Contrary to policy goals, tariffs on China didn't bring manufacturing jobs to the US. Data shows Chinese goods were just rerouted through third-party countries like Vietnam and Mexico, which in turn imported their components from China. This added complexity and cost without boosting US jobs.
Modern AI development tools are blurring traditional job descriptions in software companies. Designers can now ship code, product managers can design and prototype, and engineers can contribute to design, leading to a convergence of these once-siloed roles.
The current customs process is broken, relying on random sampling with low hit rates. The future is a system of "product passports" creating a library of continuously monitored, vetted goods. This shifts the paradigm from inspecting every transaction to verifying pre-approved product identities.
Data reveals a paradoxical trend: US manufacturing output is growing, but the number of manufacturing jobs is shrinking. This suggests that any reshoring or new production is being accomplished through industrial automation, not by bringing back assembly line jobs as policymakers had hoped.
AI's value in logistics extends beyond raw inference. Agentic systems learn from user interactions and messy data to create and refine Standard Operating Procedures (SOPs). This allows them to handle complex, recurring tasks with increasing efficiency over time, unlike models that start from scratch.
To overcome data sovereignty concerns, Altana brings its platform to its clients' data rather than centralizing it. It extracts learnings—like supply chain connections and model improvements—without copying sensitive details like pricing. This federated approach enables a shared intelligence network.
Simply restricting Chinese drone imports was insufficient to create a domestic industry. A successful US industrial policy paired these barriers with a strong demand signal—billions in committed capital from the Department of Defense. This combination created a viable market for US manufacturers.
The significant cost of advanced AI models ($20-$50 per million tokens) is no longer a trivial expense for internal development. Companies are now implementing observability, permissioning systems, and other controls to manage "token burn" and ensure a positive ROI on AI-assisted work.
Early on, founder conviction drives product. To scale, Altana injected deep domain experts in trade, customs, and logistics into the product process. Combined with new AI tools, these experts can now validate and execute on product judgments much faster than a founder could alone.