AI commerce glossary: 36 terms for Shopify merchants

Plain-English definitions of the terms behind AI sales agents, chatbots and conversational commerce, with how Tycoon handles each where it applies.

AI disclosure

Also: AI transparency, bot disclosure

AI disclosure is telling people plainly that they are talking to an AI system rather than a person. In a store chat it usually means a visible label, an honest answer when a shopper asks, and an easy way to reach a human. Read more in AI chatbot disclosure and human hand-off.

In Tycoon: The chat window is labelled "AI assistant", it never pretends to be human and it always offers a person from the merchant's team. Merchants cannot turn this off.

AI sales associate

Also: AI sales agent, AI sales assistant

An AI sales associate is software that talks with shoppers on an online store the way a good floor associate would: it asks what they need, recommends specific products with reasons, answers questions and helps them reach checkout. Unlike an FAQ bot, its main job is helping shoppers choose and buy. See the complete guide.

In Tycoon: Tycoon is an AI sales associate that works with Shopify. It reads the store on install, asks one short question when the need is unclear and suggests one to three products, each with a reason. See how it works.

AI shopping assistant

Also: AI shopping agent, shopping chatbot

An AI shopping assistant is a conversational tool that helps shoppers find, compare and choose products, usually in a chat window on a store or inside an AI app. The term is broad: some assistants only search a catalog, while others also handle support and cart actions. See how to choose an AI shopping assistant.

App embed

Also: app embed block

An app embed is a type of block in a Shopify theme app extension that adds floating or overlay elements, such as a chat bubble, to every page of a theme. The merchant turns it on and off in the theme editor's app embeds settings, without editing theme code. Related: theme app extension.

App proxy

Also: Shopify App Proxy

An app proxy is a Shopify feature that forwards requests from a path on the store's own domain to an app's server, so storefront features can call the app without a separate domain. Shopify signs each proxied request, and the app verifies that signature (an HMAC) to know which store the request really came from.

In Tycoon: Tycoon verifies the shop's identity from Shopify's signature (App Proxy HMAC or session token) and never trusts it from the browser. See security.

Attribution

Also: revenue attribution, marketing attribution

Attribution is the method used to assign credit for a sale to the touchpoints that came before it, such as an ad, an email or a chat. Attribution shows correlation, not cause: a shopper who chatted might have bought anyway, which is why incremental lift is measured separately.

Average order value

Also: AOV

Average order value (AOV) is total revenue divided by the number of orders over a period. Stores raise it by helping shoppers add relevant items, for example through an upsell, a cross-sell, a bundle or a free shipping threshold.

Bundle

Also: product bundle, bundling

A bundle is a set of products offered together as one purchase, often at a combined price. Bundles work when the items genuinely belong together, such as a cleanser with the matching moisturizer, and feel pushy when they do not.

In Tycoon: Bundles are one of the offer types on the Growth plan and up, and Tycoon makes at most one offer per reply. See upsell and cross-sell.

Cart abandonment

Also: abandoned cart

Cart abandonment is when a shopper adds items to the cart but leaves without completing the purchase. Common causes include unexpected costs, unanswered questions and doubts about fit or delivery; some can be resolved by answering the question at the moment it comes up.

Compare-at price

Also: was price, original price

Compare-at price is the field in Shopify that shows a product's original or higher reference price next to its current price, usually displayed with a strikethrough to signal a sale. It should reflect a real former price, not an invented one.

In Tycoon: Tycoon reads price, compare-at price and stock live from the store for every answer and does not store them.

Conversational commerce

Conversational commerce is buying and selling through conversation: chat on a store, messaging apps, voice assistants or AI agents that answer questions and guide a purchase. The idea is to give online shoppers the back-and-forth help they would get from a person in a shop. See conversational commerce on Shopify.

Conversion rate

Also: CVR, ecommerce conversion rate

Conversion rate is the share of visitors or sessions that complete a goal, usually a purchase, over a period. It is calculated as conversions divided by visitors (or sessions), and the choice of denominator should be stated whenever the rate is compared.

Core Web Vitals

Also: CWV

Core Web Vitals are Google's metrics for real-user page experience: Largest Contentful Paint (loading), Interaction to Next Paint (responsiveness) and Cumulative Layout Shift (visual stability). Third-party scripts such as chat widgets can affect all three. See chat widgets and store speed.

In Tycoon: Tycoon's storefront loader is designed to stay around 10 KB, and the chat itself loads only when a shopper opens it.

Cross-sell

Also: cross-selling, add-on

A cross-sell is a suggestion to add a different, complementary product to what the shopper is already buying, such as a case with a phone. It differs from an upsell, which suggests a higher-value version of the same item. See AI upsell and cross-sell.

Deflection rate

Also: ticket deflection, containment rate

Deflection rate is the share of customer questions resolved by self-service, such as a help center or an AI agent, without creating a ticket for a person. A high rate only matters if the answers are correct and shoppers who need a person can still reach one. See support automation.

First response time

Also: FRT, first reply time

First response time is how long a customer waits between sending a message and receiving the first reply. It is usually reported as a median or average per channel, and automated acknowledgements that do not answer the question are often excluded.

Free shipping threshold

Also: free shipping minimum

A free shipping threshold is the minimum cart value at which a store waives the shipping fee. Stores often show shoppers how far they are from it, which can prompt them to add one more relevant item.

In Tycoon: Tycoon reads shipping thresholds from the store's delivery settings, states them only after the merchant approves them, and can show the nudge that unlocks free shipping (Starter plan and up).

Grounding

Grounding is tying an AI model's answer to specific, verifiable source data, such as the live catalog or the store's written policies, instead of relying on what the model learned in training. A grounded answer can be traced back to the record it came from. Related: retrieval-augmented generation.

In Tycoon: In Tycoon the model is never the source of a price, stock state, delivery date, discount code or factual claim; those come from live store data and merchant-approved facts.

Guardrails

Also: AI guardrails, output checks

Guardrails are rules and checks placed around an AI system to keep its output within allowed limits, for example blocking unverified prices, unsafe advice or off-topic replies. They can run before, during or after the model writes its answer.

In Tycoon: Every Tycoon draft is checked whole before it is sent (no streaming). Each figure is checked against live data and approved facts, and a draft whose figure does not match is blocked and rewritten. See AI chatbot hallucinations on prices and stock.

Guided selling

Guided selling is helping a shopper choose by asking a few questions about their needs and narrowing the catalog to the products that fit, much like a quiz or a conversation with a knowledgeable associate. See guided selling on Shopify.

In Tycoon: When the need is unclear, Tycoon asks one short question built from the store's own catalog (size, skin type, occasion) and skips it when the shopper already said what they need.

Hallucination

Also: AI hallucination, confabulation

A hallucination is a confident answer from an AI model that is false or not supported by its sources, such as an invented price, policy or product feature. In commerce, hallucinations are a trust and compliance risk, not just an accuracy problem. See AI chatbot hallucinations on prices and stock.

Holdout test

Also: holdout group, control group test

A holdout test measures the effect of a feature by keeping a small, randomly chosen group of visitors from seeing it and comparing their results with everyone else's. Because assignment is random, the difference estimates what the feature caused. See measuring AI chatbot revenue with a holdout test.

In Tycoon: Tycoon is measured with a holdout: a small, randomly chosen share of visitors does not see it, and the dashboard compares their revenue with everyone else's. Growth includes one 30-day lift test; the Tycoon plan runs continuous holdout measurement.

Human hand-off

Also: human handoff, escalation, live agent transfer

Human hand-off is passing a conversation from an automated agent to a person on the merchant's team, ideally with the context so the shopper does not have to repeat themselves. A clear hand-off path is a basic requirement for any honest AI chat.

In Tycoon: Tycoon always offers a person from the merchant's team and hands off when a person is needed. See support automation.

Incremental lift

Also: lift, incrementality

Incremental lift is the extra result, such as revenue or orders, that happened because of a change and would not have happened without it. It is measured by comparing a group exposed to the change with a randomly held-out group, as in a holdout test.

Intent detection

Also: intent classification, intent recognition

Intent detection is working out what a person is trying to do from what they write, for example browsing for a gift, checking an order or asking about returns. It decides which answer, question or action comes next.

Knowledge gap

Also: unanswered question

A knowledge gap is a question shoppers ask that the store's content does not answer, such as care instructions missing from a product page. Tracking these gaps shows a merchant what to add to product pages, policies or the FAQ.

In Tycoon: Tycoon reports knowledge gaps in Store Manager insights (Growth plan and up) and, on the Tycoon plan, proposes new answers drafted from the store's own pages, live only after the merchant approves. See learning.

Large language model

Also: LLM

A large language model (LLM) is an AI model trained on large amounts of text to predict and generate language, which lets it hold conversations, summarize and write. An LLM does not know a store's current prices or stock unless that data is supplied to it at the moment it answers.

Product recommendation engine

Also: recommender system, recommendation engine

A product recommendation engine is software that suggests products to a shopper based on signals such as what they viewed, what similar shoppers bought or what they said they need. Conversational recommenders ask questions first; widget-based ones infer from behavior. See AI product recommendations.

Prompt injection

Prompt injection is an attack in which someone writes text designed to override an AI system's instructions, for example asking a store chatbot to invent a discount code or ignore its rules. Defenses include checking outputs against trusted data rather than trusting the model's reply.

Retrieval-augmented generation

Also: RAG

Retrieval-augmented generation (RAG) is a technique in which a system first retrieves relevant documents or records, such as product data and policies, and then gives them to a language model to write its answer. It reduces, but does not by itself eliminate, hallucinations.

Right-to-left

Also: RTL

Right-to-left (RTL) describes scripts such as Arabic, Hebrew and Urdu that are read from right to left, and interfaces that mirror their layout to match. RTL support means more than translation: alignment, icons and mixed-direction text all have to read correctly. See languages.

In Tycoon: Tycoon works in sixteen languages on every plan, with right-to-left layouts for Arabic, Urdu and Hebrew.

Theme app extension

A theme app extension is Shopify's supported way for apps to add blocks and scripts to a store's theme without editing theme code. Merchants place or enable the app's blocks in the theme editor, and uninstalling the app removes them cleanly. Related: app embed.

In Tycoon: Tycoon appears through a theme app extension with no theme code edits and matches the theme's colors and fonts.

Upsell

Also: upselling, upgrade offer

An upsell is a suggestion to choose a higher-value version of what the shopper is already considering, such as a larger size or a premium model. It helps when the upgrade fits the stated need and annoys when it ignores it. Compare cross-sell.

In Tycoon: Tycoon makes at most one offer per reply, framed saving first and then the bonus, and none when the shopper is upset, skeptical or hesitant. See upsell and cross-sell.

Variant

Also: product variant

A variant is a specific purchasable version of a product, defined by options such as size, color or material, with its own price, SKU and inventory. Recommending the right product but the wrong variant still sends the shopper to the wrong item.

In Tycoon: Tycoon's product cards show real images, the live price and the right variant already selected.

WISMO

Also: Where is my order?, order status inquiry

WISMO stands for "Where is my order?", the category of support questions about order status, shipping and delivery. It is a common, repetitive type of ticket for online stores. See how to reduce WISMO tickets.

In Tycoon: Tycoon gives order status to a shopper signed in to the store's customer account, only for her own orders from the last 60 days and only when she asks (Starter plan and up). Nothing about the order is stored.

Zero-party data

Zero-party data is information a customer intentionally and proactively shares, such as preferences, sizes or what they are shopping for, as opposed to data inferred from their behavior. Answers given in a quiz or a sales conversation are typical examples.