AI may lower the floor and raise the ceiling, but these are two very different things
The idea that AI can empower beginners and non-experts as well as turbocharge the abilities and talent of experts is nicely summarized in this expression: AI can lower the floor and raise the ceiling. In other words, it can lower the barrier to entry and expand the possibility space. There is a lot of merit to the concept. It does presuppose that the user has at least some basic ability to evaluate the output, which implies that it does not apply to every domain.
In the context of design (listen for example to @figma CEO @Dylan Field’s interview or @scott belsky’s perfect summary, links in the comments), it does make sense: after all, you are an expert in your ow taste and no one is better positioned to determine whether you like a design or not. If you are not a designer, your AI-based designs may be mediocre in the eyes of designers but perfect for you, with the added benefit that you don’t have to rely on an intermediary to explore the design space. This disintermediation lowers the barrier to entry: all you need to pay for is access to an AI and expend some time to play with it.
The end product may be perfectly satisfactory to you, the non-expert. But it could also serve as a starting point to engage an expert: “here is a mediocre example of what I want, can you make it better?”. In that case, the lowered floor enables a more effective communication between beginner and expert. In a completely different setting, imagine going to your doctor with an AI-generated list of possible diagnoses based on your symptoms and how you feel: sharing it with your doctor may take them in a direction they might not have previously considered, thereby expanding the possibility space for the expert.
As for raising the ceiling, the nature of the interaction between the expert and an AI is very different: the expert needs more control over the process, must be able to steer the AI precisely and finely, has to be able to tweak parameters to a level of detail that the non-expert would find overwhelming. By contrast, the non-expert’s interaction style with the AI is restricted, simple, mostly allowing the user to express what they are looking for, not minute details and arcane corrections. One reason the chatbot interface has become ubiquitous in AI is because it is so simple and familiar... but the degree of control is very low, which makes it tough to deal with complex situations.
In many AI business use cases, the two types of lift are often mixed or not distinguished: lowering the floor is about enablement (calculating, coding, planning, writing), raising the ceiling is about improvement (productivity, creativity). They require at least two different ways of thinking about how to deploy and insert into processes, at least two kinds of metrics, but also two very different approaches to interfaces.