Why 'Artificial Market Intelligence' is Neither Artificial nor Intelligent

History is littered with incumbents who, believing their expertise was a strategic moat, failed to evolve until it was too late. Encyclopædia Britannica, for instance, had centuries of verified, specialist knowledge but was outmanoeuvred by a more accessible, digital-first model.

This history of disruption in the knowledge economy is exactly why I’ve been so vigilant about the rise of Gen AI.

I used to be paranoid that a nimble, well-funded AI start-up could replicate what we do at Euromonitor International. My paranoia wasn’t unfounded. The technology is incredibly impressive, it’s quick to set up, and it generates answers that are astonishingly plausible. This created a sense of urgency, a feeling that if we don’t do it soon, someone else will, and fast.

But now, after being deep in the trenches building our own systems, I’m certain of the opposite: Only we can build an AI replica of what we do.

Bold statement? Yes. Bordering on arrogant? Potentially.

But my paranoia faded when I truly grasped a concept, powerfully articulated by AI researcher Kate Crawford, that is the key to our industry’s future:

“Artificial Intelligence is neither artificial nor intelligent.”

This isn’t just a clever phrase; for our industry, it’s the central, operational truth that reveals both the AI’s limitations and our unique, strategic advantage.

My recent articles on the “Plausibility Trap” and “AI Sycophancy” are direct summaries of the key challenges we’ve encountered building our new internal AI Research Agent. To be clear, this agent isn’t our client-facing Passport AI, which excels at enabling clients to search and synthesise our existing content. This new Research Agent is an internal tool to fundamentally upgrade the quality, consistency, and explainability of the very data we publish.

It’s during this development process that we’ve discovered how easy it is to produce numbers that look plausible yet crumble under scrutiny. This is because what we do is the art and science of market research. There is no single “true number” for much of what we publish. Our value is in a credible, explainable approach—leveraging the best sources and data assets available for each specific area of research. In data-rich markets, that might be scan data and granular national statistics. In data-poor markets, it’s good old-fashioned expert interviews and store checks. And to be clear, there’s a plethora of data assets and methodologies that sit in between those extremes.

AI is “Not Artificial” (It’s an Echo of Us)

This is why the “Human vs. Machine” debate misses the point. AI is “not artificial” because it’s a reflection of billions of human-generated documents—a statistical echo of our collective, generalist intelligence.

This is powerful for general-purpose tasks. But our business is built on “specialist” human intelligence. This “ocean of tribal knowledge” is our true strategic moat. An AI trained on the open web knows nothing about the hyper-niche, “on the ground” realities of global market research:

• It doesn’t know to account for cross-border illicit trade in African tobacco markets.
• It doesn’t know that direct selling sales figures are often missing from national retail statistics.
• It doesn’t understand the specific, local definitions of a “hypermarket” in Brazil vs. Japan.

Unless told…

This specialist intelligence is documented. We have always been rigorous about our sources, methodologies, and processes. But this knowledge doesn’t live in a clean, public database. It lives in the heads of our expert analysts, in the footnotes of decades-old reports, and in the nuanced logic of our proprietary taxonomies.

AI is “Not Intelligent” (It’s a Prediction Engine)

This is the second pillar, and it’s where the magic happens. AI is “not intelligent” because it is not reasoning. Even the newest models that imitate reasoning are, at their core, hyper-sophisticated prediction engines. They calculate the next most plausible word in a sentence.

But here is the key: what it predicts is heavily influenced by the context you provide.

This is exactly why our specialist human knowledge is so critical. The AI is a powerful generalist prediction engine with no driver. When we feed it our hyper-specialist context—our “tribal knowledge”—we are fundamentally changing its predictions.

It no longer just echoes the internet; it echoes our expertise. It doesn’t understand methodology, but it will follow the methodology we provide. It doesn’t know which conflicting source is reliable, but it will use the one we tell it to prioritise.

The Next Evolution: Encoding Our “Ocean”

This brings us to our core mission, the next evolution of our business. AI and automation are unquestionably the future. But our greatest challenge—and our competitors’ greatest barrier to entry—is that this “ocean of specialist logic” has, until now, not needed to live in super-structured, AI-accessible data assets.

For 50 years, much of our most valuable expertise has been handed down analyst-to-analyst through coaching, verbal training, and mentorship. Our project is difficult, but it is doable. This very difficulty is the moat.

The work of our “Insight Engineers” is to be the bridge. Their job is to translate this 50-year-old ocean of human expertise—from reports, from methodologies, and from our collective minds—into a structured, logical system. They are translating the “art” of our analysts’ nuanced decisions into the “science” of an automated workflow. We aren’t just automating a task; we’re encoding our entire research methodology. This is how we turn our most valuable ‘tribal knowledge’ from a mobile, individual asset into a permanent, scalable company asset.

This is why I’m no longer paranoid, but we can’t be complacent. The AI itself isn’t the advantage. The advantage is having the deep, proprietary, human expertise to build and train it.