.jpg)
In 1962, Marshall McLuhan wrote that we shape our tools, and thereafter they shape us. At Patterns Summit Pittsburgh last week, ADP design systems lead Lou Manning offered a quieter version of the same idea, drawn from his first skydive: “arch your back and keep your head up.” When you're in the worst possible position you can be in — suspended in open air with the ground coming up — the form is the only thing that gets you through.
He wasn't talking about jumping out of planes. He was talking about AI as the new disruptor.
The room he was talking to had collectively shipped through Photoshop, Illustrator, Dreamweaver, Flash, Sketch, Figma — every tool-chain shift the field has ever called career-ending. Each one was absorbed into the work. AI is the next one, not a different category. The teams that handle it well will be the ones who already know the move.
The Pattern Beneath the Panic
The room had come in with a specific question: are we mature enough for AI? Design systems teams are being asked to validate, substantiate, and operationalize what leadership is telling them to do, often before the org is ready and before the tooling is approved. A 2025 Gartner survey found that 77% of designers expect generative AI to materially change their job within two years. Most of the press around that number frames it as alarm. Inside the room at Patterns Summit Pittsburgh — design systems leaders from PNC, ADP, Pittsburgh-rooted independents, and a cross-section of Knapsack customers and prospects — it didn't read that way. It read as familiar.
Brad Frost, author of Atomic Design, traced his own résumé through the field's last several "oh shit" moments — MS Paint, GeoCities, the early kerning obsessions, the Agile Manifesto reading group, Sketch, Figma, every layer of abstraction the practice has invented to keep up with itself. He marked 2008 as the last big inflection: the financial crisis, the iPhone a year out the gate, responsive web on the horizon, the assumption that you could design one fixed canvas suddenly worthless. The job rewrote itself then. It's rewriting itself now.
That's the pattern beneath the panic. Tool chains shift. Practices adapt. Practitioners ship.
Same Title, Different Profession
Anna Shade, design systems lead at PNC Financial Services, opened her segment with the sentiment that landed hardest: I went from an archeologist to an advocate for design.
She didn't take the design path to get there. Anna started in bank operations on a microfiche reader — push a button, the statement spits out, envelope it, mail it to the customer — and watched the technology underneath her change overnight. She moved from BSA to project manager wanting "digital things," with no designers on the team. ("So we were like, oh yeah, let's make our BRDs perfect.") She fell hard for systems thinking through DOORS, the requirements management tool — "the speed, you know, NASA uses that" — and learned that a real spec captures the happy path and the stress points in the same breath. A system that only describes the happy path isn't a system; it's a wish.
That obsession with rigorous requirements is what carried her into hiring designers, then leading them, then building one of the early design systems contributions at PNC, then being handed design ops, and then — "pretty much overnight" — being asked to lead design systems too. She didn't see herself as the center of the work until she was already there.
So when she says the job is now advocacy, she means it the way someone who came up through specs means it. Advocacy is getting engineering leaders, product heads, finance partners, and risk teams at the same table and making the case for why a coherent system is a business asset, not an aesthetic preference. The same role rewrites itself every tool-chain shift. AI is just the latest pressure shaping the rewrite. Anna's version of the rewrite, in 2026, is a system that can answer business questions out loud.
The Only Constant Is Change
Lou Manning, who leads the design system at ADP, told his version of the same story from the other side of the city. He moved to Pittsburgh from Queens at age ten. "I've always had to adapt" is the line he uses for himself, and the skydive quote is the surface of it.
He put himself through the Art Institute of Pittsburgh, took a first job helping an engineering firm with presentations, and started a career thread that has held through every tool the field has handed him: the spec is the artifact. He delivered a spec to Merrill Lynch in person during the first month flights resumed after 9/11, with Ground Zero to his left. He once printed an information architecture map 30 feet wide on the floor of a PNC Advisors conference room to make a point about depth. At Xerox, he ran a 16-person team that downsized into 8-on / 8-off and learned to white-label at scale through Photoshop macros — systematic thinking, before there was a name for the role. At ADP, he built an Illustrator template that became his team's design artifact way before Figma or Sketch.
"The only thing that's been constant for me has been change," he said. "I don't sweat the changes."
That posture is doing real work right now. Lou can't arbitrarily run Claude inside ADP yet — security review, procurement, the usual regulated-enterprise reasons. ADP is currently in conversations with Anthropic to change that. In the meantime, he built his own AI sandbox, runs experiments out in the open on LinkedIn, and is using it to prototype the next version of the ADP design system. Dogfooding, in his own words. Drinking his own champagne. The team watches him do it, and learns the move. Positive attitude, positive outcome — this is more than a motto. It's also a survival skill.
One-to-Many Goes One-to-Agent
Brad's headline argument was about the shape of the work itself: Design systems are one-to-many now. They need to breathe. They should flex. And yes — fracture them.
A traditional product team builds one experience for one surface. A design systems team builds one set of patterns for many — many products, many teams, many tools, and now many agents generating UI on top of those patterns. The unit of consumption has changed. AI is producing components faster than any single team can ratify them. Product orgs expect the DS to absorb every new surface — marketing sites, internal tools, embedded experiences, agent interfaces — without slowing down.
A rigid system snaps under that load. A system built to flex, breathe, and fracture on purpose survives contact with the real thing.
Brad arrived at the practitioner identity that work demands the personal way. His father was an accountant, his mother an art teacher — left brain and right brain came pre-installed at the dinner table. He started college as a music major, then realized that as a music industry minor he could take more recording-tech and media classes than as a major, since being a major meant four years of being stuffed in a basement practice room mastering an instrument. He switched fully into a Media Arts and Design program. That detour into making-things-on-screens became the third leg. The career grew out of the triangle. He framed it for the room as a permanent shape every design systems person has to hold: artist, accountant, maker. You're the artist because somebody has to care what it looks like. You're the accountant because somebody has to track which token is used where, which component is on what version, what the audit will find. You're the maker because at the end of the day a thing has to ship. Pretending you're only one of those is how a system fails.
Stability Is the Product
The deeper signal across the panel was about what a design system is for in this moment.
Anna kept circling back to one idea: uncertainty is the climate, so the design system has to be the stable thing. When the org is reorganizing, the regulatory environment is shifting, and the AI tooling is changing every quarter, the design system is the load-bearing wall. It's the contract — the spec — that keeps everything else aligned.
This is what makes design systems the right control plane for AI. AI without a source of truth drifts. Constrain a capable model with a real spec — components, tokens, governance, opinions about what's right and what's wrong — and the outputs reflect that structure. AI without that grounding is improvisation. AI with it is delivery.
What Knapsack Does
This is exactly what Knapsack is built for. We make the operational layer that turns a design system from a collection of files into a system AI can actually build against, with the governance intact.
That work shows up in three places:
- The MCP server, so Claude, Cursor, and any LLM-powered tool in the stack can read a team's design system as a first-class source of truth — not approximate it, not guess. The agent stops being a wildcard and starts being a colleague who's read the docs.
- The docs platform, where a team's components, tokens, decisions, and governance live in one place that's sharp enough for humans and structured enough for agents.
- The Intelligent Product Engine, just released — the piece that turns a static design system into a system AI can build product against, on the team's terms.
When AI tools are working from Knapsack, they're not improvising. They're working from the truth.
A Paired Approach
The technology side is solvable. The harder side is the people. Adoption follows trust, trust follows ownership, and ownership follows participation — a system people contributed to and watch shape outcomes is a system they'll defend.
That's why we've launched a paid AI Readiness Advisory practice for design systems teams. The teams calling us with "AI is here, what do we do with our design system" deserve more than a product pitch. The engagement helps DS teams get AI-ready in a way that works for their org, not ours — assessing the foundation, sharpening the spec, and building the feedback loops that turn skeptical teams into invested ones.
If you'd like to talk to us about it, reach out and we'll be in touch.
Riding This One Out
The practitioners in that room weren't worried. They've watched this movie before — different tools, different tooling vendor, same arc. The form is what gets you through.
We've been here before. We'll be here again. It isn't the worst. Just arch your back and keep your head up!