Marco Kuijper - Overseeing Property Assessment in the Netherlands

The property tax assessment profession stands at a curious inflection point. While we debate the merits of AI-powered valuation models in conference rooms and oversight meetings, taxpayers are already using ChatGPT to write their appeal letters. This disconnect between our cautious approach and the public's rapid adoption reveals something fundamental about where our industry is heading, and the trust questions we must address along the way.

The AI Arms Race Nobody Asked For

In the Netherlands, where municipalities reimburse successful appeals, an entire industry has emerged around challenging assessments. These tax agents are now weaponizing large language models to generate appeal letters at scale. As Marco Kuijper from the Netherlands Council for Real Estate Assessment observes, municipalities are responding in kind, preparing to use AI systems to handle the deluge of automated protests.

This isn't the thoughtful integration of technology we might have envisioned. It's an arms race that threatens to reduce the assessment appeals process to competing algorithms. Yet it also forces us to confront an uncomfortable truth: while we deliberate about explainability and public trust, the public has already decided AI is trustworthy enough for their purposes.

Beyond the Black Box Problem

The explainability challenge has long been our industry's favorite reason to resist advanced machine learning models. We cling to our clustering techniques and multiple regression analyses not because they're superior, but because we can explain them to a skeptical property owner or an inquiring journalist.

But as Kuijper points out, "People are also trusting AI in other parts of their life. If you're driving a car in the future, it will also probably use AI to drive you from A to B. And then people are trusting AI... why shouldn't they trust AI for property taxes?"

The question isn't whether AI can explain itself in terms a taxpayer would understand. The question is whether we're holding property assessment to a higher standard of explainability than we demand from our navigation apps, medical diagnoses, or credit decisions.

The Maintenance Detection Revolution

While we debate abstract questions of trust, practical applications are emerging that could transform how we collect and verify property data. Cyclomedia's use of AI to automatically detect maintenance levels from street-view imagery represents the kind of targeted, specific application where machine learning excels.

This isn't about replacing assessors with algorithms. It's about augmenting human judgment with consistent, scalable data collection. A model that can reliably distinguish between excellent, good, fair, and poor maintenance across thousands of properties doesn't need to explain its reasoning in human terms, it needs to be accurate and consistent.

The Decentralization Dilemma

The Netherlands' 342 municipalities face a challenge familiar to many U.S. states: how do you maintain assessment quality standards across jurisdictions of wildly varying sizes and resources? Smaller municipalities often lack staff who understand basic ratio study statistics, let alone advanced modeling techniques.

This capacity gap suggests that AI adoption might follow an unexpected pattern. Rather than large jurisdictions leading the way, we might see AI-as-a-service models that allow smaller offices to access sophisticated valuation tools they could never build themselves. The democratization of advanced analytics could level the playing field, if we can solve the trust equation.

Building Trust Through Transparency, Not Simplicity

The path forward isn't about making AI simple enough to explain. It's about building institutional transparency around how we develop, test, and deploy these tools. This means:

• Publishing our accuracy metrics regularly and prominently
• Creating clear appeals processes that don't disadvantage taxpayers who question AI-generated values
• Establishing oversight frameworks that audit outcomes, not algorithms
• Investing in public education about how modern assessment works

As one oversight professional to another, Kuijper's advice resonates: "Don't look at your own property tax system as something that is static or stable for a long time." The world is changing around us. We can either shape that change or be shaped by it.

Key Takeaways

For Assessors: The AI adoption question isn't if, but how. Start with specific, bounded applications like image analysis or data quality checks before tackling valuation models.

For Leadership: The trust challenge is institutional, not technical. Focus on building transparent processes and clear accountability rather than perfect explainability.

For Policy Professionals: The emergence of AI-powered tax agents changes the game. Consider how appeal processes and reimbursement structures might need to evolve to handle automated challenges at scale.

The property tax assessment profession has always balanced technical accuracy with public accountability. AI doesn't change that fundamental tension, it merely raises the stakes. As we navigate this transition, our goal shouldn't be to make AI assessments indistinguishable from human ones. It should be to make them better: more accurate, more consistent, and more fair. The public's trust will follow results, not explanations.