Realworld
R091 - Designing Product Behavior, with María Pérez
Follow the Realworld podcast!
Listen to the episode on Spotify, Apple Podcast or YouTube Music.
Subscribe to our YouTube channel to not miss an episode
Building a product with artificial intelligence is now a matter of days. We all use the same models and build with the same pieces. And it shows. You open three new apps and they seem the same: white background, a touch of blue, and a chat box waiting for you to type. So I wonder: if we all have the same intelligence, what is left for us to design?
Today I sit down with María Pérez. María joins companies reinventing themselves with AI to lead product. She started as a designer in Barcelona, founded and sold her own company, and spent 12 years in New York leading product at Adobe, Trello, WeTransfer, or The Knot. She knows both sides very well: design and business decisions.
We talk about why chat has become too small for us, what it means to design how a product behaves, and that never-ending feeling when working with AI. If you design, build, or make decisions about digital product, this episode is for you.
What do you want us to know about you?
That I'm a girl born in Madrid, I've lived in London, New York, and Barcelona, I've spent many years in design and then moved to product, and I love cooking and talking to my dog.
I cook everything, three meals a day. I think cooking is super important, it's a mindset, and many comparisons could be drawn between how software is built and how you cook in your kitchen. I don't usually follow recipes. I find it a super creative exercise to see what you have in the fridge and put a dish in front of someone you care about. It's also very important how you source your food: where you shop, what you buy, what type of food, and the impact nutrition has on you. Cooking has many layers.
What did you take from each stage of your career?
I like to think I've always been in the right place at the right time. Until a few years ago, I thought it was luck, and I've realized it was actually a choice: I was very motivated by innovation, understanding where it could be generated, and learning what impact is and how to innovate. I started my design career when web pages were made with HTML, experienced the transformation to CSS, then moved to e-commerce and mobile apps, and later to other business models like SaaS (Software as a Service) and marketplaces.
What have I taken from those transitions? First, the richness of working in very varied environments, with very different industries that each had their reasons for being at the peak of innovation at the time. And second, a design taste, a way of designing that in my time already resembled product quite a bit. I've taken all that know-how to the product side.
Do you see much difference between how product is made in the United States and in Europe?
Yes, I find quite a difference. The first is cultural, obviously. I've worked with teams from the United States, Europe, Latin America, and India, and I think the cultural aspect is sometimes overlooked and is super important: how teams interact, how we talk to each other, even how feedback is given. In the United States, they always give you the sandwich: good, bad, good.
Second, I've encountered two very different speeds in the industry's maturation. In the United States, they've been making products for more years and have more mature software. What has been brought to Europe has also been productized in some way: processes, roles, and metrics have been brought over. What I've found in Europe is a lack of team maturation and, in many cases, a lack of talent. I like to say that the process has become the product: it seems that many companies are content with checking off Agile, having squads, and using Jira to feel like they're making a product. This has led to what I call zombie products, products a bit dead to the user, with excellent metrics.
I'm talking about team metrics and, above all, business metrics that don't represent what users experience. A very clear case is marketplaces. When you become a lead generator and your metric is the volume of leads you deliver to each vendor, your volume metrics can be very healthy and look good in front of the board and investors. At the same time, both vendors and end users experience something those metrics don't reflect: lack of quality, fragmented experiences in the UI, or products that generate no engagement or recurrence.
Metrics are a bit of an issue everywhere. Where I do see a difference is in culture and work process. In the United States, there's greater maturation of the product craft, the product mentality. Here there's a wannabe: we want to be product-driven and want to implement those methodologies. As a consequence of how talent has been formed, in a very short period, many products are led by the process, ask fewer questions, and generate less value for the end user.
How much of these companies' success is due to the product mentality?
I think it has more to do with the journey and maturation. A company like Adobe has 30 or 35 years of software, code, and business, with resources unimaginable in a European company. Adobe can afford to make mistakes, in quotes. They define themselves as a computer science company where innovation is at the center of everything. This was hard for me to understand when I joined because you're used to the business model, the leadership team, or even user feedback guiding you. At Adobe, everything is innovation: you have research teams constantly generating small technology capsules, and product teams consume that technology, transform it, and turn it into a meaningful product for users.
That's why the product cycle is very long. You have almost a whole year to generate a product that adds value. You have many resources: design committees to integrate your product into the entire Adobe ecosystem and Spectrum, their design system, and a marketing team that will take your product and grow it within the Adobe environment. All for that once-a-year delivery.
This doesn't necessarily make the company effective because you have 50 products, 50 lines. There's the cash cow of Photoshop, which generates all the income and sets the priority structure, and around it, you have all these satellite products, which are almost bets: bets to see which ones hit the mark or reach the market in time. A great example is XD. Adobe bet on competing with Figma voraciously and ended up trying to acquire Figma, which also didn't work out. The company's maturation is a factor, they have the methodology much more assimilated, and they can attract international talent with very attractive compensation. And I notice that the success rate, with all those resources, is not as fine-tuned as you would expect from a large company. It's part of that innovation, that walking through the unknown: walking where we don't know if there's an answer.
Adobe also has open lines to listen to its users. One thing I loved is that we had focus groups with very loyal users we talked to directly and launched test flights to. There are artists you work with for years who help you develop a tool like Draw. And then there's the other part: no one asked for a PDF format. There was a need to be able to read information without worrying about the format, and Adobe has that way of saying: come on, let's see how we solve this problem. Both parts balance each other. They know the creative industry very well, I think they've defined creativity and have owned that word for many years, and they also generate many research projects and investment in new technologies. At Adobe, we were already working with AI in 2016: Sensei was our AI brain, reading all the images, understanding workflows, and even giving us small doses of that intelligence to apply to our products. That's how we launched the font recognition tool from the camera.
I like to say that the process has become the product.
Did you join Adobe as a designer or already with a product role?
I joined with a product role. I had already made that transformation. I'm not very clear if it was voluntary or not, and I do think that at some point design stopped being interesting to me. There was a moment when design started to move within design systems, very necessary from a code and consistency point of view. Design systems also change the designer's profile: you have one person making the design system and another applying it to millions of product types. I remember we were making the 3D tool to generate shapes and textures, and the design system fell short: there wasn't, for example, a 3D viewer. It started to seem a bit boring to me, almost a prison or a template you always have to apply, versus the possibility of designing interfaces in a more open way.
With all due respect: obviously, I don't mean that the design system isn't necessary. We were already making design systems in Photoshop or Illustrator, with a little screen where the buttons for copy and paste were. What I'm saying is that it changes the role and the profile. Suddenly, product appears and takes away from design, in some way, the ability to think. Product starts to understand the user better and make many decisions, when I think the designer is the one with more information and empathy. So I moved to product, I switched sides.
How has your way of building product changed in the last year?
Almost everything changes. What changes the most is the immediacy with which I build product and how quickly you get context. What doesn't change is why you're building something, how you decide what to build, and what value and impact you generate.
How does it change? Since I have access to context about the seven or eight elements with which to evaluate a product or a company, I can generate context super fast. You can create synthetic users and talk to them, or access AI interviewers, which are wonderful and do that blending between qualitative and quantitative research, like Authentik. You can access metrics very quickly and, of course, prototyping and defining requirements change. Some people say PRDs, Product Requirements, are no longer needed. We still have them, only now they're not a closed document with questions: all the prompting you do with the model is your Product Requirement. And above all, you can have a prototype very quickly, which you not only share with colleagues or clients: it helps you check if your idea makes sense when you see it and iterate directly on a visual product.
It works very well for me because I come from design. I don't know if other PMs feel as comfortable in this environment.
I think in the next 12 months, the brand will live in the product's behavior.
After a year of FOMO, it seems something has settled. Do you recognize yourself there?
Totally. Recently I gave my husband access to Instinct. At first, he was a bit reluctant, and the day before yesterday he told me, literally: "I couldn't live without this. This makes me a better person." For all those things he left for later and forgot. He's not very digital and not organized at all in his head. Now he doesn't let so many people down. I feel super identified.
I think we're at a maturity moment for the tool. If you look at January or April this year, the models were just a chat: you could only generate a line of conversation, they spat out a lot of text, and you wrote a lot. It was exhausting, you had a lot to digest. The context wasn't as worked out or fed for as many months as now. Your comfort with the tool has also grown: you already know how the settings are, what the credits mean, how instructions or connectors are made. The software's own advancement and maturation have given us that learning curve, and now we're at a calmer moment. The next one is coming.
Why do all AI products look so similar?
I find this hilarious. Around December 2025 or January, when I started collaborating with the Authentik team, I did some research on AI companies' branding. I have about 25 screenshots of different AI tools on my computer, and they all have the same look and feel: a bit of white, rounded corners. It's logical, and I think there are three main reasons.
One: everyone uses the same tools and everything comes a bit templated. I think in three years we'll laugh at the first artifacts Claude gave us, all with that blue look and feel. There was already an artifact style that in a few years will be very retro. Two: time to market was very important for all these companies, you had to be the first in the race. Branding might not have been as important. What you want is to have a presence out there and even test, because many of these startups are looking for funding or haven't yet tested their business model. And three, and I'd like to think we'll have time to revisit this: I think we haven't known how to fit the brand. We haven't known how to translate what a brand is within these new products.
Where does the brand live in the product today?
Today, in October 2026, I'm not very clear. I think in the next 12 months, the brand will live in the product's behavior. If we've moved from designing 2D screens, the future of design will be decision-making and the behavior of both our agents and our interface. Therefore, the brand has to live in that behavior.
It sounds very geeky, and it's not so much. Until now, we've had 2D brands. We come from paper, then we moved to the offline-online layer and started reproducing offline brand values within the screen. Now I like to say we have 3D brands: it's not just the interface, it's also the tone of voice, when and how you speak, what kind of decisions you make, what happens when you make a mistake, how you set your security limits, where your models' ethics are, or where you get your data from. How brands engage with all those questions is what defines what a brand is.
Security, for example, will be super important. What decision-making capacity do you give your user about how much of their context they want to open to that product? How much do you need? Where does your data live? Are you open to other agents connecting with you, or are you closed and offer the complete experience? If you look at SaaS, or any digital product, I think something similar happened. Everything starts with a spearhead. Notion started as a text editor, and in the last six years since the pandemic, each SaaS has been extending and opening its capabilities to become something much more complete. TravelPerk starts with travel and becomes Perk because it already encompasses much more. Notion, which was a text editor, already gave you timelines and Kanban and positioned itself in Atlassian's territory, and now it creates agents. I imagine the same will happen now.
When all these products open up and offer you a bit of the same, what is the brand? It's how you perceive that product. It's what I hear: "I like Notion for X, Y, Z" or "I'm more of a Confluence person because of the interface, the colors, the features." And it's also the emotional attachment. If you've been using Notion for five, six, or seven years, you're not going to switch to another tool now. What did Notion do well? Build loyalty, create a style, a way of using it, like team comments. It goes a bit like that.
Not caring about the brand is a brand in itself.
You talk about behaviors. What behaviors are we talking about?
When you use Claude and ChatGPT, do you feel them different? I'm referring to the product behaviors, how it behaves in interaction with the user.
I go back to 2D. When you want to buy something on an e-commerce site, you know exactly where the photo is, where the description is, where the button is, and what the cart is like. This isn't random: it comes from years of research, Nielsen, heuristics, tests. All that is there for a reason. What will be the standard behaviors now? We have the chat, and it seems we're only going to talk within the chat. Even in those new behaviors, each product has its own personality. I ask ChatGPT something, and it doesn't answer the same as Claude, and Gemini does something else. They already have personalities, we don't talk about it, and we use them for different things.
I'd like to understand why we perceive these personalities and different uses in a type of software that has had zero onboarding. We've spent years thinking about onboarding: let's make a flow, let's put bubbles indicating where to click. There's an idea I love: how can it be that onboarding has always been so separate from the product and considered a separate part? You launch a product and then say: come on, now let's do the onboarding. Shouldn't onboarding also be the product? That's always caught my attention. And suddenly AI arrives, which seems like the most powerful tool we've been given in the last ten years, and there's no onboarding. If it's so easy to use and the model knows so much, why hasn't it guided me to use it better? Why hasn't it explained that I need to set up a context? Why hasn't it told me what a skill is?
I don't know if it's intentional because the business model is based on tokens, if it's due to lack of time and time to market, or if it's been a test and trial: let's leave it out there, listen to what users tell us during the first six months, and then choose which pieces to move. They must have heard something because from April until now, the capabilities the models have released are infinitely richer in how we relate to them, how they behave, and where they've chosen to develop.
And if the user doesn't care about the brand?
Not caring about the brand is already a brand in itself. I'd love for there to be brands whose goal is for the user to think about them as little as possible, use them as little as possible, and solve as much as possible. Where does the brand live? I think it will live in that behavior, and I don't know if there will be fewer visible brands for the user. A consequence of the digital boom is that I have to interact with too many digital products, and they all ask me for a login and password. There are many memes about that. I think we don't need half of those products. It's okay if some aren't consumed by the end user. Our agents will decide some of those brands and services for us, or we'll simply consume third-party brands without interacting with them. I could buy a plane ticket without going to the airline's website.
Why will I choose that company? For price, for loyalty, because it knows my preferences. Probably not for its interface. So, where does the brand reside? I think there will be fewer end-consumer brands and more brands made for agents, made for the agents themselves. And brands will have to elevate their value significantly. I'm talking about the software world, and also offline: they will have to develop their personalities much more, generate many more differentiating values, and create that behavioral attachment, that personality we still don't know how to describe very well and that makes us part of that tribe, in quotes.
Where do companies that manufacture products and not models have opportunities?
Anything that isn't a model is a good opportunity. The model is the brain, and you choose the right one for each moment: if you want to make an image, you choose Gemini. Around it, there's a whole layer of experience, connection or connectors, and context, which I find very important. Anything that facilitates my context will be a killer. Anything that helps my context be well-segmented, categorized, and secure is a good opportunity because it's who I am. Whoever contains my context, understands it, and categorizes it. It's a bit like what Obsidian does by generating brains at the work level, and we'll need it at the personal level too.
Anything related to personalizing your interface environment will also be very important. We've just seen that ChatGPT has released the dynamic interface. We've been talking about dynamic interfaces for months, and suddenly we see what it is: we can deliver information in a format that's not just text, still integrated within the ChatGPT window. I think this marks where we all expect our interfaces to go. If you mix agents, context, and interface, there's a world totally to explore.
The big word today is uncertainty. Knowing how to handle it and knowing how to choose within it is fundamental. Decision-making, for our teams and even ourselves to be able to make better decisions, is what can generate the most impact in the next 24 months.
The best way to know if you've finished is to answer the question of how you know if this problem is solved and how you measure it.
How do you know you've finished something when working with AI?
Super good question. You don't know, and you have to trust that it's enough. Prompting is made to keep asking. I wrote about this: it's like the "anything else?" at the market, infallible when you go to buy ham and end up taking cheese too. I think the models no longer pressure as much as a few months ago with the "I can prepare..." or the "do you want anything else?" The key is your own criteria: not entering an eternal question-answer loop and auditing what the model returns to you.
You have to audit. Your criteria are super important there: not taking everything as valid, contrasting it, reading it, and letting it rest. When you've been prompting non-stop for three hours, leave it, go for a walk, have a coffee, and the next day read it and think if all those wonderful hallucinations you've had are real and serve a purpose. You have to be aware, know when to stop, and have the criteria to make the decision.
A work of art is something much more subjective, related to the artist's satisfaction and emotional state at that moment. When we talk about a product with AI, the best way to know if you've finished is to answer the question: how do you know this problem is solved and how do you measure it? That should be your absolute guide. There's the decision: not building for the sake of building, knowing why you're building it and what you're solving.
Anything that facilitates my context will be a killer.
If you design a product that never lets you stop, whose responsibility is it?
If we take the example of social networks, the responsibility is obviously on whoever designs that product because it touches something very dopamine-related and how our brain works. And I'd like to think the responsibility is also ours. It's very easy to always place it on others. Minors don't buy a phone for themselves or spend hours in front of the screen alone. Adults also decide to spend many hours in front of the screen. I think we need to learn to take our share of responsibility. Another thing is if the product is designed in a misleading way or generates addiction, as has been proven.
With AI, the responsibility to stop is also ours. Do we have to waste water to ask Claude for anything? Sometimes I ask for something, and while it thinks, I could have gone in and changed it myself. You have to be a bit responsible. It's like choosing between the elevator and the stairs.
How do you create your impact?
I'd like to think the best impact I create is the one I have on people: sharing my knowledge, having fun with the people I work with, and spreading a bit of good vibes, ethics, and purpose in the work we do. I also create impact by empathizing a lot with users and always focusing on solving real problems. Many times you can focus on moving a metric that turns out to be the wrong one. You have to focus on the real problem. Everything else seems absolutely irrelevant to me in the era we're living in.
Conclusion
I take away an idea from the conversation with María: when intelligence is the same for everyone, the difference is back in design. In how a product behaves, in how it speaks to you, in how it accompanies you to where you wanted to go. That's where a company can continue to be itself.
And that brings a new responsibility. Designing a product's behavior is also deciding when it invites you to continue and when it lets you stop. Building something engaging is relatively easy. Building something that helps you finish, think, and decide calmly is much harder.
Perhaps that's the impact worth seeking: products that help us move forward and respect our time. A big hug, and I'll see you in the real world.
Let's talk
Tell us about your challenge, and we'll explore how to turn it into a growth opportunity together.