Why not tackle the declaration engine itself?
Stratech, Softpak and Stream are packed with integrations, tariff logic and years of operational settings. Rebuilding that costs a fortune and delivers nothing new. The pain is not in the engine. It is in everything that precedes it: documents arriving as PDFs, being read manually, and then keyed into the declaration by hand. That is where the hours go and where errors occur. Particularly now that the revised Union Customs Code takes effect in 2026 and compliance requirements are becoming stricter, that risk is greater than ever.
What does AI with Stratech and Softpak actually do?
The AI reads invoices, packing lists, certificates of origin and other customs documents, regardless of format. PDF, email, scan: structured data arrives in the declaration in an organised way, ready for the declarant to review. Incomplete or contradictory information is flagged before the declaration goes out, not afterwards when a correction has to be filed. The system adapts to the document flows your clients supply. No generic model that your team has to configure themselves.
How does HS classification work with AI?
HS classification remains human work. Responsibility for a declaration lies with the declarant, not with an algorithm. But the groundwork, looking up previous classifications, comparing descriptions, proposing a code based on similar goods, does not need to be done by a person. The AI proposes a classification and the declarant reviews, adjusts and confirms. Faster, with less risk of inconsistent codes across shipments, and with a demonstrable audit trail.
What changes for the team?
The team still works in Stratech, Softpak or Stream. There is no migration, no new system to learn, no data loss. The AI connects to the existing workflow: documents come in, the AI handles the reading and the proposals, the declarant reviews and submits. The time savings come from eliminating the tedious middle section. That is also precisely why adoption on the work floor is greater than with a system replacement: the system people know stays in place.
When is this not the right solution?
If the declaration engine itself is outdated, integrations are no longer functioning correctly, or you are actively considering moving away from Stratech or Softpak, then adding an AI layer around it is pointless. In that case, we rebuild the system from the ground up, with AI built in, using the Bonsai AI Digital Twin approach. But for most customs offices, that is not the question. The question is: how do I process more declarations with the same team, without the error rate rising as regulations tighten? For that, an AI layer around the existing software is the shortest route.