White paper

From manual to intelligent

How AI automates HS code classification and turns it into a strategic advantage

April 21, 2026

13 min

The new reality of HS classification

Your classification team receives an urgent alert that new tariffs take effect in 72 hours, impacting thousands of SKUs across multiple product lines. The regulatory change affects high-volume import categories that drive significant revenue. Your team needs to validate duty impacts, update financial models, and coordinate with procurement while maintaining customs clearance and avoiding compliance exposure.

This scenario has played out repeatedly as regulatory volatility becomes the dominant reality facing global trade teams. The elimination of de minimis thresholds in certain import lanes forced one electronics retailer to reclassify 12,000 SKUs within 45 days, a volume impossible to handle manually without significant outsourcing costs and error risk. Meanwhile, volatility in the application of Sections 232 and 301 tariffs, along with the introduction and reversal of tariffs under the International Emergency Economic Powers Act (IEEPA), resulted in tens of millions of modifications to HS codes and tariff information annually, triggering repeated cycles of classification review and adjustment.

It’s a challenging enough situation even for companies whose classification procedures are airtight; imagine, then, how much worse the situation is for companies who have thousands of wrong or inconsistent classifications. This is no longer an operational inconvenience, it's a compounding supply chain risk. Classification errors flow directly into landed cost miscalculations, sourcing strategy failures, compliance exposure, and speed-to-market delays. In today's interconnected supply chains, classification has become a risk multiplier that determines whether organizations can execute their business strategies or scramble reactively to regulatory changes.

The acceleration of everything

The 2026 Thomson Reuters Global Trade Report reveals the magnitude of this transformation. Supply chain management has surged from 35% of professionals citing it as a priority in 2024 to 68% in 2026, nearly doubling in two years. Simultaneously, regulatory compliance concerns have risen from 21% to 33%, reflecting how classification accuracy now sits at the intersection of operational resilience and compliance integrity.

This convergence isn't coincidental. U.S. tariff volatility, cited as the most impactful regulatory change by 72% of trade professionals, represents just one pressure point. Teams simultaneously navigate export control expansions, EU Carbon Border Adjustment Mechanism (CBAM) reporting requirements, Corporate Transparency Act implications, and the elimination of de minimis exemptions in key trade lanes.

Classification has become the control point where cost, compliance, and speed converge.

The strategic elevation of classification

This pressure has fundamentally elevated classification's strategic importance. Accurate HS codes now determine the quality of every trade decision:

Duty calculations directly impact landed costs and financial projections
Compliance filings create audit exposure when they don't match operational reality
Strategic initiatives, from sourcing strategies to product launch timing, succeed or fail based on the reliability of underlying trade data

The scale alone makes manual-only approaches unsustainable. Modern portfolios span tens of thousands of SKUs across dozens of markets, with thousands of HS codes requiring interpretation of similar language and country-specific nuances. When you factor in over 155 million tariff and classification data changes published globally in 2025, human capacity struggles to maintain accuracy and speed at the required scale.

As one global trade manager put it, “We went from managing compliance to enabling strategy. Classification determines whether our procurement team can execute their sourcing plans, whether finance can model scenarios accurately, and whether we can respond to market changes faster than competitors.”

The four fault lines: Where manual classification breaks down

These challenges tend to surface along four predictable fault lines.

Fault line No. 1: Unmanageable scale

Modern classification faces a complexity explosion that overwhelms human capacity:

Code proliferation. Thousands of HS codes with frequently ambiguous language require careful interpretation
Jurisdictional multiplication. The same product may require different optimal classifications across markets
Continuous change. New, split, or expired codes force ongoing reclassification across entire portfolios
Portfolio breadth. Large, multimarket product catalogs amplify workload exponentially

The scale challenges exceed human capacity. A portfolio of 10,000 SKUs across 10 markets creates 100,000 potential classification decisions. Add regulatory changes, new products, and audit requirements, and the workload quickly exceeds any reasonable team capacity.

Fault line No. 2: Data quality crisis

Data quality issues create cascading problems that undermine classification accuracy:

Inconsistent descriptions. Product names and attributes differ across systems, creating research friction
Abbreviation confusion. Shorthand creates interpretation challenges and slows research
No single source of truth. Similar products receive different classifications, and errors propagate downstream

This creates the "AI data quality paradox"; simply adding an "AI layer" onto messy data risks accelerating inconsistencies rather than eliminating them.

Fault line No. 3: Information bottlenecks

Manual processes create information flow problems that impact cross-functional execution:

Slow propagation. Classification updates reach finance, operations, and procurement too late for decision windows
Collaboration friction. Manual processes limit the frequency and quality of alignment
Decision lag. By the time classification insights reach decision-makers, market conditions may have already shifted

Fault line No. 4: Reactive posture that multiplies risk

Manual classification inherently operates in reactive mode:

Discovery delays. Issues surface after filings or during audits rather than during decision-making
No forward signal. Limited ability to predict which SKUs will be impacted as regulations shift
Resource strain. Rising workloads increase error risk and burnout

Together, these fault lines permanently force organizations into a reactive posture.

The convergence effect

These fault lines don't operate in isolation, they compound exponentially. Scale pressures worsen data quality. Information bottlenecks slow your response times. Reactive postures increase resource strain, creating more bottlenecks and forcing your organization into a permanently reactive cycle.

What's required isn't more people or faster research, it's classification intelligence that scales with the size and pace of your modern product portfolios. Breaking this cycle requires fundamental transformation in how classification intelligence flows through your organization.

This is the operating model behind Thomson Reuters ONESOURCE Global Classification AI, which transforms your workflow by reducing manual research effort and enabling your experts to focus on validation, judgment, and higher-value trade decisions. Classification AI only delivers value when it understands the unique complexity of global trade. Here's how ONESOURCE Global Classification AI is purpose-built for this challenge, not adapted from generic AI tools.

How it works: The intelligence behind classification automation

The AI methodology: Built for trade complexity

ONESOURCE Global Classification AI combines multiple layers of machine learning specifically trained in trade classification challenges. Unlike generic AI systems, our models understand the nuances of trade language, regulatory context, and classification logic that drive accurate HS code determination.

Machine learning that gets smarter over time

The system learns continuously from expert decisions within your organization. When classification experts review AI suggestions and make final determinations, the system captures these decisions to refine future recommendations. This method creates a feedback loop where the AI becomes increasingly aligned with your organization's classification standards and risk tolerance.

For example, if your team consistently classifies certain electronic components using specific criteria due to your company's interpretation of regulatory guidance, the system learns these patterns and applies similar logic to comparable products. This organizational learning means accuracy improves over time while maintaining consistency with your established practices.

Deep global content in practice

“Deep global content” means the AI draws from our comprehensive trade intelligence database, which includes harmonized tariff schedules from 220+ countries and territories, helping the AI understand country-specific classification nuances.

This content foundation enables the AI to suggest classifications that consider not just product descriptions, but the broader regulatory and interpretive context that expert classifiers use in their decision-making.

Handling edge cases and uncertainty

The system is designed to be transparent about confidence levels. ONESOURCE Global Classification AI:

Provides confidence indicators that help experts prioritize review time on uncertain classifications
Surfaces multiple options with explanatory context when a single clear answer isn't apparent
Enables workflows to flag potential edge cases for expert attention rather than forcing low-confidence suggestions, and escalates complex scenarios to ensure expert judgment drives final decisions on ambiguous products

This approach ensures that uncertainty is surfaced early, not hidden, allowing experts to focus their specialized knowledge where it's most needed.

Seamless integration architecture

ONESOURCE Global Classification AI integrates with existing trade management and ERP systems through:

API connections that allow classification suggestions to flow directly into master data management workflows
Export and import functionality that works with existing data formats and approval processes
Audit trail integration that maintains compliance documentation within your established systems

The system is designed to enhance existing workflows rather than replace them, allowing teams to maintain their established approval processes while accessing AI-powered research and suggestions.

Quality assurance through expert oversight

Every AI suggestion includes supporting rationale drawn from relevant tariff notes, classification guidance, and similar product patterns. This transparency allows experts to quickly validate the reasoning behind suggestions and make informed decisions about acceptance, modification, or rejection.

The system tracks these expert decisions to continuously improve suggestion accuracy while maintaining complete audit trails for compliance purposes.

Implementation: Proven starting points

Successful implementations tend to begin with clearly defined use cases that align with operational priorities. By focusing on areas where scale, change, or complexity is highest, teams can apply classification intelligence incrementally while maintaining control and oversight.

Implementation principles: Designed for control, not disruption

Classification intelligence delivers maximum value when introduced deliberately, not through wholesale transformation. Leading organizations follow three proven principles:

Incremental adoption

Teams start with a focused use case — such as new product classification, regulatory-driven updates, or batch processing of legacy inventories — before expanding to broader portfolios. This approach ensures early wins while eliminating operational risk and building organizational confidence.

Expert-in-the-loop oversight

AI enhances research capabilities and pattern recognition, while classification experts retain complete authority to review, validate, and finalize all decisions. Professional judgment remains central, never automated. This workflow maintains compliance standards while accelerating throughput.

No “big bang” transformation

Classification intelligence integrates seamlessly into existing workflows and systems without requiring process overhauls or extensive retraining. Teams continue using familiar interfaces and established procedures, with AI working behind the scenes to enhance accuracy and efficiency.

This measured approach allows organizations to realize immediate value while building toward comprehensive automation at their own pace.

Why Thomson Reuters

Thomson Reuters brings together decades of trade content expertise and purpose-built AI capabilities designed specifically for global trade classification. This foundation includes comprehensive global classification content, trade specific model training, integrated workflow design, and established implementation experience across complex trade environments.

An unmatched trade content foundation

Classification intelligence is only as strong as the content behind it. ONESOURCE Global Classification AI is powered by one of the world’s most comprehensive trade content foundations, developed over more than 20 years of serving global trade professionals.

This foundation includes tariff schedules across 220+ countries and territories with real-time updates and interpretive guidance, historical classification decisions and rulings from customs authorities worldwide, and cross-referenced product data that links attributes to relevant HS code families, exclusions, and regulatory context. Industry-specific classification patterns refined through thousands of real-world implementations add further depth and relevance.

This is not just data. It is curated trade intelligence that understands the why behind classification decisions, not just the what.

Purpose-built AI for trade complexity

ONESOURCE Global Classification AI reflects years of specialized development focused exclusively on trade classification challenges. It is trained on real-world classification scenarios and trade-specific data, enabling accurate interpretation of tariff language, legal notes, and classification logic.

Proven implementation expertise

Technology alone does not transform classification. Thomson Reuters brings deep experience deploying trade and compliance solutions across thousands of organizations worldwide.

Our teams understand how classification workflows integrate with ERP, trade management, and compliance systems. A compliance-first design preserves audit trails and expert oversight throughout the automation journey, while scalable architecture supports growth from targeted pilots to enterprise-wide deployment without disrupting established processes.

Trusted for mission-critical trade operations

Global enterprises rely on Thomson Reuters for their most critical tax, trade, and compliance decisions. That same commitment to accuracy, reliability, and regulatory expertise underpins our classification intelligence.

Enterprise-grade security protects sensitive trade data. Continuous content updates keep pace with evolving regulations. A global support infrastructure provides implementation guidance and ongoing optimization as complexity grows.

When classification accuracy directly impacts duty optimization, compliance risk, and operational efficiency, organizations choose the provider with the deepest trade expertise and the most reliable technology foundation. That is the Thomson Reuters advantage.

Results you can expect

Organizations that apply classification intelligence within their workflows often experience outcomes such as:

More consistent classification decisions supported by shared data and standardized processes
Improved efficiency for classification teams by reducing repetitive research effort
Stronger audit readiness through clearer documentation and rationale consistency
Better cross functional alignment as classification updates are more accessible downstream
Greater capacity to manage regulatory change across large and evolving product portfolios
Confidence to expand usage over time, as teams validate outcomes and standardize processes at their own pace

These results support more informed trade decisions while preserving expert judgment and compliance rigor.

Ready to get started?

As regulatory complexity accelerates and supply chains expand, classification accuracy has become a foundational element of supply chain risk management. The question is no longer whether classification processes must evolve, but how organizations will scale expertise, consistency, and oversight to meet what comes next.

Discover why ONESOURCE Global Classification AI is the intelligent choice for forward-thinking trade organizations by contacting us today.