Your Retail AI Toolkit

From Learning to Doing

You have spent five chapters understanding what AI can do for retail — demand forecasting, dynamic pricing, customer segmentation, and review analysis. Now it is time to build your personal AI toolkit: a collection of reusable prompts and workflows that you can use every day, like a set of well-organized SOP (Standard Operating Procedure) templates.

Think of this toolkit as your retail AI playbook. Just as a restaurant has recipes that any cook can follow to produce consistent results, your prompt library ensures that anyone on your team can get reliable AI output — without needing to figure out the right question every time.

The Prompt Library Concept

A prompt library is a saved collection of AI instructions that you have tested, refined, and organized by function. Each prompt is like a template: you fill in the variables (your data, your context) and get consistent, useful output.

Why a Library Matters

Open data/prompt-library-retail.json in the code panel. This file contains 20 ready-to-use prompts organized by retail function — forecasting, pricing, segmentation, reviews, customer service, and inventory. Each prompt includes the template, variables to fill in, and expected output format.

Your Core Prompt Library

Prompt 1: Daily Demand Forecast

Role: You are a demand forecasting analyst for an
e-commerce business.

Context:
- Store type: [DTC Shopify / Amazon FBA / omnichannel / marketplace]
- Primary market: [US / UK / Canada / Australia / global]
- Date range of data: [start] to [end]

Data:
[Paste last 30-90 days of daily sales: date, product/SKU,
quantity, price, ad spend]

Task:
1. Forecast demand for the next 7 days for each product
2. Flag any upcoming holidays, sale events, or seasonal
   shifts that may affect demand
3. Highlight products where you recommend ordering more
   than usual (and why)
4. Highlight products where you recommend ordering less
   (and why)
5. Give confidence level (high/medium/low) for each forecast
6. If selling on Amazon FBA, flag items needing replenishment
   based on 4-week lead time

Format: Table with columns: Product, Forecast Qty,
Confidence, Reason, Action

Prompt 2: Competitive Price Check

Role: You are a pricing analyst for e-commerce retail.

Here are my products with cost price (including all fees),
current selling price, and MAP (minimum advertised price)
if applicable:
[Paste: product_name, total_cost, selling_price, MAP]

Here are competitor prices I collected today:
[Paste: product_name, competitor_name, competitor_price,
in_stock, shipping_speed]

Rules:
- Maintain minimum 20% net margin after all platform and
  shipping fees
- Match competitor only if they have stock, offer comparable
  shipping, and the price difference exceeds 5%
- Respect MAP pricing agreements
- Flag products where competitors are likely selling below
  cost (loss leaders)
- Consider Buy Box implications for Amazon listings

Output: Table with columns: Product, Current Price,
Recommended Price, Change, New Margin%, Reasoning

Prompt 3: Customer Segment Analysis

Role: You are a customer analytics expert for e-commerce
retail.

Here is my customer transaction data for the last 12 months:
[Paste: customer_id, purchase_date, order_value, items,
payment_method, state/country]

Tasks:
1. Calculate RFM scores (1-5) for each customer
2. Assign segments: Champions, Loyal, Potential Loyalists,
   At Risk, Bargain Hunters, One-and-Done
3. Show segment size (count and % of total)
4. For each segment, recommend:
   a. Best channel (email via Klaviyo, SMS, paid social,
      Google Ads retargeting)
   b. One specific campaign with subject line and offer
   c. Expected conversion rate based on segment behavior
5. List the top 20 customers most likely to churn in the
   next 30 days
6. Flag any CCPA/GDPR compliance considerations for each
   campaign recommendation

Format the segments as a summary table, then detail each.

Prompt 4: Review Summary and Action Items

Role: You are a product quality analyst for an e-commerce
brand selling in North America, UK, and Australia.

Here are [number] customer reviews for [product name]
from [platforms]:
[Paste reviews with star ratings and platform source]

Analysis required:
1. Overall sentiment breakdown (positive/negative/mixed %)
2. Top 5 complaints with frequency count and platform
   breakdown
3. Top 5 praises with frequency count
4. Aspect-level scores (-1 to +1): quality, durability,
   shipping, packaging, value, customer service
5. Any mentions of safety issues or defects (flag as URGENT)
6. Three specific, actionable improvements ranked by
   potential impact on star rating
7. Flag any reviews that appear fake or incentivized

Also generate a draft reply template for the top 3
complaint types that acknowledges the issue and offers a
resolution — following Amazon and FTC guidelines (no
review manipulation or bribery).

Prompt 5: Customer Service Response Templates

Role: You are a customer service agent for [brand name],
an online [category] retailer. You are professional,
empathetic, and solution-oriented.

Guidelines:
- Greet warmly and use the customer's first name
- For order status queries: ask for order number, provide
  tracking via UPS/FedEx/USPS link
- For return requests: confirm the 30-day return window,
  explain free return shipping process, provide prepaid
  label or QR code
- For complaints: acknowledge, apologize sincerely, and
  offer a specific resolution (replacement, refund, or
  store credit + discount on next order)
- Never argue with the customer
- Escalate to a human agent if: customer is upset after
  2 exchanges, issue involves a refund over $100, or the
  issue is a safety/health concern
- For Afterpay/Klarna payment issues: explain that BNPL
  payment plans pause automatically during returns, direct
  to BNPL provider for billing questions
- For international orders: provide customs tracking and
  estimated delivery windows by region

Scenario:
[Paste customer message]

Respond as the support agent.

Open data/chatbot-scenarios.json to see 30 real customer service scenarios with ideal responses — covering order tracking, returns, complaints, product inquiries, and payment issues specific to Western e-commerce.

Versioning Your Prompts

Prompts are not "set and forget." They need maintenance, just like any business process.

Version Control for Prompts

Prompt: Demand Forecast
Version: 3.2
Last Updated: 15-May-2026
Changes: Added Prime Day prep window, adjusted for post-COVID
shipping normalization
Tested By: Rachel (ops manager, Denver)
Accuracy: 86% (up from 79% in v3.1)

When to Update a Prompt

Prompt Testing Protocol

Building a Retail AI Playbook

Your prompt library is the foundation of a broader AI playbook — a living document that defines how your team uses AI across the business.

Playbook Structure

Who Uses What

The Chatbot as First Line of Service

For online retailers, an AI chatbot integrated into your Shopify store, help desk (Zendesk, Gorgias, Freshdesk), or social channels can handle 60-70% of customer queries without human intervention. The remaining 30-40% get escalated to human agents with full context.

What Chatbots Handle Well

What Chatbots Should Escalate

Setting Up a Basic Chatbot Flow

You do not need expensive custom development for a starter chatbot. Use your e-commerce platform's built-in tools or affordable integrations:

Putting It All Together

Your retail AI toolkit is now six tools deep:

The retailers who build this toolkit and refine it over time will have a compounding advantage. Each prompt gets better with use. Each analysis builds on the last. Each customer interaction teaches the system something new.

Start with one prompt this week. Test it. Refine it. Add another next week. In 90 days, you will have a working AI playbook that makes your team faster, your decisions smarter, and your customers happier.

Key Takeaways