Renato Valdés-Olmos

Your company's head of design is whoever ranks highest and cares most. That is almost never the person with the title.

I have held the title four times and I misunderstood it for at least two of them. I thought I had been hired to be the head of design. I had been hired because somebody above me already was, and needed the caring turned into something the company could operate without them.

I have also coached a dozen or so heads of design since, some of them stepping into the seat for the first time and some running large organizations. The same misunderstanding turns up in almost every one of those conversations, which is most of the reason I am writing this down rather than continuing to explain it one at a time.

This is the sequel to an argument I have made before, that the function is subordinate to the business. That one is about accountability: design exists to move the company, and work you cannot defend in the company's own terms goes first. This one is about where design authority actually sits, which turns out not to be inside the design organization at all. And it has become the whole job in the last two years, for reasons that have nothing to do with org charts.

Authority does not live in the function

Watch what happens in a company where the founder has taste and exercises it. Design is good. It is good whether the design team is six people or sixty, whether it reports to product or to the CEO, whether the Head of Design is excellent or merely competent. It holds because the person who can end an argument cares where it lands.

Now watch a company where nobody senior cares. Every mechanism the profession has invented gets deployed. Critique rituals. A design system. Research readouts. A seat at the table, fought for and won. None of it holds. Work drifts to whatever the loudest stakeholder asked for. The design leader spends their tenure explaining why, to people who are politely uninterested.

Same title in both companies, and the craft is not the difference either. What separates them is whether anyone above the role cares, and that is the one thing the title has no control over.

Whoever holds the title is not the head of design. They were hired to build somebody else's caring into the organization, so it survives that person leaving the room.

What I got wrong

At Honor I had it right without knowing why. Our founders cared, visibly and expensively. My job was to turn that into decisions still standing in year five. Clearest of them is the landline. Engaging with the product had to work as well over a landline as over a smartphone. It sounds like an accessibility footnote. It was the most constraining decision in the company's history. It survived because it was a rule, not a preference. Nobody had to remember why to keep obeying it.

Later, running larger organizations, I inverted it. I started behaving as though the design bar was mine to hold personally, in reviews, by being the most senior person in the room with an opinion about quality. That works. It works right up until you are in a different meeting, and then it stops, and nobody notices for a quarter.

A standard that lives in your judgment is a standard that leaves when you do. Same is true of caring. If the only reason the work is good is that you are watching, you have not built anything. You have been performing the job rather than doing it.

Why this is now the entire job

Everything that used to justify the title is thinning at once.

The AI in Design Report puts numbers on it this year, from nine hundred designers in sixty countries. Roughly two thirds report doing more product and engineering work, and about forty percent say product managers and engineers are now doing design work. Roles are dissolving into each other. One finding I would not have predicted. Design leaders now ship code as much as their individual contributors, and sometimes more.

Meanwhile the thing a Head of Design was historically measured on is disappearing. That job was largely a hiring plan with a philosophy attached: grow the function, defend the headcount, install process to coordinate the people you added. Teams are getting smaller and more senior. Coordination is the cost that collapsed hardest. Whatever remains of the role, it is not that.

So the part I thought was the soft half of the job, building the caring into the company so it operates without a person, is now most of what is left. It was always the part that compounded. It is now also the part that is scarce.

What survives you

So what I would now defend is narrow and slightly unglamorous.

Find the person who actually cares most and outranks you, and work out what they care about specifically enough to write down. Not values. Rules with edges. Honor's landline. What the product refuses to require. Who it is for when there are two plausible answers.

Notice that every one of those is an opinion rather than a standard. The landline was contestable. Reasonable people argued against it, it cost us real speed, and nothing measurable defended it for years. That is what made it worth encoding. A quality bar can be inferred from the work around it. An opinion has to be written down or it does not survive the first person who disagrees and outranks whoever holds it. I have made that argument at length separately, because I think opinion rather than taste is what the profession actually has left.

Then encode it where it operates without attention. In a component rather than a checklist. In the definition of done. In what the model is allowed to generate and what it is not. David has the better half of his own argument here, and it is the one I would keep. When the output is cheap, the decision trail becomes the proof. Not the artifact, and not the taste that selected it. The record of what was considered, what was rejected, and why. Which is a record of opinions held, and of who held them. That is why it functions as proof at all, and why a trail with no name attached to any decision in it proves nothing. This is the argument I have made about seam work in a frozen organization, and it turns out to be the same argument in an accelerating one. Both are questions about what happens when nobody is watching.

And accept the implication, which is that a good design leader makes themselves progressively less necessary to the quality of the work, and no less necessary to the opinions in it. Somebody still has to be wrong in public. That is a strange thing to build a career on. It is also the only version of the job that leaves anything behind.

The part that should worry you

There is a structural failure running underneath all of this and almost nobody is fixing it.

Around three quarters of designers feel expectations rising on output, quality and speed. Under a third of leaders report making any formal change to evaluation, levelling or hiring. We changed what the job is and left the ladder where it was. People are being measured against a description of a role that no longer exists.

Sharper end of it is the junior rung. Entry-level design work was mostly production, which is exactly the part that has collapsed in cost. So the economic reason to hire a junior designer is going. It is going faster than anyone is replacing what that production work was quietly doing, which was teaching. That is how taste was transmitted. Not in a course. In two years of making the unglamorous version of somebody else's decision and being corrected.

That is the story our generation tells, and I notice it is an elegy. It describes the apprenticeship we had and assumes its absence leaves nothing. Look at what is actually being built and the picture is stranger than that.

People at the front of this are not waiting to be given somebody else's decision to execute badly. They are building their own instruments. A designer's toolstack went from three tools to seven in a year, and a growing share of them are self-made. One manager in that survey put it plainly: they had stopped paying for tools and started building them. Amelia Wattenberger is not designing screens for an agent, she is asking what an interface should be at all when the work is intent rather than lines. Brian Lovin ships Shiori, a bookmarking tool he built because he wanted it, and Gavin Potenza ships a Figma plugin that turns a pile of images into a moodboard. Small tools built around one person's way of working. None of it was made to be shown to anybody.

So the apprenticeship has not disappeared. It has moved. It used to be doing a senior's work badly and getting corrected. It is now building your own thing entirely and finding out, in public, at a speed no studio ever offered. Which is a real education, and it is faster, and I would not want to argue that the old one was better.

What I do not know is whether it produces the same judgment, and neither does anyone else. Whoever ran that survey worries about it in their own report: that outsourcing parts of the process may limit how younger designers develop judgment. Correction from somebody who has seen the failure before is not the same as correction from a build that did not work, and only one of those teaches you what you were not looking at.

It also quietly selects. Building your own thing alone rewards whoever already had the confidence, the time and the tolerance for public failure. That is the same mechanism I have described elsewhere as a description of comfort, arriving in a new form. The new route in is not open to everyone the way a junior seat was, whatever else is wrong with junior seats.

I do not have a solution I have tested. What I have is a way of choosing. The companies I would bet on are the ones planning the curriculum before they plan the headcount, and the ones where somebody senior is still in the room when the build fails. Almost nobody is doing either, including me.

I've led design and product at Honor, Lyft, Grammarly and Pitch, today I build at Noord, and I have coached around a dozen heads of design, incoming and senior. Figures come from the AI in Design Report 2026, published by Designer Fund with Foundation Capital in August 2026, from a survey of more than nine hundred designers in over sixty countries, and from Figma's State of the Designer 2026. David Hoang on the decision trail is the argument the middle of this rests on. Read one way, this assumes the caring is fixed and only its encoding is in question. In companies where nobody senior cares, a persuasive design leader sometimes creates the caring, and I have watched it happen. I think it is rarer than the profession believes and I would not plan a career on it.

© 2026 Renato Valdés-Olmos