Scaffolding, Not Substance: On the Responsible Use of AI in Creative Practice
James Richmond
Thursday 19th March 2026
This essay suggests some practical guidelines for the use of AI within the realm of creative practice. This can be summarised most succinctly as ‘use it for scaffolding, not content’. This distinction is more important than any particular model, prompt or workflow that you may choose to use. It cuts directly to the question of ‘who is doing the thinking?’ AI is remarkably convincing at generating content- sentences, ideas, narrative arcs that convincingly resemble finished work. It can imitate the tone of an author convincingly. What it cannot do is hold the concept of intention. It doesn’t care whether something is true or meaningful. It is, at least at the time of writing, simply very good at what is likely to come next.
That difference is not entirely philosophical in nature. It is also practical. If you allow AI to generate the substance of your work, you are not extending your practice, you are displacing it.
The Role of Scaffolding
In architecture, scaffolding is a temporary aid. It exists as a form of support, necessary for the construction of a building and then later removed. It is not a fundamental part of the structure itself.
When used properly AI occupies a similar role.
It can assist in exploring different approaches, to break through writer’s block, to test structural issues in trains of thought or reframe problems from different perspectives. When used in this manner AI is not a creator, but rather a catalyst. It is an accelerator pedal, rather than a GPS. It can accelerate the pace at which you can move but it does not determine the destination.
The danger here is when this scaffolding becomes indistinguishable from the building itself. When, as a creator, you lazily accept AI generated text, melodies or ideas as finished work, rather than as provisional material, the locus of authorship begins to blur. I argue that this is not because AI has become the author, but rather than the human has stepped back from the act of authorship.
AI is new enough that we are still all figuring out where our personal lines exist for what is acceptable for AI use. This is demarcated differently based on the use case. Advertising copy has less rigour than a Ph.D. for instance. What makes this paradigm shift particularly problematic is how seductive it is. The experience of using AI can feel effortless, even rewarding. A few lines of a prompt and all your thoughts are coalesced into a nuanced argument that says everything you wanted to say but couldn’t find the words or the time to articulate. There is a degree of satisfaction watching something so coherent emerge so quickly, especially in contrast to the more difficult, uncertain and significantly slower pace of, dare I say it ‘traditional’ creative work.
This ease has a cost. The more we accept what is given to you without interrogation, the more natural it begins to feel. The line between generating and decision making blurs. Are you shaping the work? Or curating it? Are you ideas shopping, rather than ideas generating. I am not coming from a place of blame here. This is not meant to shame anyone into turning away from AI tools. I think, on the whole, when used well and most importantly, thoughtfully, they can assist in creative work. When used poorly, without thought or an ethical consideration there is a not so subtle withdrawal from authorship under the guise of efficiency or progress.
Authorship and Resistance
Friction is a necessary part of the creative process. There are many moments when creating something new that you meet resistance. Perhaps it is a melody that feels awkward, or a sentence that doesn’t say what you intended, a narrative arc that resolves to a place that you did not intend. Working through, around or with that friction is where creative intention becomes form.
Resistance in a creative endeavour is not a flaw in the process, it is evidence of the artist in process.
AI is extremely good at removing friction. It is a large part of the appeal, it is why 92% of UK university students are using it. Why we see it literally everywhere, all the time. It can offer immediate completion and immediate coherence. Along with that you also can get untested resolution. You haven’t decided upon anything. It has been assembled through acceptance.
Voice and the Homogenisation of Weirdness
Much of the anxiety around AI in creative fields centres on voice. Individual style that is diluted or replaced with something generic. I see it everywhere, I am sure you do too. It is fantastic at pattern generation but I would argue that voice is not a pattern. It is the accumulation of decisions made under restriction, under constraint.
It is shaped by what you reject, what you consciously choose to keep and what you fail to articulate. That is an artist’s voice. It is a palette. And no artist has an endless palette. But AI does.
If AI is allowed to supply the content, those decisions are reduced. The work becomes smoother, more coherent, and often less distinct. Not because AI is inherently bland, but because it removes the necessity of choosing.
Rick Rubin said “The reason we go to certain artists, writers, or filmmakers is because of their point of view. AI doesn’t have a point of view.”
And he is right.
My concern with the use of AI in creative fields is the erasure of the voice of an artist. AI David Lynch would be a lot less weird than actual David Lynch. Or Salvador Dali. I want the weirdness, the uniqueness that made those artists who they are. I don’t want my own weirdness to be distilled down to a palatable version of who I am.
When used as scaffolding AI preserves the need for choice. It offers possibilities but it does not resolve them for you. You, the creator, still need to decide what the work actually is. And that is your artistic voice. And it matters. The longer we use these tools, the more it will matter.
The Prison of Efficiency.
We can view AI purely as a tool for productivity. Yes, it can produce more material in a day than many artists can output in a lifetime. That can be useful in some industries. In creative work, I am not so sure.
Efficiency is only a valuable commodity when it preserves or enhances meaning. I don’t need or want ten thousand Tom Waits songs. I can happily listen to ‘Hold On’ and several dozen others in his oeuvre that make my mind still, my heart yearn and my eyes weep. When the output is faster but thinner then something important has been traded away.
AI as scaffolding can improve efficiency without sacrificing depth. It can reduce the time you spend staring at a blank page but it doesn’t remove the requirement of thinking to produce something meaningful. It can fast forward you through the advertisements and trailers but doesn’t race you to the end credits.
AI as content makes the entire endeavour a pointless exercise. You are not engaged in the act of creation, you are showing the world what you did not create, taking ownership of aggregated voiceless knowledge. The work is completed more swiftly but it has less of you in it.
Practical Guidelines
“Use it for scaffolding, not content” is not a restriction. It is a discipline.
It means:
generating ideas, then rewriting them completely
using AI to outline structure, then building it yourself
prompting for variation, then selecting and refining with intent
treating every output as provisional, and not final
It also means being willing to discard what the AI produces, even when it appears to work. Perhaps, especially when it appears to work.
Because conflating ease with accuracy and fluency with meaning is ultimately to the heart of the artist’s process.
Towards Conscious Practice
AI is here to stay, I think we all know that. It shouldn’t be ignored. It is a powerful and capable piece of technology. The question is not whether to use it but rather how it is used.
If it is used as a substitute for creative thinking then it will gradually erode the capacity that an artist has that makes creative work meaningful and necessary.
When used as scaffolding it can expand that capacity, support experimentation and allow ideas to develop more freely. This difference may appear subtle but it seems important.
Maybe I’m being incredibly naive and the horse has already bolted? But I don’t want to live in a hellscape of AI generated slop where everything is made by everyone all at once with no individual voice to fire my imagination, to make me want to follow an artist’s development throughout their career and hopefully have a few people follow mine.
One approach is faster and smoother but increasingly interchangeable. The other leads to output that still carries the scars of resistance, intent and purpose.
In other words, something that still feels human.
Notes on this Essay.
I used ChatGPT to generate some chapter headings. Most of them were changed, for instance ‘Voice and the Homogenisation of Weirdness’ started as ‘Voice Is Not Generated’. The essay text was not AI generated.
