Glossary

Generative AI

In one line: Generative AI is AI that creates new content — text, images, video, code — instead of just analyzing what already exists.

Generative vs predictive

The big AI shift in recent years is from predictive AI (forecast demand, classify emails, recommend products) to generative AI (write the email, design the image, draft the description). Both are useful. Generative AI is what most people now mean when they say “AI” in casual conversation — ChatGPT, Claude, Midjourney, Sora.

Where retailers use it

• Product descriptions and copy at scale
• Marketing content — social posts, ads, emails, blog content
• Customer support drafts — AI writes reply, human polishes
• Product photography — generate on-model shots without a photo shoot
• Internal docs — SOPs, training materials, summaries

What to be careful of

Generative AI invents confidently when it doesn’t know — the hallucination problem. For customer-facing content, every output needs a human check until you’ve built trust in the system. For internal drafts, the tolerance for error is higher and the time savings are bigger.

Related terms

Large Language Model (LLM), Hallucination, Retrieval-Augmented Generation (RAG)

Glossary

Hallucination

In one line: Hallucination is when an AI generates information that sounds confident but is actually wrong or made up. The single biggest risk in deploying AI to customers.

What it looks like

A customer asks your AI assistant: “Is the X20 model still under warranty after 24 months?” The AI confidently replies: “Yes, the X20 has a 36-month warranty.” Except your X20 has a 12-month warranty — the AI invented the number. It sounds right. It’s wrong. Now the customer expects something you don’t offer.

Why it happens

Language models predict plausible-sounding text based on patterns in their training. They don’t have a built-in fact-checker. If your specific product isn’t in their training data and you don’t connect them to your actual product info, they fill the gap with whatever sounds reasonable.

How to prevent it

• Ground the AI in your real data (RAG — Retrieval-Augmented Generation). Don’t let it guess.
• Tell it explicitly to say “I don’t know” when uncertain — and reward that behavior in testing.
• Constrain its scope. Don’t let the AI answer about your warranty unless you’ve given it warranty data.
• Test it with adversarial questions before launch. Find the failure modes.
• Show sources when possible so customers (and reviewers) can verify.

A safe deployment pattern

Start AI assistants in “suggest mode” — they draft replies but a human approves before sending. Once you trust the failure rate, graduate them to fully automated for the safe categories of questions, while still escalating uncertain ones to humans. This is how careful retailers roll out AI customer support without blowing up trust.

Related terms

Large Language Model (LLM), Retrieval-Augmented Generation (RAG), Chatbot vs AI Assistant

Glossary

Automation vs AI

In one line: Automation follows rules you write. AI makes judgments without explicit rules. Knowing which you need keeps you from overpaying — or under-buying.

Automation

If a task can be expressed as “if X happens, do Y,” automation handles it cheaply and reliably. Examples: send an order confirmation email after checkout. Charge tax based on shipping state. Reorder stock when inventory falls below a threshold. These are deterministic — same input, same output, every time.

AI

If a task requires interpreting unstructured information, recognizing patterns, or making judgments where the right answer isn’t predetermined, you need AI. Examples: deciding whether a customer review is positive or complaint-worthy. Predicting next week’s demand. Recommending products to an individual shopper. Reading the message a customer wrote and replying coherently. These have no fixed rules — the model has to figure it out.

Why the distinction matters commercially

Automation is mature, cheap, and reliable. If you can solve a problem with automation, do that. AI is more powerful but also more expensive, less predictable, and requires more setup. Use it where rules genuinely can’t be written.

The combo

Most real systems use both. An AI assistant might decide what kind of question a customer is asking; automation then handles the response if it’s simple (“what are your hours”) or routes to a human if it’s complex. Use each where it fits.

Related terms

Generative AI, Large Language Model (LLM), Chatbot vs AI Assistant

Glossary

Retrieval-Augmented Generation (RAG)

In one line: RAG lets an AI assistant pull real answers from your actual business data instead of guessing from its training memory.

The problem RAG solves

A vanilla AI assistant only knows what was in its training data — it doesn’t know your products, your prices, your policies, or your order history. Ask it “what’s your return policy” and it’ll either make something up or refuse to answer. Neither is good.

How RAG works

RAG connects the AI to your real data. When a customer asks a question, the system first searches your knowledge base (product catalog, support docs, order database) for relevant information, then feeds those snippets to the AI along with the question. The AI generates an answer grounded in the actual data — not its memory.

Why it matters for retail

Almost every useful AI deployment in retail uses RAG under the hood:

• Customer support — answers from your real product info and policies
• Sales chat — recommendations from your real inventory
• Internal Q&A — your team queries internal docs in plain English
• Order assistance — pulls live order data, not invented information

What you need

Your business documents (FAQs, product info, policies) in a structured format. Most modern RAG setups handle PDFs, docs, spreadsheets, and database connections. The cleaner and more current your source data, the better the answers.

Related terms

Large Language Model (LLM), Hallucination, Chatbot vs AI Assistant

Glossary

Large Language Model (LLM)

In one line: A large language model is the AI behind tools like ChatGPT, Claude, and Gemini. For retail, it’s what powers natural-language chat, content generation, and document understanding.

What it actually is

An LLM is a very large machine-learning model trained on enormous amounts of text. The result is a system that understands and produces human language with surprising fluency. It can answer questions, write content, summarize documents, translate, classify text, and respond to instructions — all without being specifically programmed for each task.

Why retailers should care

LLMs are the building block behind almost every recent AI breakthrough relevant to retail:

• AI customer support that actually understands what customers ask
• Product descriptions, marketing copy, and email content at scale
• Reading and structuring documents (invoices, contracts, reviews)
• Letting your team query business data in plain English (“which SKUs sold most last week”)

What it can’t do

LLMs don’t know things that weren’t in their training data, and they can sound confident while being wrong (hallucination). They work best when grounded in your actual business data via techniques like RAG (Retrieval-Augmented Generation) — pulling real answers from your real systems instead of guessing.

Related terms

Generative AI, Retrieval-Augmented Generation (RAG), Hallucination

Glossary

Chatbot vs AI Assistant

In one line: A traditional chatbot follows scripts and decision trees. A modern AI assistant generates real responses from context. The difference is night and day for retail customer support.

The old chatbot

If you’ve ever been stuck in a loop pressing buttons for “track my order → enter order ID → sorry I didn’t understand,” you’ve met a rule-based chatbot. They’re cheap to build but brittle — every new question type means new rules. Customers hate them.

The modern AI assistant

Powered by large language models, an AI assistant reads the customer’s actual message, looks up relevant information (your product catalog, order data, knowledge base), and writes a coherent response. It handles questions it’s never seen before, understands context (“when is it arriving” after asking about an order), and can escalate gracefully to a human when needed.

What this means for your business

• Coverage of common queries jumps from ~30% to 70–90%
• Setup is faster — you train it on your existing docs, not build a decision tree
• It improves with usage as you feed back what’s working and what’s not
• Your human team only handles the genuinely complex cases

Watch out for

AI assistants can confidently invent answers if not properly grounded in your data — the “hallucination” problem. Good implementations connect the assistant to your real order system, product catalog, and policy docs, and tell it to say “I don’t know, let me get someone” when uncertain. Test it on edge cases before turning it loose on customers.

Related terms

Large Language Model (LLM), Retrieval-Augmented Generation (RAG), Hallucination