“Instead of doing one little pixel at a time, we should build the infrastructure that engineers use to create new features”: How Sierra designs with AI

Following the launch of our AI in Design 2026 report, we’ve been spending time with the design teams behind the products shaping tech, including Linear, Stripe, and Notion, to understand how AI is changing the way they design and build. We’re excited to start sharing those case studies with you.

First up: Sierra, the startup building AI agents for enterprise customer experiences.

Read the full case study

Co-founded by Bret Taylor and Clay Bavor, Sierra has quietly become one of the fastest-growing enterprise companies ever, crossing $200M in ARR in just over two years. But behind that hyper-growth is an incredible creative constraint: Sierra’s design team is just a handful of people supporting an engineering organization nearly twenty times its size.

Keeping pace requires a completely different mindset when AI acts as such a massive force multiplier for code. Sierra’s designers simply cannot personally touch every product decision anymore. Instead, they’ve had to completely rethink their leverage. As designers Wayne Fan and Nick Hiotis put it, the focus has shifted from polishing every feature themselves to building the systems, patterns, and components that allow hundreds of engineers to ship high-quality UI on their own.

At the same time, the team is highly protective of their craft. While AI has made building those technical foundations incredibly fast, they are intentionally keeping certain parts of their process slow, human, and unautomated.

Key takeaways from Sierra’s case study

1. Shrink the distance between idea and reality

The faster you can test an idea, the less time you waste debating hypotheticals. This is massive for Sierra because their product is incredibly data-dense. A layout might look flawless in Figma, only to completely break the moment it hits real-world data. Wayne laughs about finishing a Figma file that “looked sick,” only to see it in a working prototype and have to apologize to his engineer.

So, they shifted their workflow. Instead of treating Figma as the ultimate source of truth, they use it for quick, rough concepts, then immediately jump into Cursor to build a prototype with real data. Within hours, they know if an idea actually holds up.

Wayne calls the new rhythm: “Cook, prototype, be wrong, go back”... all before lunch.

2. Scale design by building systems, not screens

When AI first made it easy for designers to write production code, the team’s gut instinct was to just fix things themselves. They have a Slack channel where engineers post design nits, and Wayne would jump in with Claude to tidy them up. He’d point to a weird spacing issue, type “I don’t like this, fix it,” and it was done.

But he quickly realized that didn’t scale. “Instead of helping 50 engineers level up their craft, taste, and consistency, if I’m doing all the polish myself, I’m really just the 51st engineer,” Wayne says. He was treating the symptom, not the cause. “That’s not a good use of my time at all, even though it seems intoxicating and so satisfying to do.”

There was another side effect: the healthy friction that usually sharpens a designer’s judgment was disappearing. Instead of debating why a specific component or layout worked better, it became too easy to just accept the AI’s first suggestion and move on. “The muscle building on trying to dial into quality is kind of lost,” Wayne notes. “It’s being subcontracted out.”

That realization changed how the team approaches their role:

“If we can touch production code and the systems more easily with stuff like Claude or Cursor, what is a designer’s role in trying to empower at scale? Instead of doing one little pixel at a time, we should build the infrastructure that engineers use to create new features.”
- Wayne Fan, Product Designer at Sierra

Take their data tables. By last December, the product had accumulated around 18 different versions of the same pattern, each solving a slightly different problem. Instead of adding yet another review cycle, the team consolidated them into a flexible table component with sorting, search, pagination, and extended it per team as needs came up.

This idea of building shared foundations is a philosophy that goes beyond design at Sierra. The engineering team recently wrote about consolidating role-specific AI agents into a single company-wide agent for the same reason: the biggest gains come when improvements compound across the organization, rather than helping only one person or team.

Sierra’s designers scale their impact by looking across the product and spotting places where shared foundations could help teams build faster and more consistently. As Nick puts it: “We have to be extremely horizontal, aware of all the different services and features coming up, things being built. Recognizing patterns is a muscle we’ve had to become really good at.”

3. Protecting craft on purpose

All that said, some things are still very much worth making by hand.

Sierra runs two agents that work alongside each other: Explorer, the agent-optimizing agent, and Ghostwriter, the agent-building agent. When CEO Bret Taylor asked for a set of icons for them, Wayne and Nick built a small visual system - a compass rose as the anchor for anything AI in the product - and worked through sparkles and a Lego-minifig-style face before landing on a ghost for Ghostwriter.

Then, they obsessed. They spent two solid hours staring at a 16×16 grid, debating exactly how many squiggles a ghost needs to feel alive. Once the static asset was perfect, they animated it in Cursor so its eyes subtly drift around and its body morphs while it “thinks.”

Was it efficient? Absolutely not. But it brought back the joy of why people get into design in the first place.

“You look up and realize hours have passed,” Nick says, “and you’re like, wow, that felt so rewarding. I did not use my time the most efficiently. And how great was that?” He calls these “moments of joy,” the details that keep design from becoming a boring loop of typing a prompt and picking an output.

The ghost became a mascot - people put stickers of it on their laptops, and it gave the team something to rally around. “It could’ve been a pencil,” Nick says, “but do people want to rally around a pencil? Having a mascot, a sense of personality, is worth the heartache and the sweating of the pixels.”

As AI makes building easier, Sierra’s story is a reminder that the details worth obsessing over often aren’t the ones that save the most time. Spending an extra hour drawing something by hand or chasing a weird idea is often exactly what turns a purely functional product into one people actually love.

Explore the full case study

What we’re reading

• How tech workers are feeling in 2026. AI is creating a strange divide: some people feel amplified by it, while others feel less sure about where they fit. Lenny Rachitsky’s latest survey found that designers and researchers are among the groups feeling the most anxiety about AI’s impact.

• What happens when the terminal becomes your canvas? Pablo Stanley puts into words (and illustrations) how it feels to work this way, and why designing used to feel “more playful. Less optimized and robotic. Less shadcn and more color. Less inter/geist and more typefaces I had recently fell in love with…it had more of me in it.”

• Design has been too settled for too long. Andy Budd unpacks our AI in Design report and a shift we’re seeing across the industry: As designers take on more product and engineering work, the question becomes less “who owns design?” and more “how do teams make better decisions when everyone can build?”

• Paper went all-in at Config for their first real marketing push, featuring 3,000 branded coffee cans, ten wrapped taxis, two LED trucks circling Moscone, even custom claw machines, all designed in their own product and built in-house.

Open design roles

• Product Designer and Design Engineer at Paper (Remote)

• Design Engineer at Ambrook (SF, Denver, NYC, Remote)

• Motion Designer, Experiential at Stripe (SF, NYC, Seattle)

• Design Engineer at New Generation (SF)

• Senior Product Designer at Via (SF, NYC)

• Architectural Designer at Primer (Remote)

• Product Designer at DICE (London)

• Senior Product Design Manager at Gusto (SF, NYC)

• Product Designer at Tracksuit (Sydney)

• Head of Brand at Primer (SF)

Explore 1,000+ open roles at Designer Fund portfolio companies on our job board or forward this post to someone who’s looking.

And if you’d like to work with founders and companies where design is central to their success, join our talent network to get on our radar for future opportunities.

One of the narratives around AI and design in tech is that designers are getting squeezed out—that as software becomes easier for anyone to build, designers matter less.

The data we’ve uncovered in our second annual AI in Design report points to a more nuanced shift. Designers are owning more and shipping more. But they’ve also gone up a layer of abstraction to become builders. They’re not just producing deliverables—they’re spinning up new design tools to produce them. They’re inventing systems that help their companies harness more creative potential of non-designers and designers alike.

We believe this is the start of a foundational shift in how work gets done and how companies see their people. And it’s happening fast. In the year since our 2025 report, weekly AI usage for design tasks has jumped from 54% to 91%, and 75% use it daily. But the most significant changes have been qualitative. Last year was about early experimentation. Now, AI has spread into all parts of the design process—and brought designers into areas that most didn’t touch, like coding, prototyping, and accessing business data at scale to influence decisions.

We explore these themes and more in AI in Design 2026, created by Designer Fund and Foundation Capital. The report draws on a survey of more than 900 designers across 60+ countries, 25+ in-depth interviews, and case studies with the design teams at Anthropic, Framer, Linear, Notion, Shopify, Sierra, and Stripe. Each video goes inside their workflows, the tradeoffs they’re navigating, and how they’re operating differently now. If you’d like to be notified as they’re released, sign up here.

For design leaders wondering how others are running their teams, we want to shine a light and help you ask the right questions for the path ahead. For designers, we share what advanced builders are creating and what managers are looking for in candidates now.

Read the AI in Design 2026 report

And even as we surfaced patterns, we found that attitudes and behaviors vary widely by company and by individual. Depending on where you sit in the industry, these findings may feel surprising, validating, or somewhere in between. Our goal with this research is to bring more clarity to a conversation that’s evolving incredibly quickly.

Below are a few findings from the full report that we want to highlight. You can also read Fast Company’s coverage of the research.

Designers are becoming builders

Half of the designers we surveyed—across product and brand design, not just design engineers—said they’ve shipped AI-generated code to production.

We’re now moving from “should designers code” to “when and how should designers code?” AI tools are lowering the barrier for designers to prototype, build, and ship ideas themselves.

One of the more surprising findings was that leaders and managers are shipping code as much as (and sometimes more than!) individual contributors. Many described using AI tools to quickly prototype ideas, test workflows, or move projects forward without waiting on engineering bandwidth.

We also saw continued movement away from static deliverables. Prototypes are becoming a more central design output, and in some teams they’re replacing mockups altogether.

And more designers, especially at enterprise companies, are building their own custom tools. They’re inventing microtools to automate manual work, creating the design software they’ve always wished they had, and orchestrating agentic workflows that let non-designer teammates create on-brand work using prebuilt design system components.

The toolstack is expanding and fragmenting

The average designer in our survey now uses 7 off-the-shelf AI tools regularly, more than double last year’s average of 3. That doesn’t include the internal tools many teams are building themselves.

Rather than converging around a single “standard” workflow, design stacks are becoming more personalized and fluid. Teams are stitching together different models, prototyping tools, and coding environments depending on the task and how they like to work.

When we asked what challenges designers face when using AI in their workflow, the most common answer was “unreliable output quality.” Output quality is also the single most important factor in whether an AI tool becomes part of a designer’s regular workflow.

Roles are continuing to merge

65% of designers said they’re taking on more product or engineering responsibilities, while 40% said the reverse is also happening: PMs and engineers are contributing more to design work. As designers move closer to code and prototyping, traditional role boundaries are continuing to blur, with product development becoming more shared across disciplines.

But that increase in overlap doesn’t always translate to closer collaboration, both within design teams and across product orgs. Last year, only 5% of respondents said collaboration had decreased because of AI. This year, that number rose 4x to 20%.

Several designers described a shift toward more solo work: spending more time in prompting and in terminals, and less time in live back-and-forth with teammates. Work is moving faster, but some of the connective tissue of collaboration hasn’t kept pace.

What design leaders are hiring for

50% of the leaders we surveyed said they’re placing greater emphasis on AI fluency when hiring, followed closely by systems thinking and strategic skills. Compared to a few years ago, there’s less focus on narrow specialization and more demand for designers who can navigate ambiguity, connect disciplines, and shape direction.

Part of that shift comes from how AI is changing the design process. As tools become more integrated into product development, designers are increasingly expected to operate across the full arc of creation, moving from strategy to prototyping to execution with more fluidity. Several leaders described designers as “orchestrators”: people who can guide workflows and translate ideas into outcomes.

At the same time, stronger tools haven’t lowered the bar for quality. Only 5% of leaders surveyed said they’re placing less emphasis on execution quality. If anything, many argued that taste, judgment, and the ability to uphold a high standard matter even more now that AI makes it easier for anyone to generate “good enough” outputs.

Expectations are changing faster than company policy

73% of designers say they feel rising expectations around output, quality, and speed. Yet only 28% of leaders say their companies have made formal updates to evaluation, comp, or hiring. Just 8% have changed performance metrics, and only 4% have adjusted compensation structure.

Learning is still happening bottom-up: peer learning more than tripled year over year (from 24% to 80%), while taking recommendations from leadership dropped from 32% to 16%. Designers are figuring this out from each other vs from above.

The teams pulling ahead are the ones whose leaders accept that and build for it—creating conditions for tinkering, spotlighting employee-built tools in company forums, running hackathons, and protecting space to share workflows. This directly translates to employee satisfaction: Designers at companies with a culture of tinkering are 2x more likely to say they feel more creative and capable, even though they also face a higher bar.

Where this leaves designers

We believe it’s critically important that designers remain part of building software. Designers bring something unique to the table: user empathy, taste, an inherent sense of what to build and what not to build, delight, and the discipline to care about positive outcomes rather than just more software. As AI makes it easier for anyone to produce interfaces, that judgment becomes more valuable, not less.

Whether you’re a design leader making decisions on tooling and hiring, a designer thinking about how your craft should evolve, a founder trying to understand how AI is changing the way design gets done, or a student preparing for what’s next—we hope this report gives you something useful to work with.

We’d love to hear what you think.

We’re also publishing a series of case studies alongside it, with teams at the forefront of this shift: Anthropic, Framer, Linear, Notion, Shopify, Sierra, and Stripe. Each one looks at how a team is navigating these changes in practice. Sign up here to get them as they’re released.