categories, extract key features, and add context about why those features matter to customers.
This is where most people fail. They create one giant "products.md" file and wonder why their AI can't find relevant information. Instead, organize into focused, searchable files.
Product names and core descriptions, key features and benefits, price positioning and value props, bestsellers and seasonal highlights.
Every fabric and technology you use, why you chose each material, performance benefits and care instructions, how to explain technical features to customers in plain language.
How your brand sounds and feels, words you use vs. words you avoid, tone for different situations (ads vs. customer service), examples of on-brand vs. off-brand language.
Real sizing info in customer language, common sizing questions and actual answers. Plus what makes you different from key competitors, your unique selling propositions, how to position your products vs. alternatives.
Product data tells your AI what you sell. Business context tells it who you are and why it matters. This is the difference between an AI that recites features and one that sells your vision.
Not marketing fluff — the real story. Why you started, what problem you're actually solving, your mission beyond selling products, what your community stands for.
Actual customer demographics. Real language from reviews. Common objections. Purchase triggers. Use cases. Source from customer service transcripts, reviews, surveys, social comments, and return reasons.
Build a knowledge file of actual customer questions and your best answers. Pull from your CS team's most common tickets. Don't guess — use real data.
Different AI agents need different knowledge. Your ad agent doesn't need detailed care instructions. Your customer service agent doesn't need competitor ad positioning. Create focused knowledge sets for each role.
Brand voice and messaging · Key differentiators and USPs · Current promotions and seasonal focus · High-performing copy examples · Competitor positioning · Target audience insights
Brand voice for email communication · Product recommendations logic · Customer lifecycle messaging · Seasonal content calendar · Promotional strategies · Personalization data points
Complete product catalog · Sizing guides and fit advice · Care instructions and materials · Return/exchange policies · Shipping information · Common issue resolutions · Escalation procedures
knowledge/
├── shared/
│ ├── brand-voice.md
│ ├── product-catalog.md
│ └── customer-personas.md
├── ad-agent/
│ ├── competitor-positioning.md
│ ├── high-performing-ads.md
│ └── seasonal-messaging.md
├── email-agent/
│ ├── email-voice.md
│ └── lifecycle-sequences.md
└── cs-agent/
├── sizing-guide.md
├── policies.md
└── common-issues.md
A stale knowledge base is worse than no knowledge base. Your products change, your positioning evolves, your customers shift. Your AI's knowledge needs to keep up.
Product updates: New launches, discontinued items, price changes, seasonal availability, new bestsellers.
Market updates: Competitor changes, new customer feedback patterns, seasonal messaging shifts.
Performance updates: Which knowledge files your AI references most, common questions not covered, agent performance gaps.
#!/bin/bash # Weekly cron to flag knowledge gaps # Check for new products not in knowledge base # Flag customer questions not answered by current knowledge # Report on most-referenced knowledge files
Use these as starting points. Fill in your specifics and your knowledge base will be live in days, not weeks.
# [Product Name] ## Overview **Category:** [Primary category] **Price:** $[XX] **Key Features:** [2-3 main benefits] **Best For:** [Customer type/use case] ## Product Details **Materials:** [Fabric + technologies] **Fit:** [Fit description in customer language] **Available In:** [Colors/sizes/variants] ## Positioning **Vs Competitors:** [How this compares] **Unique Selling Points:** [What makes it special] ## Customer Language **They Call It:** [How customers refer to this] **Common Questions:** [Frequent questions] **Review Highlights:** [Key themes from reviews] ## Cross-Sell **Pairs Well With:** [Complementary products]
# Brand Voice Guide ## Who We Are **Personality:** [3-5 adjectives] **Mission:** [What you're really trying to accomplish] ## How We Sound **Tone:** [Professional/casual/friendly] **Perspective:** [Brand as friend/expert/etc.] ## Words We Use **Preferred Terms:** [Brand-specific language] **Emotional Triggers:** [Words that resonate] ## Words We Avoid **Generic Terms:** [Overused industry language] **Negative Associations:** [Words that hurt positioning] ## Voice Examples **Good email subject:** [Example] **Bad email subject:** [Counter-example] **Good ad copy:** [Example] **Bad ad copy:** [Counter-example]
# Competitor Positioning ## [Competitor Name] **Their Positioning:** [How they position themselves] **Their Strengths:** [What they do well] **Their Weaknesses:** [Where they fall short] **Our Advantage:** [Why customers choose us instead] **Messaging Against:** [How to position in ads/content] ## Overall Market Position **Our Unique Space:** [Where we fit] **Key Differentiators:** [Top 3 things that make us different] **Competitive Advantages:** [What competitors can't copy]
# [Persona Name] ## Demographics **Age:** [Range] | **Location:** [Geographic distribution] **Lifestyle:** [Activity level, interests, values] ## Shopping Behavior **How They Find Us:** [Traffic sources] **Price Sensitivity:** [What drives vs stops purchase] ## Language & Communication **How They Describe Our Products:** [Real customer language] **Their Pain Points:** [Problems they need solved] **Common Objections:** [What stops them from buying] ## Product Preferences **Bestsellers for This Persona:** [Top products they buy] **Seasonal Behavior:** [How buying changes by season]
"Premium athletic wear designed for active lifestyles. Shop our collection of high-quality activewear. Free shipping on orders over $75."
Generic. Boring. Sounds like every other brand.
"Finally, shorts that keep up with your 6am HIIT class and your 8pm yoga flow. Our coconut-blend fabric moves with you while keeping you comfortable all day. Join 50,000+ who've made the switch."
Specific benefits. Real use cases. Community proof.
"Thank you for contacting us. Please refer to our size chart for sizing information. If you need further assistance, please don't hesitate to reach out."
"Great question on sizing! Our shorts fit true to size with a relaxed feel. Since you're between a medium and large, I'd go large — the coconut fabric has just enough stretch without losing shape. And if it's not perfect, our 30-day exchange has you covered."
| Metric | Before | After | Change |
|---|---|---|---|
| Email open rates | 18% | 27% | +50% |
| Ad click-through rates | 1.2% | 1.8% | +50% |
| CS satisfaction score | 6.8/10 | 8.9/10 | +31% |
| Time editing AI responses | 40 min/day | 8 min/day | -80% |
| Brand consistency score | 3/10 | 9/10 | +200% |
Join THE AI INCOME LAB — free community where e-commerce founders share how they're running AI-powered operations that actually drive revenue.
Join Free →