As we barrel towards a world where AI can generate “good enough” designs for most startups using output mocks from LLMs, the question of “Ok, if an AI can do this, what’s my deliverable?” is one that’s been on everyone’s mind. And it’s a really good one. Designers have been judged, evaluated, and hired for decades based on the outcomes they produce. The process is important to shed light into how they got there and what decisions led to that output, but the final deliverable has historically been the go-to tentpole when judging a designer’s capabilities.
A common answer to this question is along the lines of taste. If an AI is handling the surface-level execution, a designer should then be focused on strategic thinking and imbuing their own taste into the product through their understanding of what’s desirable to users so that the product feels intentional and personalized. Taste is an important component to building opinionated products and it’s going to become incredibly important in this coming era. So how do we build it?
Taste is composed of many things: it’s a combination of your aesthetic sensibilities as a designer, your personal opinions about things that you think are well-designed, and your judgement about what differentiates the good from the bad. Most importantly, it’s your conviction about what needs to be true based on your lived experience of using and interacting with many, many products.
“Having great taste isn’t just knowing that the food is good, it’s knowing whether or not it needs a little more salt” — Emily Campbell, The Shape of AI
The only way to truly know whether a dish needs more salt is by having sampled hundreds of dishes with dozens of combinations of ingredients across the entire gamut of being undersalted, oversalted, or perfectly salted. You need to have developed your culinary palette enough to definitively say with conviction when a new dish needs more, less, or no salt. You simply can’t do it without this.
Designers need to be sampling more dishes all the time. Way more than they’re currently doing. They also need to be cooking their own dishes to truly understand why adding a specific amount of salt throws off the taste profile by diluting or messing with the other flavors. The only way to build conviction here is by dramatically increasing the sample size of dishes you’ve tried and rigorously evaluating their flavor. This process can take years or an entire lifetime to get to a point where you’ve honed this skill enough to know right versus wrong.
How, you might ask, can designers in their twenties and thirties imbue taste into a product if they haven’t even had a chance to cultivate it by sampling hundreds of dishes? The senior and more experienced designers definitely have a leg up here. They’ve worked in companies before the AI era where they made decisions about how flows and interactions should work. Those tenured enough to work through the dot-com era before the rise of A/B testing did this entirely based on gut feel and what simply felt right to use.
For the junior or early career designers starting out, there is a way. We’re in the beginning of an era where we’re seeing thousands of AI products hit the market that were entirely vibe coded and even entire startups launched that were built entirely with AI. Use them all. Play with them. Dissect them and reverse engineer them to figure out why the ones you like work well and what doesn’t work well about the ones you don’t like. Better yet, start building stuff yourself. Figure out why the AI makes certain decisions by default and see if there’s a way you can guide it with prompts to get to a better, more opinionated outcome. Find the hidden tips & tricks to improve the recipe across dozens of dishes so that you eventually have enough of a repertoire to write your own cookbook.
If you figure out how to crack prompting an AI to get to a design output that genuinely feels useful and desirable to users, you’ve started to develop your own personal taste. If you have a specific look or feel in mind, start making stuff. Try every single model and every single IDE to see what works. The best chefs have cooked in a variety of kitchens and utilized a large set of tools to get to a point where they know exactly which knife, whisk, pan, and cutting board to use before cooking the meal. They know the pros & cons of working with a gas stove over an induction stove and have a preference for one or the other depending on the dish they’re cooking.
If personal robots could cook every meal for you, you would still want a dash of whimsy or fusion in your dishes to balance out the stale, same-y flavor that you’ve gotten used to from the preset recipes programmed into the robot to guarantee the same level of precision and finesse in every dish. You will crave some wild flavors or unexpected taste combinations in your dishes. You would want to hire a human expert to add this variation to the robot’s programmed recipes, and maybe have the human expert cook entire dishes for special occasions. The next generation of designers are going to be these experts, and the ones that begin refining their taste sooner by sampling the widest range of dishes are the ones that will rise to the top.