Macarena Zañartu

Most AI never makes the jump. Not because the model is not good enough, but because not enough people trust it to ship it, to let it touch money, to stop checking its work. Closing that gap is the job.

At NotCo, using Giuseppe, a food development AI platform, was written into the scientists' KPIs. They still would not use it. The obvious read was usability, so we made the interface better. Adoption did not move. The real problem was bigger than the tool. It asked food scientists to change how they worked, to trust a model earlier in their process than their own judgment, and nothing about how it explained itself, or how it had been sold to them internally, gave them a reason to. Most of it was not on the screen. Building trust is not only a design job, it takes the teams that shape the model, the workflow, the incentives and the way a product explains itself, not just the one that owns the screen, and chasing it past the screen is why I have never stayed exclusively in the design lane.

Trust does not sit in one department, so I could not either. Design, strategy, engineering, AI. I move across all of them, because that is where the problem actually lives. But range like that does not arrive on its own. I left a role in banking to work with startups as an independent consultant, because I wanted to learn the ecosystem from the inside, across as many kinds of problem as I could get my hands on. That took a year. Then I wanted in properly, employed, close enough to shape product strategy and not just advise on it. That was the bet. It is why I can work at a collision now.

At NotCo I created Usability Nights to change who shaped the platform, putting the scientists and chefs who were meant to use it in the same room as the ML and software engineers who had built it, and what came out of those rooms changed both the interface and the optimization model underneath it. Every session asked some version of the same question, how much of the black box a person needs to see before they will act on what a model tells them, and since the answer changes with the work and the person, it was never something we could settle from a desk. We ran more than eight of them over three years, in Santiago and then in San Francisco, some with forty people in the room, and the format never repeated, because each one was built around whatever we most needed to answer next.

I started leading teams at BBVA in 2016, and since then I have led teams of up to thirteen designers and built a research function, some of it across the region, without the pitch ever changing. When someone is writing their own success instead of clearing deadlines, they do the best work of their life, and they stay honest with me while they do it. And leading is not only the people. It is holding the direction of the work when a room of engineers, product managers and executives all want it pointed somewhere different. Usually they already agree on the goal and only disagree on the route, so I build the session that gets them to one, rather than pulling rank to pick it.

The lessons started when I was four. I still play, and I sing while I do. Years later, a piano arrived in the house because my brother played it. I taught myself by hunting down the guitar chords I already knew, one at a time, until songs I loved came out of a different instrument.

I think that is still how I learn. I find the underlying structure, work out what I do not yet understand, and keep experimenting until something begins to work in a form it was not originally built for.

A pandereta, a harmonica, and a violin my daughter plays have joined the collection since then. I used to record, layer the parts, and invent the percussion. I want to get back to writing my own music.

I am still chasing the same thing I was at four: making something that makes another person feel something.

What I do does not have a clean title yet, and I am fine with that, because it usually means I am early to something. The roles I want are the ones where product, design, AI and engineering have to be shaped together, whether the title on the door says design or product.

The way I approach it is not new. I have been listening closely, learning the structure, and making things in my own way since I was four.