The fear isn’t replacement
We’ve seen the tokenmaxxing free-for-all, complete with literal leaderboards ranking who could light the most tokens on fire. Now the pendulum’s swung back: orgs are throttling capabilities and hunting for cheaper models. As organizations continue maturing how they’ll operationalize AI, many are also measuring AI use by their employees in performance reviews. I’ve seen it firsthand: a new AI section added to the review framework as part of a broader AI strategy.
Declarations like this, announced without thinking through the second- and third-order effects, make anxiety spike. And the anxiety isn’t mainly about being replaced. In Lenny Rachitsky’s 2026 survey of tech workers,11 Lenny Rachitsky and Noam Segal, “How tech workers are feeling in 2026: a workforce splitting in two,” Lenny’s Newsletter, July 7, 2026. 22% worried about losing their job to AI, a real number but not the top fear. More respondents worried about doing more work for the same pay (51%), an unsustainable pace (46%), and declining quality (41%). Those fears come from an undefined, ever-rising bar for what a designer told to “use AI” is supposed to deliver.
- 51%Doing more work for the same pay
- 46%An unsustainable pace
- 41%Declining quality
- 22%Losing their job to AI
Share of respondents naming each worry. Lenny Rachitsky and Noam Segal’s 2026 survey of tech workers.
It’s tempting to dismiss this environment the way my running coach does when I miss my intensity targets: stop whining and get more comfortable being uncomfortable. But discomfort isn’t the problem. The problem is a leadership gap. As of May 2026, only 25% of U.S. employees said their organization had communicated a clear plan for integrating AI.22 Gallup, “Global Indicator: Artificial Intelligence,” data as of May 2026. “Just use AI” isn’t a plan. So what is? And what does it mean for each person on the team?
The same survey found that founders are the happiest people in tech,33 Same survey: Lenny’s Newsletter, July 2026. and the likeliest reason is control over their own destiny. Designers don’t need to become founders to feel agency. They need leaders who define what good looks like, give them a say in how they get there, and show them a path forward. Here’s how I’ve tried to do that.
Define what good looks like
If “Use AI” is vague and ambiguous, let’s address that head on by defining AI usage across levels and situations.
First imagine the primary criteria you expect to measure a junior designer off of, from discovery & ideation through delivery and communication. How do you, as their manager or the leader of their organization, expect them to use AI in these themes?
Check out the illustrative tables below to see how designers from Junior to Staff and even management can clearly understand what’s expected of them when using AI. Also note how the themes shift as the person grows in their role or is interviewing for a more senior role.
| Theme | Junior | Mid | Senior | Staff | Manager |
|---|---|---|---|---|---|
| Production & delivery | 55% | 45% | 20% | not weighted | not weighted |
| Discovery & framing | 15% | 20% | 30% | 15% | not weighted |
| Collaboration & partnership | 30% | 35% | 30% | 20% | 30% |
| Judgment & ownership | not weighted | not weighted | 20% | 25% | 20% |
| Influence | not weighted | not weighted | not weighted | 25% | 15% |
| Coaching | not weighted | not weighted | not weighted | 15% | 20% |
| Resourcing | not weighted | not weighted | not weighted | not weighted | 15% |
Weights show what each level is measured on most. A blank doesn’t mean it stops mattering. Related themes are grouped. Production fades as judgment, influence, and ownership grow.
Junior designer
| Theme | What using AI looks like | Weight |
|---|---|---|
| Production & delivery | Builds screens and variants from the design system with AI, then checks accessibility and pattern fit. Drafts specs, edge cases, and annotations with AI, and verifies them against the build | 55% |
| Discovery & framing | Summarizes research notes and generates a range of concepts before narrowing, checking sources | 15% |
| Collaboration & partnership | Shares AI-assisted explorations early, labeled as drafts, and asks for critique. Uses AI to tighten writeups but explains decisions in their own words | 30% |
Mid-level designer
| Theme | What using AI looks like | Weight |
|---|---|---|
| Production & delivery | Prototypes interactive flows with AI to test ideas with users quickly. Generates states, edge cases, and content variants, and owns their accuracy | 45% |
| Discovery & framing | Synthesizes interviews and analytics to frame problems, and flags gaps in the evidence | 20% |
| Collaboration & partnership | Treats AI prototypes as shared hypotheses with product and engineering, not finished answers. Presents tradeoffs clearly, including where AI output was wrong | 35% |
Senior designer
| Theme | What using AI looks like | Weight |
|---|---|---|
| Production & delivery | Produces less; uses AI to explore and pressure-test alternatives fast, and reviews others’ AI-assisted work for quality and consistency | 20% |
| Discovery & framing | Frames the research questions AI helps answer, challenges thin synthesis, and analyzes usage and research data with AI, then turns it into a recommendation | 30% |
| Collaboration & partnership | Sets norms with product and engineering for when an AI prototype needs design before it ships. Shares AI workflows that work, and the ones that didn’t | 30% |
| Judgment & ownership | Decides which AI-generated work is good enough to ship, and can explain why | 20% |
Staff designer
| Theme | What using AI looks like | Weight |
|---|---|---|
| Discovery & framing | Frames complex or ambiguous new work before it is scoped, and spots where AI could remove friction in the product or process | 15% |
| Collaboration & partnership | Works with engineering on the design system and tooling so AI builds on the right foundations | 20% |
| Judgment & ownership | Owns the quality of a product area end to end, including AI-generated parts, and decides what is good enough to ship | 25% |
| Influence | Shapes how teams use AI through patterns and guidelines others adopt | 25% |
| Coaching | Mentors designers on where AI helps their craft and where it hurts it, and pilots approaches for others to adopt | 15% |
Manager
| Theme | What using AI looks like | Weight |
|---|---|---|
| Collaboration & partnership | Aligns with product and engineering leads on how the triad works and what to measure. Translates leadership’s AI goals into clear, level-specific expectations | 30% |
| Judgment & ownership | Sets review standards for AI-assisted work and holds the bar | 20% |
| Influence | Shapes how the wider organization approaches AI, bringing the team’s practices and results to peer leaders and executives | 15% |
| Coaching | Helps each designer find where AI helps their craft (and be aware where it doesn’t) | 20% |
| Resourcing | Plans capacity on net gains, not gross, so no one gets quietly overloaded | 15% |
This isn’t just for design, either. Partnering with your peer leadership teams in Product and Engineering and encouraging them to build similar matrices makes everyone feel connected to one standard. The goal posts don’t move, either within Design or across the product development organization.
Write the playbook together
Agency also applies to how teams will work together. There are no shortage of articles and research indicating disciplines are blurring, that skills matter more than titles, and that role clarity may be a vestige of a recently bygone era (and by bygone era I mean checks watch a few years ago at best).
Designers can’t throw up their hands and rue the day a PM produced a prototype in an afternoon that would have taken weeks in Figma. That day has come and gone. What reduces the anxiety isn’t pretending it didn’t happen. It’s agreeing, as a team, on how we engage, collaborate, and build together now that anyone can make something that looks real. In other words, a playbook.
Every prototype is a provocation, not a product.
Prototypes are cheap to make and expensive to get attached to. The real skill in AI prototyping isn’t generating the thing. It’s being intentional about what the prototype needs to prove, then spending your time on those details and nothing else. Polish everything and people assume it’s further along than it is. Skimp on the critical interactions and the prototype can’t resolve the ambiguity it was built to resolve, so the team never connects it to the why. That gap between what a prototype looks like and what it’s meant to answer is where the anxiety lives: the feeling of having slipped the dock and drifting toward the horizon, busy but unsure where you’re headed.
Everyone knows where prototypes live and who’s seen them.
AI prototyping can’t happen in secret. When prototypes live in someone’s DMs, designers learn about a direction after it’s already been decided, and nothing feeds anxiety like finding out late. So make prototypes shareable by default. Critique them in public, not in side conversations. Let people fork a prototype and build on it rather than starting over. None of this should add process. The tools can carry the load: ask your LLM to draft a short markdown file alongside each prototype covering who it’s for, what it’s meant to test, what changed along the way, and what you’d measure if it ever made it to production. Five minutes of context saves five meetings.
Five minutes of context saves five meetings.
Know where the gains are, and where they aren’t.
AI speeds up the early work dramatically: discovery, ideation, first drafts. Then the gains thin out in the last mile, in final design and handoff (sorry, not everyone is pushing to production just yet), where the details and edge cases live. Plan around the upstream gains and you’ll promise timelines the team can’t hit, which is exactly how designers end up doing more work for the same pay at an unsustainable pace. Plan for net, not gross.
Plan for net, not gross. Illustrative.
Show the way forward
Clear expectations and a shared playbook answer the question “what do you want from me today?” They don’t answer the one that keeps people up at night: “what happens to me next year?”
Start with an honest picture of where the team stands. I use a readiness model that looks at six systems: the design system, AI and tech infrastructure, workflow, how the triad holds itself accountable, track record to base future predictions on past results, and whether customers are ready for what we ship (I work in governance, risk, and compliance, which means our corporate users are pretty risk-averse). The rule that matters is that the weakest system sets the ceiling. The most AI-pilled designer can’t prompt their way past a design system that isn’t ready for AI, and they shouldn’t be judged as if they could.
Six readiness systems, illustrative. The faded part of each bar is strength the ceiling wastes.
Then make the career path map to the work. Look back at the tables: production and delivery shrink from 55% of a junior designer’s expectations to little or none by staff, while judgment, influence, and ownership climb. You’re laying out a path that reflects a progression, and it should show up in how you hire and promote. Your teams will also recognize what they have to do to get to the next level instead of believing promotions are arbitrary or based on who goes to happy hour with the boss. When designers see that the skills AI can’t replace are the ones that get rewarded, the future looks less like a cliff and more like a staircase heading up (though in this world it may be an MC Escher-like staircase).
The tools won’t be the difference
Here’s what I keep coming back to. Within a few years if not quarters, the tools will be roughly the same everywhere. Every team will have the same models, the same prototyping tools, the same agents. When the tools stop being the difference, the difference will be how intentionally an org leads and works: whether people know what’s expected of them, how they build together, and where they’re headed. The orgs that chase every new tool and feature will stay chaotic. The ones that invest in their people will compound.
That isn’t a soft conclusion. In Lenny’s survey,44 Same survey: Lenny’s Newsletter, July 2026. manager effectiveness was the strongest driver of burnout, stronger than role, company size, or how people felt about AI. The biggest lever to reduce your team’s anxiety isn’t another tool and it probably doesn’t burn tokens. It’s how you operate and show up as a leader.
We’re not done. My team is further along than we’d have been without this work, and some weeks it still feels like we’re making this up as we go along, or can find ourselves surprised and reacting off our back foot. “Use AI” means something now, and it also means we have a clear path forward and a map to navigate us on our journey.