The Compliance Challenge Has Become An Engineering Problem. AI Can Help You Solve It
Danny Fields, Chief Technology and Customer Operations Officer Avalara, an agentic AI leader in global tax and compliance.
gettyWhen I talk to technology leaders about global trade compliance, I tend to get one of two responses. Either their business has already faced disruption—shipments held at customs, a refund claim that cascaded through the financials, a market entry delayed by months because the regulatory environment shifted—and they’ve been handed a mandate to make sure it doesn’t happen again. Or they haven’t, and the mandate has arrived anyway.
Either way, the message is the same: Compliance is no longer a problem that finance teams can manage through spreadsheets and experience. It’s now an engineering problem that can be solved with AI.
The volume, velocity and interconnectedness of global regulatory change has crossed a threshold where human experience alone can no longer keep pace. Tariff schedules update. HS classification rules shift. There were more than 417,000 taxability updates across the U.S. and Canada alone in 2025, according to the Avalara Tax Changes 2026 Midyear Update.
Then add in the transactions themselves subject to those changes. Over five days of Cyber Week last year, we processed 1.14 billion API calls, representing $187 billion in tax transactions and peaking at 9,124 transactions per second. The moment a compliance decision needs to be made faster than a person can make it, at a scale no team can match, across a level of complexity no spreadsheet can model, is when you’ve crossed from a compliance problem into a systems architecture problem.
The pace of change is what makes this so urgent. This year alone, companies have had to navigate legal challenges to tariff policies, fluctuating duties across major trade corridors and renewed threats of retaliatory tariffs across global markets, all while managing geopolitical uncertainty that has fundamentally changed how boards think about international growth. According to our recent data, 83% of business leaders said their cross-border operations are more complex than they were just 12 months ago.
What concerns me from a systems perspective isn’t the pace of change itself, but what happens downstream when systems can’t keep up. Every regulatory update ripples through thousands of transactions simultaneously. Any delay in implementing that change creates compliance exposure that cascades: shipment holds, customs penalties, disruptions that travel up the supply chain and into the P&L. It’s essentially a data latency problem.
Consider what this means in practice for classification alone. The World Customs Organization updates Harmonized System (HS) codes on a five-year cycle, but country-specific tariff rates tied to those codes can change multiple times a year, and in the current environment, they are.
An AI system trained on a comprehensive classification database and updated continuously can absorb a regulatory change and propagate it across an entire product catalog in minutes. A human team cannot. AI-based classification systems now draw on databases of millions of real-world product classifications. At this scale and pace of change, I believe AI is the only viable architecture for the problem.
In fact, nearly 40% of businesses we surveyed said they’ve delayed expansion specifically because of regulatory uncertainty. Companies that should be growing aren’t because their systems weren’t built to handle the pace. When compliance becomes a growth constraint, that’s the moment technology leaders have to step up.
The old model assumes a regulatory environment that is stable enough to configure once and revisit quarterly. It’s not. The companies getting this right are building systems designed from the ground up for constant change. That means a few things in practice:
• Real-Time Regulatory Ingestion: Your compliance systems need to consume regulatory changes as they happen, not after a quarterly review cycle. This requires API-connected content pipelines that surface rate changes, classification updates and policy shifts at the moment they’re published, not after they’ve created exposure.
• Agentic Compliance Workflows: This is where I think the most interesting engineering work is happening right now. Rather than building systems to alert human reviewers of compliance changes, the more powerful architecture would be one where AI agents identify the change, assess the impact across transaction data and implement required adjustments—with human review of edge cases.
• Transparency And Traceability By Design: Audit readiness is often an afterthought in system design, bolted on after the fact. The better approach is to build traceability into the core data model so every compliance decision has a logged chain of evidence: what changed, when it changed, what transactions it affected and what the system did in response.
According to our data, nearly half of U.S. businesses are already using AI as a core component of their cross-border risk management. But there’s a significant laggard population, with 13% of organizations not yet using AI for compliance at all. That gap is going to widen. The companies investing now in adaptive infrastructure are building a structural advantage, while others are accumulating a compliance debt that will eventually come due.
For a long time, compliance has been treated as finance’s problem. The tools were built for tax professionals and the processes were designed around human expertise, not system architecture. But that meant that when regulatory volatility spiked, as it unmistakably has, the systems weren’t ready.
Today, however, compliance needs the expertise of technology leaders, not to replace tax expertise, but to engineer the infrastructure that makes tax expertise scalable. A tax professional who understands customs classification can’t manually review 10,000 SKUs every time a tariff schedule changes. But a well-architected system with AI-assisted classification, real-time rate feeds and agent-driven change management can. That combination creates compliance certainty at velocity.
More than half of global respondents in Avalara’s research said that tariffs and changing duties have created the greatest operational burden in recent times for their businesses. The path forward is building systems that turn complexity into a managed, data-driven process, treating AI as production infrastructure. By turning complexity into a managed, data-driven process, organizations can stay ahead of risk, maintain consistent compliance and position themselves to grow confidently across borders. The time to build it is now.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
