OpenPOS stores its tills as a wcpos_pos_outlet custom post type, its cashier sessions as rows the plugin writes during open and close, and the resulting transactions as standard WooCommerce orders in wp_posts tagged with an _op_session_id postmeta key. SleekAI's data-source wizard maps those records into named variables, so the bot's system message can reference the current register, the cashier on duty, and the count of orders in the open session.

That gives a public storefront bot the same stock view as the till, and gives a staff-only bot a real picture of the day's takings, refunds, and outstanding cash drawer. Display conditions scope the staff bot to logged-in shop manager and cashier roles so customer-facing visitors never see till totals or session ids. Multibot mode runs the public, staff, and account bots side by side without any of them seeing the others' data.

For shops with thousands of past tickets, push the order history into an OpenAI Files vector store of up to one gigabyte per file and the model retrieves only what matches the question. Every conversation is logged with model name, token usage, and page URL inside WordPress so managers can audit what the bot told a customer or a cashier.

The two questions that drive every retail shift are stock and balance. A cashier needs to know what is on the shelf and what the drawer should hold before they can close out. A customer wants to know if the size they want is here today and whether they can pick it up before the store closes.

A generic chatbot bolted on the storefront cannot answer either question correctly because it does not see OpenPOS at all. SleekAI changes the floor of acceptable. The staff bot quotes session 482, names the cashier on duty, and lists the two parked tickets blocking the close.

The public bot quotes the live stock count for the named outlet the customer is asking about and says whether pickup is realistic before 8pm. Both bots read the same WooCommerce database OpenPOS writes to, so neither contradicts the till. Retail margins do not survive an angry email caused by a wrong stock promise, and they do not survive a misbalanced drawer at the end of the week.

Having a tool that flattens both failure modes by reading real data and explaining it in plain language pays for itself the first time a cashier finishes a shift in five minutes instead of twenty. Owners also keep cost control. SleekAI is a one-time plugin licence with no per-message markup.

The AI calls run on the merchant's own provider key. OpenPOS licensing stays separate. The chatbot is part of the same WordPress install everyone already opens, not a fourth dashboard.