AI Can Predict People: But Can It Truly Understand Them?

This week, I read a fascinating paper exploring the impact of Generative AI (GenAI) on consumers in the hospitality and tourism industry. While the paper focuses on a specific sector, it made me think about a broader question that applies well beyond: 

As AI becomes better at predicting human behavior, are we becoming better at understanding people, or simply better at predicting their next action?

As a psychologist now working in market research, I find that distinction fascinating, and honestly a little uncomfortable. AI can identify patterns, anticipate decisions, and process enormous amounts of data at a scale no human team could match.

But understanding why people make those decisions is a different challenge altogether, and it's the challenge our industry keeps quietly skipping over because prediction is faster to ship than understanding is to build.

What the paper set out to do and what they found out

This is a systematic review, meaning the authors didn't run one new study - they mapped the entire current landscape of GenAI research in hospitality and tourism at the consumer level, pulling together what the field has learned so far about why consumers adopt (or resist) these tools, and what effects that adoption has on their behavior.

Two things stood out to me as relevant well past this one industry.

1. First, most existing research has concentrated on GenAI's positive impact on operational efficiency, largely from the perspective of industry practitioners, service providers, and technology suppliers. In other words, the field has mostly studied this technology from the business's point of view: how much time it saves, how much it cuts costs, how much it scales - far less from the consumer's point of view.

2. Second, and more strikingly, critical concerns related to consumer experience: ethics, psychology, social dynamics, and security have received limited attention, despite posing significant risks to the successful adoption and sustainable development of GenAI technologies in the industry. The paper names those four areas directly. Not as a footnote, as a warning.

You can build a highly efficient system that consumers still don't trust, don't feel comfortable with, or quietly abandon, and the research base right now isn't equipped to tell you why, because it wasn't set up to ask.

Because of that, the authors call for research that identifies strategies to mitigate GenAI's risks while maximizing its benefits and argue future work should take a consumer-focused approach to understanding what actually motivates or blocks adoption.

As they put it directly, the field's attention to “ethics, psychology, social dynamics, and security” has lagged far behind its attention to efficiency. That's a paper about hospitality and tourism, essentially asking the same question I keep coming back to:

We've built excellent tools for measuring what GenAI does for operations. We haven't built the same rigor around what it does to the person on the other end of it.

Prediction vs. Understanding: where the line actually is

Prediction asks, “what will this person likely do next?” Understanding asks, “why does this person do what they do, and what would change if their circumstances did?” Prediction is built entirely on what's already happened: past clicks, past purchases, past behavior captured in a dataset.

Understanding is what lets you anticipate what hasn't happened yet: a shift in trust, a change in life stage, a need nobody typed into a form because there was no field for it. AI is remarkably good at the first. It has no access to the second unless a person builds that context in through the questions asked, the segments defined, the follow-up conversation that catches what the dashboard missed.

This is exactly the gap the paper points to when it separates “operational efficiency” research from “consumer experience” research.

Efficiency is a prediction problem: did the system do the task faster, cheaper, at scale?

Experience is an understanding problem: did the person feel respected, safe, and genuinely helped? A model can ace the first metric and fail the second completely, and you often won't see it in the data until adoption stalls or trust erodes, by which point it's expensive to fix.

Human behavior is shaped by emotions, experiences, context, relationships, and trust: factors that aren't always visible in the data. AI can recognize patterns and generate insights, but it doesn't inherently understand the people behind those patterns. It can tell you that something is happening well before most humans would notice. It still can't tell you why, and increasingly, that's the more valuable question, and, per this review, the one our field has spent the least time answering.

I see this gap in something as ordinary as my own workday:

In my own work, I speak both Spanish and English, but there are moments: writing outreach emails, phrasing something delicately for a prospect, where I lean on AI to help me get the English right. And it genuinely helps - it predicts natural phrasing, catches grammar, suggests a register that reads well. What it doesn't know is that a specific contact prefers a warmer opening because of how a past conversation went, or that a phrase that sounds perfectly direct in one business culture can land as too blunt in another.

AI can get the words right. Knowing which words are right for that person, in that relationship, is still on me.

That's the whole distinction in miniature: prediction gets me fluent text; understanding is what makes the message actually land.

Why this matters for how we work

That's why the future isn't about choosing between AI and human expertise, it's about combining both, deliberately, rather than assuming one will eventually absorb the other.

Generative AI is an incredibly powerful tool, but its value depends entirely on the expertise that guides it. AI doesn't develop judgment or understand context on its own - it learns from the data, prompts, and direction provided by people. Without that guidance, it can identify patterns, but it cannot fully understand the motivations, relationships, and complexities that influence human behavior.

This is where experts continue to play a critical role, and I don't think that role shrinks as the technology improves - if anything it gets more important. They ask better questions, challenge assumptions, interpret findings, and provide the context that allows AI to generate meaningful insights rather than simply more information. A model without a good question behind it just produces confident noise faster.

At 4xi, we believe AI enhances human expertise rather than replaces it. We combine AI-powered insights with deep expertise in workplace, customer, and employee experience to help organizations make informed decisions, facilitate meaningful conversations, and design solutions that technology alone cannot deliver.

The tools change every year. The discipline of asking why doesn't.

But it's not enough to combine both quietly and hope people notice. Communicating that combination - being explicit that the insight in front of a client is AI-accelerated and human-verified, not one or the other - is, to me, one of the most important things we can do right now.

Organizations are being sold two competing stories: that AI will replace expertise, or that AI barely matters next to human judgment. Both are wrong, and both are easier to sell than the truer, more nuanced answer. Saying plainly “this is what the model surfaced, and this is what our team added” builds a kind of trust that neither story alone can.

That's the part of this conversation I think we, at 4xi and across our industry, need to say out loud far more often than we do.

As AI continues to evolve, the greatest opportunity isn't simply to use it more, it's to combine its capabilities with the expertise, empathy, and judgment that only people can provide, and to be unmistakably clear that we're doing both.

The organizations that win this decade won't be the ones with the best predictions, or even the ones with the best people. They'll be the ones that let both do what they do best, together, and aren't shy about saying so. Because in the end, the question was never AI versus understanding, it's whether we're willing to let one earn the other.

If you wish to dive deeper into this topic or subject, I encourage you to check out: The Effects of Generative Artificial Intelligence on Consumers in Hospitality and Tourism: A Systematic Review and Future Research Directions, or feel free to email me with more of your thoughts.

---

Constanza Diaz

Research Associate

Email Constanza

---

4xi Global Consulting & Solutions is a team of talented leaders from both the client-side and service provider side, impacting the Human Experience (HX) for people at work, in education, rest, and at leisure.

We believe in a people-first, experience-led philosophy. Whether client, employee, or guest – their experience is the fundamental foundation of success.

WHAT WE DO

We work with corporations, service providers, and innovators:

• Strategic Advisory & Special Projects (SPx)

• Headquarters Fractional Support On-Demand

• Evolving Experiences© - Employee (EX) & Customer Experience (CX)

• Design4Life©: Environmental, Physical, and Experiential Design

• Global Amenities Strategy, Design & Operations

• TRUE NORTH©: Strategic Partnership & Growth

• Sustainability Simplified©: Supply Chain & Innovation

• Market Research Reports & Benchmarking

• Project Management Office

• Food Craft Culinary Consulting

• CREATE. : Graphic Design and Creative Services

• DATAxi: Data Ingestion and Visualization

• Explorers Innovation Directory: Gateway to Innovation

---

4xi is proud to be Chair of WORKTECH Academy for North America and a member of its Leadership Advisory Board. 4xi is a Global Ambassador for WORKTECH Academy.

San Francisco | New York | Philadelphia | Orlando | North Carolina | Seattle | Silicon Valley | Santiago | London | Tokyo