On AI and the Work We Do

A letter to my design team at Filevine, written as AI swept into our industry and the layoff stories started piling up. The message is simple: I hear you, we're going to be fine, and you're still designers with a real role to play.

Something real is happening to our roles, and I won't pretend otherwise. AI is changing what a designer, a PM, and an engineer each do, and changing it fast.

You're feeling the pressure. Some of it comes from the tools, some from the noise online telling you that if you're not shipping production code by now, you're already behind.

I want to tell you where I stand, because I've changed my mind more than once getting here.

There's a popular argument that the roles are collapsing. That with AI a designer becomes an engineer, an engineer becomes a designer, a PM becomes all three, and the lines between us dissolve. I don't buy it. The people making that argument are chasing the wrong variable. They're optimizing for looking cutting-edge instead of for doing better work.

Here's what's actually true. The boundaries between us are getting more porous, and that's good. A designer who can stand up a working prototype in real code doesn't become an engineer, but they do change the conversation with engineering. Feasibility stops being something we learn about three sprints too late. Lean into that.

Work in the representation you can evaluate

But porous boundaries are not collapsed roles. The thing I want to protect, fiercely, is this: we should keep working in the representation we can evaluate.

Producing the artifact was never the hard part. Evaluating it is the job: knowing what's right and what isn't, and being able to reach in and fix it yourself. The moment you're handed a solution in a language you can't judge, you've lost the part of the work that matters most. That's the test for any tool. Not how fast it produces, but whether you can see the result and change it in a medium you actually think in.

Use the tool the job asks for

Right now that principle is split across more than one tool, and that's fine. AI is extraordinary for the early, exploratory stretch: generating directions, pressure-testing ideas, getting six versions in front of people fast.

Figma is where we go deep, where we get pixel-honest and hands-on. Use the right tool for the job in front of you. And within that, use the tool you're most fluent in. Job first, competence second. The phase of work picks the tool; your competence picks how you move inside it. What I don't want is anyone reaching for a tool because it looks impressive, or staying in one out of habit when the job calls for the other.

A word on handoff, because this is where people get confused. Right now, handoff still lives in Figma, and not because AI needs it to. Our engineers may tell you they don't much care what format they get. They paste it into Claude Code and let it do the heavy lifting. Figma owns the handoff for a different reason: it's the surface we all gather around. One link everyone can open. Comments in context. A single source of truth. It's a language we speak fluently and that still translates cleanly to everyone else. That won't be true forever. Figma replaced Sketch, Sketch replaced Photoshop.

Something will replace Figma. But not until it's genuinely better at those things, and we'll switch the day it is, not the day it's fashionable.

One caution on using AI to move between all of this: don't let "faster" become "frictionless." A lot of the value in a handoff is that it surfaces disagreement, where we catch the bad assumption before it ships. Keep the arguments. Just make them cheaper to have.

We each own our own ground

Here's the rule that keeps the blurring honest: we each own our own ground. Designers own design, engineers own code, PMs own direction. The boundaries can blend at the edges, and they should. But the further you push into someone else's territory, the more it belongs to them, not you. You can prototype in code; engineers still own what ships. A PM can sketch a flow; the design call is still yours.

Ownership follows who can truly evaluate the work, and the stakes rise with it. The closer a decision gets to something a client's case could turn on, the more fully it belongs to the person whose expertise it is. That isn't a limit on anyone's ambition. It's how we stay responsible to the people who depend on us.

Speed is not the point

Now the part I most want to drive home. AI is going to make us faster, and I want us to own that. Output will speed up because AI eats the tedium, and yes, speed will get measured. That's fair. But it is one measure, not the only one, and it is a result, not the point. If faster output ever becomes the single thing we're judged on, we've lost the plot, because that's not what the speed is for. The speed is what buys us the room to do the part that actually matters: more discovery, more research, more testing, more time making sure we're solving the right problem at all. I don't want more designs, faster. I want better designs, and the time to get them right.

We were never valuable for the artifacts we produce. AI makes that undeniable, because the artifact is now the cheap part. What's left is the work that was always the real work: figuring out what's right, and communicating it clearly enough that it gets built well.

That's what I want us to get better at. Not becoming someone else. Becoming more fully what we already are.

–Neil E