Ecommerce and POS integration is finished when the voice can answer the question

A storefront (Shopify, BigCommerce, WooCommerce, a custom headless build) and a point-of-sale (Toast, Square, Clover, NCR Aloha, Revel) agreeing on one catalog and one order object. The storefront pulls items, prices, variants, and inventory from the POS on a schedule. The storefront pushes orders into the POS through an order-injection API when a web shopper checks out. Payment settles on the POS side. Refunds go back through the POS. That is the standard two-system contract, and it is well covered by connectors and docs. The thing that is not well covered is what happens when a third channel, the phone, joins the same contract.

Because callers ask questions a product page answers visually. On a web PDP, a photo, an ingredient list, and an allergen icon do most of the work. On the phone, the same caller asks 'is the palak paneer spicy?' or 'what is in the Caesar?' or 'can I do a half-and-half pizza?' or 'does the teriyaxx bowl have gluten?' None of those are answerable from SKU, title, price, variant, and stock count alone. The catalog record has to carry descriptive metadata (spice scale, ingredient list, dietary flags, combo compatibility, pronunciation hints) that a Shopify-to-POS connector does not ship. The integration is not finished when the POST works. It is finished when the voice can answer the question.

Six that come up every week on real restaurant menus. First, spice or heat level on a scale the voice can reason about (mild, medium, hot, very hot, or 0-4). Second, sweetness for dessert and drink menus. Third, ingredient list in plain English, not kitchen shorthand. Fourth, dietary flags the caller will actually ask for (gluten-free, vegan, vegetarian, dairy-free, nut-free, halal, Jain). Fifth, combo and modifier compatibility: whether this item can be halved on a pizza, paired with a specific side, or swapped for a different protein. Sixth, pronunciation hints for menu items that are hard for text-to-speech to say correctly. None of these live in the default Toast, Square, or Clover catalog schema. They have to be enriched onto the POS item ID during onboarding.

No. PieLine sits on top of the same POS item IDs your web connector already reads. During onboarding, PieLine's team scrapes your live menu (the one your Shopify or ordering page already shows), maps every scraped item to the matching POS item ID, and then layers voice-only enrichment (spice, allergens, pronunciation, combo rules) on top of that mapping. When a phone order is built, it is written into the POS through the same order-injection API your web store uses (Toast POST /orders/v2/orders, Square POST /v2/orders with an idempotency key, Clover POST /v3/merchants/{mId}/orders plus /line_items). One contract, one catalog root, a thicker record on top.

The POS item ID is the join key. When the POS catalog changes (new item, price change, 86'd out-of-stock flag, seasonal special), the enrichment layer re-reads the POS catalog and reconciles. New items are flagged for enrichment during the next onboarding pass. Price, modifier, and availability changes on existing items flow through automatically because the voice catalog points at the POS record, not a snapshot of it. The operator never maintains a second menu.

Sometimes part of it. Shopify product descriptions often carry ingredient prose and dietary tags because the web merchandising team wrote them. BigCommerce and custom headless builds can carry a metafield for spice level if the team added one. In practice, almost nobody has pronunciation hints, combo compatibility, or consistent allergen flagging on the web catalog, because the web presentation does not need them. PieLine's onboarding normalizes whatever already exists in the ecommerce description and fills in the gaps, rather than forcing the web team to rebuild their PDP copy.

Two paths. If the question is about something the catalog does cover but the voice is uncertain (an ambiguous modifier, a near-match item name, a pronunciation the speech-to-text mangled), the voice asks a clarifying question. If the question is genuinely outside the catalog (a catering quote, a complaint, a one-off dietary concern the flags do not capture), PieLine transfers the call to a designated human with the full transcript and customer context already attached, so the staff member does not ask the caller to start over. About 90 percent of calls get handled end-to-end by the voice; roughly 10 percent transfer.

Same business day on supported POS systems. PieLine's onboarding team scrapes the online menu, maps items to POS item IDs, configures rules (delivery zones, minimum orders, hours, specials), and adds dish descriptions covering spiciness, sweetness, ingredients, and dietary info. The operator forwards the phone line or sets PieLine as overflow in about 10 minutes. Active call monitoring and AI refinement run in the first month at no extra cost, so any enrichment gap surfaced by real calls is closed against production traffic, not a made-up test set.

Clover, Square, Toast, NCR Aloha, and Revel are the deepest integrations, with the same order injection APIs your web connector uses. Fifty-plus more POS systems are supported on the standard tier. Orders are tagged with a channel source so phone shows up as a distinct line in the revenue-by-channel report your ecommerce team already reads.

Five questions. Which POS endpoint do you call on order creation. Do you set a channel tag on the order so reporting can split phone from web and in-store. Do you pass an idempotency key. Do you read menu and stock state directly from the POS rather than maintaining a separate menu file. And what enrichment fields do you layer on top of the POS item ID for voice use. If any of those answers is vague, the voice channel is running off-contract and your integration story is only two-thirds real.