I will admit that my use of generative AI tools and platforms is limited. I’ve been a designer since the year 2000. Which means I’m possibly more sluggish than the design industry young guns in succumbing to the lure of AI and all its workload-easing temptations.
I still enjoy using ‘old school’ processes and methods. That said, I have used genAI tools in a number of recent projects. My go-to for image adjustment or concept ideation has been a mix of Adobe Firefly and ChatGPT. With both platforms, the process involves text prompting to generate the desired output. In fact, when generative AI first entered the mainstream creative conversation, text prompting quickly became one of its defining characteristics.
For most designers, this was a new way of working: describe what you want to create, generate an image, and refine the result through further prompting. The process required us to think about the task differently, learning how to communicate with a generative model rather than relying solely on established creative software.
Photoshop’s Generative Fill, introduced in Beta in May 2023, brought this approach directly into an application many designers were already using every day. It allowed users to add, remove, or extend image content using text prompts. This was an early example of integrating a generative AI tool into an existing design environment. But even so, it was still clearly a genAI tool. It relied on that most tell-tale of indicators; the text prompt.
Something different – and less clear
Photoshop’s new Relight Layer tool, added to its latest Beta edition, is different.
Instead of writing a prompt to describe the lighting you want, you manipulate familiar controls. Add a light source. Move it around the image. Adjust its colour or intensity. Refine the result. The interface feels closer to the conventional tools we’ve used for years than to the prompt-driven experience many of us associate with generative AI.
This seems to me like a very important development. This is melding the capabilities of generative AI with the familiar interaction experience of a long-standing design application. So, is this an indication of a move away from a world in which we consciously choose to use AI, and towards one in which we simply choose the tool that gets the job done?
Just part of the furniture
I’m not suggesting that AI is new to Photoshop. Adobe has incorporated machine learning into features such as selection tools and Content-Aware Fill for years. Nor is Relight the first AI-powered feature to use familiar controls.
What’s changing is how seamlessly the capabilities of generative AI are being integrated into established design applications. As that integration becomes less conspicuous, the decision to use generative AI may also become less deliberate.
Instead of switching between Photoshop and a separate AI platform, designers can increasingly access these capabilities within the environment in which they are already working. The technology becomes another part of the application, alongside layers, masks, adjustment tools, and filters.
It reduces friction, minimises the need to learn entirely new interfaces and allows us to apply new capabilities within workflows we already understand. It also potentially makes the technology accessible to people who might never have felt the need to explore a dedicated generative AI platform. But it also changes our relationship with the technology.
When we open a separate application such as Midjourney or Firefly, the context is clear. We’re entering a generative AI environment, and the way we work reflects that. We make decisions about prompts, variations, and outputs with an awareness that we’re using a particular kind of technology.
When a similar capability sits behind an ordinary-looking control in Photoshop, that distinction becomes less obvious. We might still know that AI is involved, of course. But it becomes easier to think of the feature simply in terms of what it does, rather than how it works. Which makes me wonder: what is the impact of genAI tools becoming less visible?
The cost of making AI invisible
1. The financial impact
As I mentioned at the start, my experience of working with generative AI tools is still developing, but one difference I’ve noticed is how the consumption of credits is presented to the user.
When using Firefly as a standalone tool, the relationship between generating content and consuming credits feels relatively explicit. You enter a generative workflow, make your request, and understand that the process operates within a credit allowance. From my experience, I can see immediately what I will be spending in credits when I change AI model, or up the resolution, or change the length of the video.
When a generative capability is presented as another editing tool, it’s easy to focus on the immediate task rather than the processing happening behind the interface. The action feels like part of the normal creative workflow, even though it consumes generative credits.
Relight Layer is an interesting example. At the time of writing, Adobe’s current Photoshop Beta announcement specifies a cost of 30 generative credits when you commit to the result. In my testing, at no point in the process was it flagged up that using the tool would cost credits. To be fair, Adobe does provide information about generative credit usage, but I needed to do a quick Google search to check.
The thing is, it’s not that these costs are hidden altogether, but that they are less prominent in the moment of use, particularly when the feature feels like an ordinary part of the application. And as AI becomes more deeply integrated into our everyday tools, I think it’s reasonable to ask how visible its resource consumption should be.
Should using a generative feature feel no different from applying a conventional adjustment? Or should the interface make the underlying process and its cost more apparent? If we become accustomed to using generative features without thinking about the resources they consume, we may also become less deliberate about when they’re appropriate.
2. The impact on intentionality
That lack of deliberate-ness in using genAI tools also has a different consequence; removing the intentionality of using generative AI. I have a number of reservations with the explosion of AI into our lives over the last few years, but I’m not entirely averse to using genAI when I think it’s genuinely beneficial in a project.
I believe that my use of it is very intentional, and the requirement of moving from one app to another in order to use it, or having the obviousness of text prompting within Photoshop as a flag that I am applying genAI, is useful in ensuring that my decision-making is clear and understood. Removing that friction presents a particular challenge, especially for designers who want to maintain a non-AI or minimal-AI practice. If generative capabilities become indistinguishable from conventional tools, it becomes harder to maintain a clear boundary around when, and why, we choose to use them.
At this point one could argue that it is incumbent on any designer to fully understand the apps and tools that they are using, and that if they are unaware of genAI underlying certain processes, then that’s on them. But as things develop, and more tools are applied to design apps in a fashion that is not immediately clear as being genAI, it makes that argument a little fuzzier.
What does this mean for designers?
I’m not suggesting that the answer is to resist the integration of AI into creative software. Making powerful capabilities more accessible is a worthwhile goal, and tools such as Relight Layer offer an interesting glimpse of how that might develop.
And I don’t think I’m saying that every AI-powered feature needs to announce itself with a prompt box or a separate interface. If the technology can solve a problem through controls that are intuitive to the user, that seems like a sensible evolution. (But a small tooltip somewhere along the workflow indicating the credit usage would be an easy addition that would help frame the implications of using it to the designer.)
As generative AI becomes a more routine part of our work, I suspect we’ll need to become more discerning about the tools we use, the results they produce, and the trade-offs involved. Understanding when a tool is useful, what it contributes to the work and what it costs, in time, credits or creative control, remains part of being a thoughtful, intentional designer.
When does AI stop feeling like AI?
Relight Layer is a relatively small development in the wider evolution of creative software. But I find it interesting because it points towards a future in which the distinction between traditional tools and generative AI tools may become increasingly difficult to draw. As a technology matures, we stop treating it as something separate and start naturally incorporating it into the way we work.
It seems inevitable to me that AI will continue to become less visible. If it allows us to achieve more through familiar, intuitive tools, that’s a positive development. But as the technology becomes more deeply embedded in our workflows, we need to retain an understanding of what it costs and why we’ve chosen to use it.
I think we can say that AI is starting to feel less like AI, and Relight Layer is a good example of that shift. If this change continues with future updates to design apps, are we able to accept them feeling like just another tool without losing the ability to make a conscious choice about using it?
