Trade Agent

Objective and outcome

Most products in a catalog classify and resolve cleanly, and you clear them in bulk. A small share do not, and those are where the time goes: reading through rulings looking for the one that matches your product, and often still not knowing which piece of information is missing.

The Trade Agent is aimed at those. It reads the rulings and rules that bear on the product, works the determination, and where it cannot get there on its own, asks you the specific questions that would settle it. The outcome is that a hard determination stops being a research project and becomes a few questions to answer, with the reasoning saved to the record either way.

Where you meet it

There is no separate agent surface. The agent works inside the lens you are already in, through the guide box in the product side sheet. You will see it in HS Classification, in Country of Origin, and on those attributes from Trade Readiness.

The guide box sits alongside the data already on file for that attribute, so you always see what you hold today next to what Altana proposes.

The flow

1. Run the analysis in the lens. The agent's work happens as part of that run, so there is nothing separate to start.
2. Open a product that has a suggestion or needs attention.
3. Read the guide box. It leads with the suggested next step, so you know what Altana thinks you should do before you read anything else.
4. Where confidence is high, the next step is to accept. Review the reasoning if you want it, then accept.
5. Where confidence is low, the next step is the questions. Answer the ones the agent has already surfaced, and watch the confidence move as it recomputes against your answers.
6. Open a chat if you would rather work the product through in conversation than answer questions one at a time.
7. Accept, choose an alternative, or enter your own value. The determination is yours.

Key features

A suggested next step, always

The guide box never leaves you on a blank page. Every state carries a recommendation for what to do next, and it follows the confidence: accept when confidence is high, answer the questions when it is low, and supply or request the input when something is missing.

Questions you answer without driving a chat

Where the agent needs more to reach a determination, it surfaces the specific questions in the guide box for you to answer directly. This is deliberate. Chat is useful, but most of the time you do not want to run a conversation, you want to answer the two questions standing between you and a defensible answer.

The questions are specific to the determination rather than generic prompts for more detail. On a classification they are the attributes that would decide the heading. On an origin determination they are the input whose origin or cost would change the result.

Chat when you want it

Choosing to chat opens the same agent that raised those questions, with the product's context already loaded. Use it when a product needs discussion rather than a couple of facts, or when you want to push back on a suggestion and see how the reasoning holds up.

CROSS ruling checks

Part of what the agent does is read the CBP rulings that bear on your product. It evaluates, per HS code, whether there is specific ruling logic that has to be applied, and where there is, the determination follows the ruling rather than the general rule.

This is the work that otherwise falls to you: searching CROSS for the analogous case and deciding whether it governs your product. The ruling that applied is shown as part of the reasoning, so a determination is grounded in CBP's own published decisions and stays defensible if someone asks about it later.

Confidence and the reasoning behind it

Every suggestion carries its confidence, and a short explanation of why Altana reached it. The intent is that you never see only the answer: you see the answer, how much weight to put on it, and the reasoning you would need if someone asked you to defend it. On classifications, the full General Rules of Interpretation reasoning is available on the suggested code and on the alternatives.

Alternatives and override

You can see the alternative determinations from the guide box, each with its own reasoning, and accept one instead. You can also enter your own value at any point. Nothing is written to your catalog until you accept it.

Your answers persist

What you tell the agent is saved to the product as a fact about that product, not left in a conversation that disappears. The next determination that needs it already has it, so answering a question once raises confidence on that product going forward.

FAQ

Do I have to use the agent? No. Where a suggestion is high confidence you can accept it from the table without opening anything. The agent is there for the products that need it.

Do I have to use chat? No. The questions in the guide box are answerable in place. Chat is an option, not the path.

Can I turn it loose on my whole catalog? Not today. The agent works a product at a time, as part of a run you trigger. Bulk clearing of confident suggestions is how you move volume.

Why do I have to trigger the run? Analysis runs on demand rather than continuously, so you control when it happens. Each lens has its own control, described on that lens's page.

Will it decide anything on its own? No. Every determination is a suggestion until a person accepts it, and anything Altana is not confident in is routed to you rather than settled quietly.

What if it is wrong? Choose an alternative or enter your own value, both from the guide box. If you want a second opinion before it stands, send the product to manager review in the HS Classification lens.

Which lenses does it work? HS Classification and Country of Origin.