AI multiplies every decision, good and bad — and four design leaders argue on stage over what to hand to the machine and what to keep in your own hands.
What it feels like to build in 2026 | Config 2026 · Tammy
30 min total·Actually worth watching closely: ~6 min·2 must-watch clips
- 0:17 – 3:31Listen
Opening: the thirds hiding behind 72%
Opens with the room's numbers: 72% are using generative AI or have at least tried it, but designers split neatly into thirds — never touched it, dabbled a little, fundamentally changing how they work.
Adoption is far from uniform, and nobody can say whether the change is all for the better or all for the worse.
A purely spoken opening plus room data, no key visuals — fine to listen to on a commute.▶ Jump to 0:17Speaker · Tammy - 3:35 – 10:36Listen
Is AI changing design, or the production layer?
Each panelist answers which layer AI actually changes: the value grows the closer you get to production and delivery, while early visioning and direction still need your own hands.
Be hands-on early and build your own point of view first, then hand off to the tools — the order can't be reversed.
A pure discussion stretch with no visual dependency; works fine as a podcast.▶ Jump to 3:35 - 10:36 – 15:30Listen
Judgment gets multiplied
A revaluation of fundamentals and judgment: the new tools are a multiplying force that ramps every good and bad decision up exponentially. Also covers generating several interaction versions side by side with AI to speed up exploration.
Whether an element should exist has to live in your head, not in a markdown file.
Mostly spoken argument, dense with quotable lines — worth close listening, but you don't need the picture.▶ Jump to 10:36 - 15:30 – 17:39Watch
When output is this fast, storytelling is the steering wheel
When AI makes shipping fast enough to get out of hand, "why this direction" becomes what matters most: storytelling gives everyone the same language to move in the same direction, and counters the pull into individual silos.
Once speed is no longer scarce, the ability to articulate direction is.
At 952s a hand gesture on stage makes the organization's diverging paths visible — a moment worth actually watching.▶ Jump to 15:30Speaker · Tammy - 17:39 – 24:41Listen
Debate: what to automate, what to do by hand
The sharpest disagreement of the session: one side wants error states, design systems and putting UI together automated as production design; the other warns that stopping for the error state can turn up a hidden gem, and that craft is where differentiation comes from.
There's no standard answer for where ideation ends and production begins — but you have to draw that line for your own team.
No demos here; the value is entirely in the clash of arguments, so listening beats watching.▶ Jump to 17:39 - 24:41 – 29:50Listen
Conviction is a culture question, and the double diamond isn't dead
Closes on conviction: the best builders have strong opinions; conviction can't be handed over, only drawn out by questions and by the safe space an organization gives for exploring. The diverge-converge rhythm of the double diamond stays, with AI as an instrument that does each part faster.
Step one in practicing your convictions is to have a conviction; AI flattens the rhythm of the process but doesn't replace the process.
A purely opinion-driven close with no visual dependency — good for writing down one conviction of your own afterward.▶ Jump to 24:41