Smaller teams want design leaders who make, again. Here’s how.

AI is shrinking teams we run and raising the bar what each person is expected to produce. That changes what a leader has to bring, starting with their own hands.

Bill Russell coached the Boston Celtics for three seasons while playing center, and won two championships doing both jobs. Player-coaches were common then because one salary cost less than two, according to Wikipedia’s Player-coach entry.

As the leagues got rich, the role faded.

Design leadership followed the same arc. A leader’s status was counted in headcount as teams grew huge. That’s no longer the case.

Plenty of design leaders are what Texans call big hat, no cattle: a vice president title, three layers of managers, and no understanding of the business that goes with nothing they’d made in years. That concept is gone.

In June 2025, Andy Jassy told Amazon employees to “get more done with scrappier teams” in Message from CEO Andy Jassy: Some thoughts on Generative AI. The player-coach is back because money tightened, tools let one designer do a small team’s work, and the metric now rewards output per person, not total headcount.

We will need fewer people doing some of the jobs that are being done today, and more people doing other types of jobs. It’s hard to know exactly where this nets out over time, but in the next few years, we expect that this will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company.
 — Andy Jassy

I’ve been a player-coach my whole career, leading teams at Icertis, Evisort, and Workday while doing the work.

Doing the work never meant keeping the best parts for myself; designers on those teams often led the most important work, like a full redesign, while we learned new tools together, but it meant keeping my hands in the game.

Name the Hangover

Design teams grew fast for more than a decade, much of it on cheap money. In a 2024 interview, Peter Merholz — State of the Design Nation, Peter Merholz traced what he called design inflation to roughly 15 years of near-zero interest rates.

He described Capital One growing from about 90 designers in 2014 to 500 or 600 within a few years, and this played out in a lot of places. In that environment, team size was the business case. Headcount leadership worked because a leader who could hire, staff, and defend a roadmap looked successful whether or not anyone could trace design to revenue.

Merholz’s own diagnosis is that design leaders have been “suboptimal advocates for design in business contexts.”

I would agree.

The pandemic pushed hiring further. In Mark Zuckerberg’s Message to Meta Employees in November 2022, Mark Zuckerberg wrote that he’d expected the pandemic’s online surge to be a permanent acceleration, so he significantly increased investment.

It didn’t play out that way, and Meta let go of more than 11,000 people.

Much of what gets blamed on artificial intelligence today is that hangover.

Oxford Economics found about 55,000 job losses were attributed to artificial intelligence in the first 11 months of 2025, less than 5% of reported layoffs, according to Evidence of AI-driven job losses remains limited, says Oxford Economics report.

The cuts are the correction for over hiring, and artificial intelligence shapes what the smaller teams look like afterward.

Learn the Business

In 2022 the bill came due twice: interest rates rose, and a tax change took effect. Under Section 174, companies had to capitalize software development costs, salaries included, and deduct them over five years, or 15 for work done abroad at the same time that generative AI took hold. Only 10% came off in the first year, as Unique Impacts of IRC Sec. 174 Changes on the Software Industry lays out.

A product team’s payroll stopped being an capital expense and started behaving like equipment.

The July 2025 tax law restored immediate deductions for domestic development, which Gergely Orosz covered in The Pulse: Section 174 is reversed! Mostly, that is.

Orosz credits the end of near-zero rates more than the tax change for the hiring slowdown.

The reversal didn’t bring back cheap money, and artificial intelligence changed the math again. A hire is a multiyear commitment, while a model call is paid for per use, and a design decision can move its price.

A player-coach like myself reads a profit and loss statement, knows what the product earns per customer, and can say what a resourcing change is worth before anyone asks.

Shrink the Team

In September 2024, Jassy’s memo Message from CEO Andy Jassy: Strengthening our culture and teams asked Amazon’s senior leadership organizations to raise their ratio of individual contributors to managers by at least 15% by the end of the first quarter of 2025.

His reason: “Most decisions we make are two-way doors.”

A one-way door, like a pricing change, is hard to undo. A two-way door can be reversed next week. Headcount leadership made nearly every door look one-way, because each layer added a reviewer who would rather wait than be wrong.

On a small team, the person closest to the work makes the two-way calls the same day, and goes deeper on the one-way calls.

That frees the leader to focus on the few one-way doors, and a player-coach can tell which is which because they know what undoing the work would take. Layered teams rarely can.

Gusto’s The Manager Mass Exodus: How SMBs Are Flattening the Org Chart found the share of workers at small and midsize businesses who managed anyone fell from just under 25% in 2019 to about 15% by late 2024.

Google eliminated 35% of its managers overseeing teams of fewer than three people, with many of them staying on as individual contributors, according to Google Eliminates 35% Of Managers Overseeing Small Teams, citing CNBC’s August 2025 report.

Some people in that gap became what’s now called super individual contributors, senior people with a leader’s scope and no direct reports. Catt Small describes the design version in The Staff Designer as team leadership plus “in-the-pixels design work.”

I’m one of them and proud of it.

Gartner’s Gartner Unveils Top Predictions for IT Organizations and Users in 2025 and Beyond warned that mentoring paths can break when the middle layer goes, but we’ll work through it as we go.

Small teams need the leader working next to the newest designer, or they stop producing senior people.

Use the Tools

Designers have trailed developers at bringing artificial intelligence tools into their everyday work. Figma’s 2025 AI report: Perspectives from designers and developers found that 59% of developers used artificial intelligence for core tasks like code generation, while 31% of designers used it for core design work.

Design has often met platform shifts late, inheriting tools and workflows engineering already chose. Getting ahead of this one is a leader’s job, and the player-coach tries the tools before the team has to.

CEO’s are taking notice of this change.

Shopify’s chief executive, Tobi Lütke, made the expectation explicit in April 2025. His memo, Reflexive AI usage is now a baseline expectation at Shopify, required its use when prototyping new products, added artificial intelligence questions to performance and peer reviews.

It named who was covered: “Everyone means everyone.”

A rule that exempts leadership tells the team which work is beneath the people setting the rules.

I wrote a 50-skill reference for designers using Claude Code because I couldn’t advise a team on workflows I hadn’t broken myself, and most of what I know about where these tools fail came from building it.

A leader doesn’t need to be the team’s best at any one tool because they need enough hours in it to tell a real limit from a skill gap.

Redefine the Craft

In Figma’s State of the Designer 2026 survey of 906 designers, 58% defined craft as visual polish, ahead of thoughtful problem solving at 47% and clear, intuitive UX at 36%.

That sounds backwards, especially in our new age.

Polish is the part of the work models now produce fastest, making the pixel pushing designer a relic, which you would realize only if you’re a player coack.

Craft has moved past the pixels to the elegance of a problem well solved: the fewest steps, the clearest choice, the edge case nobody notices because someone already handled it.

Pixel pushers need not apply.

In artificial intelligence features, much of that solving happens in behavior. An artificial intelligence feature has thousands of possible outputs, and no designer will see most of them. Someone has to decide what a good answer looks like, write it down as examples and evaluation criteria, and keep checking real outputs against it.

Teams call those evaluations, and they’re becoming the specification for how a feature behaves. For a contract summary feature, that might mean 10 real agreements, the five facts a lawyer needs from each, and a rule that the summary never invents a renewal date.

That work looks less like a mockup and more like an editor’s markup.

Judgement still decides which of 40 generated layouts is worth keeping, and it’s hard to judge you haven’t exercised recently.

The player-coach sets the quality bar by meeting it in public, reviewing outputs in the same channel as everyone else and showing the edits.

Blur the Lines

The specialist organization chart assumed handoffs.

• A product manager wrote requirements
• A designer drew screens
• An engineer built them

In the 2000s, the web designer was often the front-end developer too, writing the markup and style sheets for the pages they drew, and the specialties split apart as teams grew in the 2010s.

Artificial intelligence shortens several of those steps at once, so the roles overlap again.

Figma’s 2026 AI report: Can AI help us collaborate better? found that the share of designers participating in development doubled in a year, from 21% to 41%, while developers doing design work rose from 44% to 60%.

In Why LinkedIn is turning PMs into AI-powered “full stack builders”, chief product officer Tomer Cohen explained that LinkedIn replaced its associate product manager program with an Associate Product Builder program. It teaches coding, design, and product skills together, and LinkedIn added a Full Stack Builder title and career ladder open to people from any function.

I’ve done my own informal polls, and I’ve see org charts of everything from product managers leading the charge to designers driving the product roadmap.

The lines are now blurred, forever

Specialists don’t disappear in a builder model. Someone still has to know accessibility law, query performance, or pricing strategy cold. What changes is who carries a feature from idea to launch.

My own job doesn’t respect the lines between design, product, and engineering. Most weeks I do a little (and sometimes a lot) of all three.

Job titles will change to match, as LinkedIn’s already have, and it’s up to design leaders to speed that up.

That means rewriting job descriptions, career ladders, and promotion criteria around which part of a feature someone can carry alone, before someone outside design writes them for us.

I’m the lead on a feature that is redefining the future of work for organizations, so I’m knee-deep in this transition and it’s happening for every role, not just us.

Ship Real Work

Gone is the era of design leaders and principals as professional meeting attenders, filling the room with words that aren’t accompanied by their own work.

Product management has always made requirements and strategy documents. Design is making too, and designers who don’t make shouldn’t be in the field. A working prototype now costs less than the deck describing it and settles arguments a deck can’t.

Dan Maccarone built an agency on that idea. Hard Candy Shell, the New York studio he co-founded in 2007, did product work for NBC, Bloomberg, and Disney by prototyping with clients, with, as the studio put it, “not a ton of presentations, meetings, and documents.”

Leaders have to show, not tell to have street cred now.

Paul Graham named a version of this for founders in Founder Mode in 2024, after a talk by Airbnb’s Brian Chesky. The standard advice to hire good people and give them room, Graham wrote, had gone badly for founder after founder, and he predicted skip-level involvement would become normal. He also warned that founders who can’t delegate will use the idea as an excuse.

Contributing to what ships means owning a real piece without taking the interesting problem away from a designer: the prompt behind a feature, a component the design system is missing, or a prototype engineering builds from.

The test is simple: if the team would be slower without the leader’s piece, it was real work.

Directors of large teams can start without a reorganization.

Pick one team close to a launch, cut one review layer for its two-way decisions, and spend a day a week working inside it as a contributor rather than a reviewer. After a quarter, compare its cycle time and quality with the teams around it, and let the numbers make the case for the next one.

Understand the System

This is where the principles of user experience really come in for smaller teams.

Melvin Conway observed in 1968 that organizations produce designs that copy their communication structures, which I cover on my site under Conway’s Law. The seams in a product tend to fall where people find it hardest to talk to each other.

Fred Brooks made the companion point in 1975, now known as Brooks’s law: communication channels multiply faster than people do.Six people share 15 possible pairs, while 40 people share 780.

Each pair is a place where a decision can stall or a handoff can drop, and the product will show it.

A smaller team has fewer seams, and the player-coach is the person who should decide where they go.

The system includes the data model under the screens, the components, the agents acting on both, and the teams who own each piece. Agents raise the stakes, because they act on whatever structure the product exposes, and a leader who knows only the screen layer will approve flows that break two layers down.

I learned this on an internal workflow application I helped build years ago. The screens tested well with users, and within a year the application was a nightmare to maintain, because nobody had owned how its pieces fit together. The design held up at the level we reviewed and failed at the levels we never looked at.

Understanding the system is what lets a player-coach decide where the team’s boundaries sit, which pieces an agent may touch, and which problems are worth a designer’s week.

Conclusion

Russell’s later career makes the case better than his championships do.

As the non-playing coach of the Seattle SuperSonics from 1973 to 1977, he took the team to its first playoffs but left with a 162–166 record, and his Sacramento stint was shorter and worse.

Without the playing, the coaching lost what he had in Boston.

Design has been here before. In the 2000s, after the dot-com bust, most design teams were small, and the people leading them still opened the files.

I designed the MySpace developer documentation site in that decade, when setting a team’s direction and drawing its pages were often the same job. The large in-house design organizations came later, during the cheap-money years, and they’re what’s shrinking now.

None of this ends the need for management.

Large organizations still need people whose job is hiring, planning, and protecting their teams, and a player-coach model doesn’t stretch to 40 direct reports. Headcount leadership measured a leader by how many people they could direct.

The player-coach gets measured by what the team ships, including the part they built. For design, that’s a return to the 2000s, with better tools and far fewer people waiting for a review.

Action items

• Price one feature. Work out what something your team shipped this year costs to run per customer and what it earns, with finance in the room.
• Ship something monthly. Own one real piece of production work each month, such as a component or a prompt, and put it through the same review as everyone else’s.
• Sort the doors. List this week’s pending decisions, mark the few that are hard to undo, and hand every two-way door back to whoever’s doing it.
• Rewrite one ladder. Redo one role’s job description and promotion criteria around what someone can take from idea to launch, and bring it to human resources before they ask.

Smaller teams want design leaders who make, again. Here’s how. was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.