Accurate within a few percent if you have HS codes assigned to your products and current rate tables loaded. The bot does not replace a customs broker for high-value or commercial imports, but for typical DTC orders the estimate is reliable enough to set the customer's expectation. The bot quotes 'estimated' rather than 'final' to set the right framing, and surfaces the actual carrier or customs broker as the authority on the final invoice.

HS codes are usually assigned per product in postmeta, often _hs_code or as a product attribute. Some stores assign at the category level; the bot can read either layer. If your catalog does not yet have HS codes, this is a one-time setup that takes a few hours for most catalogs and pays back many times over in better duty accuracy and customs clearance speed.

Yes. The bot reads whatever data the DDP service writes back into WordPress (final landed cost, included duty, included VAT) and quotes it as the final number. For DDP customers the bot's role shifts from estimation to explanation, since the cost is already calculated at checkout. The bot tells the customer what each line of the landed cost represents so they understand they will not be charged again at delivery.

Honestly, complex jurisdictions need a specialist tool. The bot can quote the general duty and tax rates but it should not pretend to handle every nuance of Brazilian ICMS or Indian GST on imports. The system prompt is configured to be honest about uncertainty for those destinations and to suggest the customer use the formal customs broker or DDP carrier for an authoritative estimate. Honesty about edge cases preserves trust on the easy cases.

Yes, if your data flags trade-agreement origin. Goods of US origin shipped to Canada qualify for USMCA preferential rates, EU origin to UK qualifies for the TCA, and so on. If your products have an origin field set in postmeta, the bot can apply the preferential rate. Without origin data, the bot defaults to MFN rates, which is the safer overstatement for the customer.

Thresholds are stored in your data, not in the bot's training. The data source wizard maps a thresholds file or a custom post type that you maintain. When a threshold changes (the EU eliminating the 22 EUR exemption in 2021, for instance), you update one record and the bot quotes the new number immediately. This is the right architecture because relying on the model's training would mean the bot quoting outdated thresholds within months.

Yes. SleekAI reads what the tax engine writes to postmeta and order data. WooCommerce Tax, Avalara, TaxJar, and custom engines all leave their output somewhere readable in WordPress. The bot quotes the engine's calculation rather than running its own, which keeps the customer-facing estimate consistent with the actual checkout calculation. No drift between the bot's number and the checkout number.

Within the rules, yes. Splitting an order to stay under a de minimis, shipping to a different destination country with lower duty, or using a different carrier service can all reduce landed cost. The bot can suggest these when the math supports them. It will not suggest under-valuation or false declarations, which are illegal and damage the store's relationship with the carrier and customs authorities.