Drive north from Cartagena along the Caribbean coast and you pass gated condominiums, luxury shopping malls, universities and hospitals alongside much older Afro-Colombian fishing and farming communities. Punta Canoa is one of them.

To many people in the affluent communities now growing around them, villages like Punta Canoa are perceived much like American inner cities, except here they are on the outskirts.

The physical distance between these worlds can be a few hundred meters. The social and economic distance can be enormous.

This year, something happened that made me wonder whether AI might help bridge it. Through Albalab, a youth innovation program in Punta Canoa, a group of local young people had developed puntacanoa.co and its social media channels with the goal of connecting the village to the communities and markets growing around it.

A representative of an international real estate company in Cartagena came across the site, liked what she saw, and wanted something similar. So she asked to speak with “the developer.” The developers she was looking for were young people from the village. She might have passed them on the road many times without either side having any reason to enter the other's world.

We had been working with young people in Punta Canoa on digital production since 2020: graphic design, video, images, text and web design. They learned to make things that worked in the village, but one of our ambitions was to help them reach beyond the local market.

Generative AI has obviously changed what young people can produce. With limited formal training, they can now build websites and applications, produce images and video, design graphics, and work with text at a level that previously required much more specialized expertise.

But producing something is one thing. Producing something that people in another social world will value is another.

Marketing has long described this through the four Ps: product, price, place, and promotion. AI can assist with the production behind all four, but they also require judgment about what people want, what they will pay for, where and how they expect to find it, and how they expect it to be presented.

Much of that judgment is not technical. We learn what a serious presentation looks like, how an email to a potential client should sound, which details make a product appear professional, and what might make a customer hesitate. We learn what looks credible, appropriate, attractive, or simply wrong.

We might call some of this taste.

Eventually, we stop experiencing it as knowledge at all. It simply becomes part of how we see the world. AI can help make the product. Can it also help the producer see it through the eyes of the people it is meant for? Much of professional and social competence is not just knowing what to do, but knowing what to notice. If beauty is in the eye of the beholder, competence may sometimes depend on knowing what the beholder sees.

The anthropologist Mary Douglas famously used the idea of dirt as “matter out of place.” The difficult part is out of place. Before something can look wrong, we must have learned where it belongs and what order is supposed to look like.

The same applies to maintenance. We do not only learn how to repair things. We learn which things count as broken. Once we have learned this well enough, it feels as if we are simply noticing reality.

We have seen the same phenomenon at a small tourism business in Colombia, where Albalab youth have been developing FincaSmart 360, an AI-assisted property management system.

The caretaker is good at fixing things. Give him a broken pump, a leaking pipe, or a damaged door, and he will usually find a way to repair it.

But prospective customers who come to inspect the property notice things he does not: paint cans beside an entrance, tools around the parking area, a plastic bag in the garden, materials stored where a visitor would not expect them.

Telling him to “keep the place tidy” may miss the point because he may genuinely not perceive some of these objects as disorder. To someone maintaining a property, a paint can may be useful material waiting for the next job, and keeping tools close to where they are used may be practical.

A prospective customer sees the same objects differently. The paint cans may signal unfinished work. Scattered tools may suggest poor organization.

They are seeing the same objects through different expectations, and the caretaker will notice things the customer may completely miss. But when the property is a tourism product, the customer's perspective has economic consequences.

The caretaker does not need to see the world like a tourist. He needs to know what the tourist will see.

Can AI help him borrow the tourist's eyes?

As part of FincaSmart, the caretaker regularly uploads photographs from different areas of the property. Until now, they have mainly served as documentation. Multimodal AI makes another use possible. We have begun testing an AI system that reviews the photographs from the perspective of a prospective customer.

In one recent test, the system reviewed a photograph of the entrance to the property. It noted that the area appeared generally clean and natural, but identified some objects and materials lying beside the road and commented on them from a “visitor's perspective”:

“A visitor might notice the objects and materials on the ground beside the road. They could give the impression that work is unfinished or that the area is not yet fully prepared to receive visitors.”

The system classified the issue as low priority and suggested removing or storing objects that did not need to remain there.

An explanation like this may do more than solve the immediate problem. It may help him recognize the perspective behind it and eventually anticipate what a tourist will notice. Even if he does not, he can borrow the tourist's eyes again tomorrow.

Expertise and socialization shape what we notice. Until recently, gaining access to another professional or social perspective usually required spending time with people who already saw the world that way.

AI may offer another route. We do not have to possess another person's perspective permanently to have access to it when it matters.

This is where the question becomes one of inequality.

Pierre Bourdieu used the term cultural capital for knowledge and dispositions we absorb from the social worlds around us.

A child raised around academics does not need a lesson called How academics speak. Much of the code comes with growing up.

We all acquire knowledge this way, and a fisherman in Punta Canoa may know the sea in ways a real estate executive never will. But some forms of cultural capital provide far greater access to universities, professions, and markets, including access the fisherman's children may need if they want to enter those worlds.

Some people therefore grow up knowing the codes through which they will later be judged. Others first encounter those codes when they are being judged by them.

AI cannot give someone a well-connected family, years at an elite school, the right accent, money, or a professional network. A chatbot cannot reproduce decades of socialization.

If AI can make some of the perspectives produced by that socialization available when they are needed, while writing an email, preparing for an interview or designing a product, it may provide some practical advantages without requiring someone to have grown up with those codes.

AI can make some of this tacit social knowledge available on demand. What once had to be inherited or absorbed over years can, at least in part, be queried.

This brings us back to the real estate company in Cartagena.

For years, our challenge had not simply been to help young people produce, but to help what they produced reach people beyond the village.

This time it did.

The product was valued in another professional world before the people who made it were known. The product opened a door between two social worlds that rarely meet.

One more example I should mention: this essay.

As I have been writing about a Colombian caretaker seeing a tourism property through a prospective customer's eyes, I have been using AI to do something similar myself.

I have asked how these paragraphs might look to you. Does describing the caretaker sound paternalistic? Am I presenting one group's conventions as superior? Will a reader outside Colombia understand the social distance? Can I describe the consequences of social exclusion honestly without making the people who suffer from it sound like the problem?

I may not have everything right, and I have tried to identify where I might be mistaken.

The caretaker and I are doing essentially the same thing. His relevant beholder is a tourist. Mine, at this moment, is you.

We are both borrowing another pair of eyes.

The way we see the world is shaped by where we were born, who raised us, where we went to school, and the communities and cultures we have been part of.

AI cannot change where we come from, but it may allow us to see beyond what those experiences have taught us.

The AI Collective is built by volunteers across 180+ chapters in 40 countries.

Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.