Marta Sanaú Hornero

Your story: how did you end up in product?

My first product job had no proper title. After studying International Business, I joined a venture builder, and at one of its Covid e-commerce ventures I ended up managing physical products and the teams around them. Nobody called it product management. I only understood that later.

I wanted to move toward design, so I did Ironhack’s UX/UI bootcamp. Someone from their careers team told me, “First business, then UX? We might have a harder time placing you.” I took that personally: in a month I sent 2,500+ CVs and did 25+ interviews. One of them, at North, changed the plan. I’d applied for user research, and the person interviewing me said, “I think you’d make a great PM.” He went on to reshape my path, and became my manager.

The rest is history. Well, almost. At North I worked as an embedded PM inside one of Spain’s main banks, building everyday banking for freelancers and small businesses, and learned Scrum inside out. Then, at a later fintech, I owned an investment product and sat between Operations, Accounting and Compliance, and learned that half the job is translating between teams that see the same problem differently.

Along the way, once an interviewer told me I lacked technical depth, I took it as homework: I sat next to my developers and asked what they were building and why. Machine learning was my missing piece, so I moved to an AI company, where I helped build the CARS suite for an automotive SaaS company. For the past five years, I’ve done all of this remotely, travelling the world.

How do you define product success?

Two signals, in this order: adoption and how easily people can explain the product.

Adoption is the obvious one: are customers really using it? And if they are, do they keep coming back? For Connect, the first thing I look at is month-over-month growth in conversations, because it tells me whether visitors actually engage. When it dips, I don’t assume the product broke: I ask whether there was a campaign or a website change. The numbers tell me what happened, and the people closest to the dealer tell me why. Then I check how many of those conversations end in a captured lead, because that will tell us whether it’s doing its job.

The second one is less common, but it’s how I check whether a product has a chance. My favorite products can be described in 5-10 simple words that anyone understands. If you need a paragraph, your teams will end up selling different concepts and value propositions; you’ll reach fewer users and adoption will be lower. Most of the time, the biggest effort is condensing all your wonderful features into the words that capture their essence: “I build AI conversational bots for the car industry.”

What's your decision-making philosophy?

Startups of different sizes and stages taught me that uncertainty is the default, not the exception, so I no longer wait for certainty. I ask the people closest to the problem the questions that matter, and when nobody has the answer, I write the assumption down alongside the data we’ve been observing over time, and share it. Written down, a hunch becomes something the team can challenge, and disagreement shows up early instead of in a post-mortem. That’s what lets us move fast without moving blind.

A recent example: whenever we launched a new LLM model, we couldn’t know the exact unit economics up front. So we went with an assumption, shared it, and revisited it once we had real usage.

And just as important is the culture around it. I want people to feel safe saying “I don’t know” or “I got that wrong”. If we get something wrong, that’s the cost of running a business: we learn, move on, and make a better estimate next time.

What's your biggest PM lesson or mistake?

You need to learn to say no. I can’t stress this enough, and most of my biggest career learnings come from it. And sometimes the no isn’t to the request itself, but to being the permanent answer.

For a while, I was the one doing account-level analysis and reporting. I had built the pricing and cost structure behind our products, so when the request came, I said yes: nobody had more context than me. But I knew from the start that I would become the bottleneck and that the solution couldn’t live in me. So I said yes to the work and no to it staying with me, and used that same context to help build, with the product team, a tool that equips Account Managers to do the analysis themselves.

What I took from it is the need for boundaries: being good at something doesn’t mean it should stay with me. Product teams tend to absorb the messy problems simply because we’re good at solving them. Now, whenever I can, I train someone closer to the problem, so the knowledge doesn’t stay siloed and they can solve it on their own.

Boundaries apply to the product too. A PM who can’t say no and stand by it ends up in a feature factory that offers too many things and nothing at the same time. Not every idea is a feature, and not every feature is worth building.

How do you manage stakeholders?

Always with a ray of sunshine! I want everyone in the room to feel comfortable speaking up, and that takes a subtle balance between the people who would happily take over a 60-minute meeting and the ones who stay quiet. I watch for who hasn’t spoken yet and make space for them. That’s where I put my energy.

When people disagree, I usually find it comes from unequal information. A recent example: one capability was being offered differently depending on the customer, and sales, account management and product each had a different picture of it. So I looked at the data to see how much of our base actually used it, estimated the real demand, and we sat down together to work out the right cost structure. The capability stayed, but it was now sold the same way for everyone.

If someone pulls rank, I listen and factor in their argument, but I’m the one who decides whether something is worth building and at what cost. And when a decision affects other teams, I put it in writing, including what we’ve deprioritized or sacrificed in favor of it, so everyone knows the trade-off, or at least nobody can say it wasn’t made clear.

How do you prioritize?

It depends on the stage of the company. In a startup still fighting to become profitable, I prioritize by the size of the business opportunity: we need revenue to live long enough before we start building fantasy features. In a profitable company that needs to grow, I prioritize by the value for customers: what keeps them coming back and increases adoption.

In CARS, the clearest “not now” was a big, exciting capability that sat on the table for months. I argued it wasn’t a priority until we had solidified our positioning and had real customer feedback, so we’d build from how they actually used the product. A feature we dreamed up on day one is a wish; one built on real usage pays for itself, and you can see it in the metrics. The same goes for POCs: I push for the minimum lovable and useful product, and anything fancy that doesn’t move adoption or the numbers is out of scope.

How do you do customer discovery and validation?

B2B discovery works very differently from B2C. In B2B, which is where Seez focuses, you rely on your customers, and the bigger they are, the more weight their opinion carries. Account managers are essential in bringing that feedback to product, but any intermediary filters what they hear, which makes it harder to see the underlying problem. So I ask about the same topic from different directions, and whenever I can, I read the real conversations people have with the product. It’s the closest I get to the end user, and it shows me what they actually say, not only what the ones who sign the contract tell me. I also value written feedback (emails, Slack messages, meeting transcripts) over spoken feedback, because memory distorts what was actually said.

Testing the idea is the part where it gets real, and the one that makes me most tense: there is a lot at stake (churn, revenue...), so getting it right matters. We release progressively: first with a couple of dealers for a couple of weeks while we watch how they use it, then we iterate or roll out to everyone. Account Management and QA are vital partners in that, and I try to keep them close from the start rather than handing them a finished product.

What's your favorite product you've shipped?

My pricing calculator, which sits on top of our whole suite of AI products.

Our products are very different from each other, and every competitor prices theirs in its own way, so there was no market standard to copy. What we could do was standardize our own logic, so that everything we sell follows the same rules. I spotted the problem because sales teams kept coming to me for quotations. We had tried to structure pricing differently, but it hadn’t reached everyone. Our Chief Revenue Officer said it should be as simple as using a calculator, so one day I said: “Then I’ll build the calculator.”

Together with leadership, I worked out the cost structure of each product, the levers for growth, the add-on structure and the margins, and adapted pricing to local purchasing power and currency. But what made it work was the people. I gathered requirements from the sales and account management teams who would use it, built their instincts for discounting and ROI into the logic, and then tested it with them on real quotations, checking whether the results matched what they would have priced and refining where they didn’t.

The goal: anyone in the company can quote our products correctly in different markets, with little room for second guesses, and large discount requests go through an approval step with sales leadership.

It also changed how I see product. Product builds the functionality, explains it and helps sales position it, but I had never been this deep in how it gets commercialized: packaging, pricing and discounting.

What product inspires you?

Hevy, a fitness app to log your workouts and the weights you lift. I found it last year through friends who were already using it, and the moment I opened it I understood where to find the exercises, how it worked and who it was built for. My first reaction was, “This is what it does, and it fills a gap I have,” because I wasn’t tracking how my lifts progressed. Hevy does one job and does it cleanly, so it never gets in the way of the workout. The advanced features take more navigating, but you don’t need them to get value from day one.

What I love most: the exercise search with clear instructions, the small animation when you hit a personal record, the history of your progress over time, and the ready-made routines you just open and follow.

As a PM, it’s a reminder that the best products feel obvious, and that “obvious” is usually the result of a lot of saying no.

How is AI changing product management?

What we’re all observing is that the technical part is getting cheaper every month. This year I’ve built automations for onboarding reconciliation and templated release notes, and vibe-coded internal tools faster than a sprint would have delivered them.

What AI still can’t do is the human side of the work. I notice when someone is having a bad day, when someone is tense, when someone is already halfway out the door, and that’s when I write to them just to talk. In a remote team there’s no corridor where these things surface, so I try to make the distance less impersonal, one conversation at a time. No tool will tell you that a “yes” in a meeting really means “I’m worried”.

Understanding how the technology works is the entry ticket. Critical thinking, saying no and understanding people are the edge, and they matter more the more of the production gets automated.

Which AI product do you use and respect?

Claude, and Claude Code in particular. I had coded very little in my professional career, and Claude Code changed that. It explains each step, what every command does and why, so I’ve gone from reading code to building my own tools and automations. It’s like having a hyper-personalized mentor focused entirely on you, and that level of detail is what has helped me learn faster.

Day to day, connectors, projects, skills and automations have taken the repetitive work off my plate.

The one area where I still look elsewhere is design, where I often turn to ChatGPT, and I’d love to see a closer integration between Claude and tools like Figma.

What's one thing about AI everyone gets wrong?

AI is not magic, and it’s not always right. At an industry conference in the US earlier this year, AI was the main character of the show, and many people still treat it as a genie or an Einstein. Two things get lost in that:

It’s only as good as what you feed it. If your website has little or poor content, your assistant will return little or poor content.
It makes mistakes. The internet is full of contradictory and fake information, and AI learns from it, so it can sound sure and still be wrong. No AI has been 100% accurate yet.

The model gets the headlines, but adoption depends on whether people trust it and know when to step in. That’s why the best AI products are designed around the human: guardrails, handover and clear limits, and it’s why our assistants answer only from the dealership’s approved information.

What's your unpopular product opinion?

Agile does not work for every team. My old self and my team would be shocked to read that coming from me.

I was raised in the era of Agile, Scrum and Kanban, and I learned Scrum inside out as an embedded PM at a major bank. I also learned the hard way that you can’t impose a methodology on a team. No two people work the same way, and sometimes no single methodology fits two teams within the same company.

My manager once told me: “Let them embrace chaos, let them realize the structure they’re missing, and then they’ll come back at you and ask you about it.” I had tried for months to implement proper Agile, but the team felt forced to do it. So I stepped back, and they started picking the bits that worked and dropping the ones that didn’t. That’s how we built our own way of working: not out of a book, but by understanding the team, the dynamics, and the work.

The result was a much more engaged team. We handed a lot of the project management over to each person’s self-management, so PMs didn’t have to be ticket-creating machines.

Which author, thinker, or influence shapes you?

I learn most from people outside the tech bubble: chefs, writers, entrepreneurs,... People who question the norm, break it down and build it up again from scratch.

Ferran Adrià and Máximo Cabrera are my references in culinary creativity, for how they take apart what’s established to build a new paradigm. Alessia Zecchini, a professional freediver, was stubborn enough to break multiple records and reach depths nobody had before her. She taught me that “impossible” is often a limit you set for yourself, and that you’re the one who has to be stubborn enough to defy it.

From writing, the lesson that stuck came from the Writer Unleashed podcast: cut ruthlessly. Don’t be precious about your pages or your characters, and be curious enough to let the story detour from your original idea. In product, that’s saying no. Your outcome isn’t that specific feature; it’s your way of working: putting the best result above your own ideas.

And for anyone starting out, I’ve kept a product library for years, with articles, courses, books and tools that shaped how I think (it’s featured on my LinkedIn). I’d also point people to Productbeats; the product publications on Medium; and Adriana Carvajal (@adri.zip), who’s been doing a phenomenal job creating educational content on career growth, professional development and using AI tools to boost productivity.

What makes a great product?

A great product is easy to understand and does fewer things really well. It’s the difference between a restaurant with 6 well-crafted dishes and one with a 4-page menu: I’ll take the short menu, because whatever they do must be amazing if they only focus on those six.

My test: can a new user open the product and, within seconds, know what it’s for, what the use case is and whether they are the target user? And behind that, someone had to say no, many times, to keep it that way. You can feel it when it’s there: a product built by people who cared about the details feels calm to use.

A great product isn’t the one with the most features; it’s the one where every feature earns its place.

What does your product solve?

Car dealerships lose customers in the gaps: a website visitor who never gets an answer, a lead nobody follows up fast enough, a service reminder that never goes out. The CARS suite is four AI conversational products, each covering a specific part of the customer lifecycle:

- Connect turns dealership website visitors into leads.
- Reach nurtures those leads and helps convert them into buyers.
- Serve turns service reminders into booked service appointments.
- Assist answers dealership staff questions across all their systems.

Each assistant answers from the dealership’s own approved information, so it stays accurate and on-brand. It’s built for car dealerships, from independents to large dealer groups, and for the customers and staff who talk to them.

What's the biggest misconception about it?

“But I can do the same with HubSpot.” Well, yes and no. A HubSpot flow can send automated messages to your customers, but it can’t hold the conversation once they reply. At that point you need your sales or support teams to be available, have the context and give the right answer without missing any lead. With Reach and Serve, the AI keeps the conversation going after the reply, and the team only steps in to handle the booking, make sure it goes smoothly and collect payment.

For Connect and Assist, people tend to picture one of two extremes: a clunky decision-tree bot that follows a script and picks from pre-written answers, or an uncapped ChatGPT that can say anything. It’s neither. It writes its answers in the moment, like generative AI, but it’s built for automotive and answers from the dealership’s perspective, using only the dealership’s own approved information. Connect is like having your best salesperson available to every visitor, and Assist like a colleague who knows all your systems.

What's the market opportunity?

The automotive industry runs on dozens of pieces of software that are often poorly connected. That’s why Pinewood’s acquisition of Seez in 2025 made sense: bring the Dealer Management Systems and AI together, so everything lives in one system.

Why now: conversational AI is finally good enough to hold real conversations with customers, and it’s now reaching voice, so AI can answer the phone as well as the chat. Dealerships are stretched thin and need to respond to every lead and service customer quickly.

Next come AI multi-agents that perform tasks themselves, freeing back-office teams to focus on personal relationships with customers while the AI does the tedious, bureaucratic work.

What would you tell a PM just starting out?

Train curiosity and discipline early. The rest comes along.

For me, curiosity meant sitting next to developers, joining sales pitches and asking the obvious question, because most of the time the obvious isn’t that obvious anymore. Keep getting out of your comfort zone too: try things outside your job description and join networking events.

Discipline is less glamorous: it’s keeping the small promises you make to yourself. For years I’ve spent 1h a week building my product library. For a while, the last 1.5h of every Friday went to learning more technical content. Now it’s 30 minutes every day writing. The habit changes, the appointment with yourself doesn’t. And it compounds: you get a little better every day, and over the years that adds up to growth you couldn’t have planned.

What do you do outside work that shapes how you think?

Cooking, diving and, lately, writing.

Cooking is my escape route to try out my creative side, embrace chaos and get used to failing (my partner is the only collateral damage). Diving is how I unwind from stress: calmly observing the underwater marine life, with nothing to fix and nobody to answer to. Writing helps me put my thoughts into words, improve the way I explain things and craft stories that compel an audience, which turns out to be a big part of a PM’s job.

How do you normally structure a work day?

- I start my day with a 30-minute catch-up: Slack threads, emails and a summary of everything that needs doing.
- Then I prioritize by urgency and what’s still open. I built my own triage tool on Slack that automates this and books the slots in my calendar.
- I spend a lot of the day in meetings, so focus time is whatever fits in between. I try to group the meetings before lunch and keep the afternoon for strategic work on my own: deep dives, building automations, playing around and vibe-coding.
- I always keep a block for lunch. It’s when I decompress, process the morning and recharge, and skipping it is the fastest route to burnout.