R085 - AI in Product Management, with Nacho Bassino

AI is no longer optional in product

What do you want us to know about you?

I'm in the fourth phase of my career. The first was as a software developer. Then I moved to product management, which I studied over 15 years ago, almost 18. Then I had a third phase as a consultant, helping companies with their product practice. And the fourth, the one that fulfilled me the most professionally, is sharing what I learned and continue to learn: teaching, coaching, writing, building community. It fulfills me a lot and I feel it adds value to the community.

How much do you think starting as a developer has helped you to become what you are today?

A lot. I always say that product managers can come from any discipline, and that brings strengths and weaknesses. I was never good at design and I still rely on good product designers. If you come from design, you might need to rely more on building relationships with the technical team. My technical background helped me identify my strength and leverage my weaknesses to be more complete.

You're at dLocal in a very interesting position. What do you do?

I am VP of Product for AI, Product Ops, and the CORE service. In AI, we divide it into two main groups. One: services to connect AI and facilitate other teams to consume AI tools (text recognition, generation with LLM, image creation) through a gateway. Two: enabling anyone at dLocal—regardless of their role—to use AI tools and create their own agents. We do a lot of agentic development. We aim to be a very lean company and from the AI team, we provide support to leverage that power.

In which parts of product work does it make sense to use AI and in which not?

I am very bullish on AI. I believe that in practically the entire product development cycle, end to end, not only can it be used: it should be used. Where it is more difficult to extract strong value is in what requires real empathy with other human beings: interviews, stakeholder management… There it can help in preparation, but the last mile we cover with our “meat brains”.

Let's stop thinking about product manager, developer, or product designer… and move towards AI builders.

Why are there people against it or not using it?

I see two sides. One: they don't have context or tools due to organizational barriers (procurement, doubts about whether it can be used, etc.). The other, more “real,” is that they still haven't understood the profound impact this will have. For years, discovery and agility were crucial because the cost of development was high. Soon, the cost of development will be practically zero, and that forces us to rethink everything. Some people resist: “I have to talk to the developer,” “I have to have X dailies”… maybe not.

The AI capabilities we have today are the worst we will have for the rest of our lives.

Totally. At Runroom, it has completely changed the way we develop.

At dLocal, we are working a lot on agentic development: an agent takes a Jira ticket, implements it, and opens a pull request. Today we are at 42% of pull requests accepted without human intervention. That changes how we conceive work. And I always say: the AI capabilities we have today are the worst we will have for the rest of our lives. That it can't do X today is no excuse not to adapt.

How do you use AI in research within the product cycle?

dLocal is very enterprise, so discovery is different from a B2C. We serve a few very large companies. There is a lot of deep conversations with clients to extract insights and a lot of regulatory research, APIs, how to do it within the company: a more internal discovery. We code a knowledge base to quickly respond to compliance and execution requirements. And in more user-oriented teams, the most powerful is pattern detection.

It used to be very costly to gather interviews, surveys, NPS, reviews… and in the end, we didn't gather them. Today you can have a robot 24/7 finding patterns among dispersed sources. That doesn't replace human judgment, but it changes the game: if instead of spending 10 hours navigating information, I spend 1, and dedicate 9 to thinking about implications, I become more strategic and can generate more impact.

Where do you not trust it?

In the part that requires real empathy: the last mile of interviews, double-checking, generating connection. And also in strategy when context is lacking: if I have to give it five years of company history, market, competitors, partnerships… it's difficult to package all that for AI to use, while a person who lives there has it in their head.

Synthetic users: I've seen people create a “GPT buyer persona” with four phrases and treat it as a real user. What do you think?

I don't recommend synthetic users for validation of new ideas. In the end, you reduce to a pattern with biases: biases of what was captured, of your current base, and it doesn't represent how a user would behave with a new product. What is it good for? For preparation: testing if an interview makes sense, detecting weak points, improving questions to follow a thread or hypothesis.

And another interesting use: stakeholders. If I have a skeptical CFO, I can “code” their profile and rehearse my roadmap pitch to anticipate tough questions. The real conversation remains human, but you arrive much better prepared.

And for ideation?

Very powerful for generating volume. Before, it depended on energy, time, and who you met with. Now you can ask for 100 ideas in five minutes. Then you apply your judgment, but it gives you an almost infinite resource.

What recommendations do you give to product managers about skills and tools?

The first: don't obsess over which course to take. The most important thing is to do. Think about your cycle: discovery, interview preparation, synthesis, data exploration… and ask yourself: “how do I do this with AI?”. You start working and the AI itself guides you. Patterns emerge there: mastering chat-type interfaces, building custom agents for repetitive tasks, creating a coach for prioritization, synthetic people for preparation, etc.

And then come flashy things like vibe coding, but the trick is that mindset.

If you are not working on replacing your job with AI, the person next to you will do it and will completely replace you

At what point in prototyping does AI add real value?

It changes the entire validation cycle. Before: interviews, definition, ideation, designer, prototype, developer for something functional… weeks. Today: I finish talking to the user, ask a model for ideas, pass them to a prototyping tool, come back and show something functional in minutes. It's absurd how much it accelerates.

Last year I wrote: “AI just made my 15 years of experience obsolete”. The way of working will be completely different. AI has the combined experience of many PMs and puts you levels above. The question is: until when? How long until a child with AI guidance does product management like us? And my philosophy is: if you are not working on replacing your job with AI, someone next to you will and will replace you. Better to be the one who goes a step ahead.

Have you changed product decisions thanks to insights generated by AI?

More than changing, it helps us form them. With deep research, you can investigate hundreds of sources and synthesize in minutes. It wasn't viable before. So we already use it from the start to build decisions.

What processes have clearly improved?

Research for strategy in markets like crypto, where a payments company needs to understand evolution and trends: what took weeks is now a strong starting point. Prototyping: it accelerates and strengthens what comes next because you validate earlier. And data analysis: a human looking at dashboards can't extract patterns like AI. You give it a historical big table and it can find much stronger patterns.

Any case where using AI has caused problems?

Yes. We tried to automate processes to gain speed and promised new business capabilities (for example, faster reconciliation and reports to clients). But it couldn't be done because the previous data pipeline wasn't resolved. Humans did it manually (copy-paste, merging sources), and without that context, AI couldn't. That got us into trouble because we had set aggressive timelines and months later we are still working on it.

By putting AI in the product, what new responsibilities do we assume as PMs?

Two things. One: with GenAI, the answers are not deterministic. We need to design interfaces prepared for variation. Two: evaluation, the famous evals. We used to write acceptance criteria; now we need to evaluate relevance and quality with “judges,” decide if a response falls within acceptable patterns. It's a new “corner,” with a different perspective.

We are working on observability and model economy (tokenomics), and it's a brutal field.

Totally. You have to optimize calls, choose models, observability (prompt, response, parameters). It changes the type of traceability. And also, small adjustments can give enormous systemic gains: in cost, precision, speed, depending on the point in the flow.

Are you concerned about the social impact of AI products?

I'm more concerned about the transition period. In the long term, I think it will be beneficial: many product decisions today are made by people without criteria, and AI can raise the average. But if a huge part of the work is automated in a short time, what happens to society? That worries me more: the economic and social impact of the change.

Even if we stopped AI development today, the impact would still be massive.

Yes. Much of what we haven't yet leveraged is not due to a lack of “intelligence,” but a lack of context: organizations without structured knowledge and applications ready for consumption. That will be resolved with engineering and focus. When it is resolved, a huge part of manual work will no longer be necessary.

How do you create impact?

I separate it into two. At dLocal, the company processes payments in emerging markets: it enables access to services using local means, even without a credit card. That has impact. And from my role, my greatest impact is enabling teams to make better products faster: providing strategic context, voice of the customer, ways of working.

The other leg is community: sharing what I've learned, a learning community in Product Direction, videos, articles, books. If you help others make better products, the end user receives better experiences.

In 10 years, what would you like to be said about the role of PMs in AI adoption?

That we were pioneers. AI is erasing roles: PM, developer, designer… and we are moving towards “AI builders” with a set of skills to build pipelines and products supported by AI. The PM is at the intersection of business, user, and technology, and AI allows us to go the extra mile: prototyping, coding, executing. I would like PMs to lead that transition.

Anything you would have liked to add?

I enjoyed the roller coaster: excitement and then social impact. I would like to end on a positive note: an incredibly fun stage is coming, which takes away repetitive work. We need to set the trend. If you doubt whether to adopt it: it's absolute. If you don't adopt it, the person next to you will and you'll be left out. The recommendation is to practice and get into the subject.

Product, AI, and responsibility: final reflection

The conversation with Nacho leaves a clear idea: AI can be a brutal accelerator. The product provides speed, alternatives, and synthesis, but it cannot replace judgment, context, and the responsibility of deciding which problem deserves to be solved and at what cost.

We also take away an important reminder. When we use AI to understand users, the risk is not just getting it wrong, it's impoverishing the view, losing nuances, making invisible those who don't fit the data, and confusing a well-written response with a human truth. And when we put AI inside the product, the role of the product manager becomes more demanding. I would say that delivering value is not enough, but we must also answer for the side effects, the incentives we create, and the social impact of what we put into the world.

A big hug and I look forward to seeing you in the real world.