Design & AI
My two-cents on keeping designers in the equation.
I’ve been 13 years in the design and marketing industries. Here’s what I’m learning about AI as it affects our work. Just note, I’m writing about design in the context of things that will be read and absorbed by humans; marketing for AEO/GEO is a whole other beast.
Materials generated with AI, such as Claude, by non-designers skew heavy on content and light on interesting layouts and design elements — to say the least. When a generated deck or flyer crosses my desk, I’ve experienced and observed an interesting phenomenon: creative lock-in.
Generated designs tend to instill the feeling that design decisions on content placement, typographic hierarchy, data visualization, etc. have already been made for us.
There’s a freedom in receiving plain content from which to create a design. Our whole job is taking human thought and laying it out, visualizing it, and translating it into something that catches the eyes of the audience and conveys its information with ease and interest. Not so for AI-generated vestiges of human thought.
Let me describe the difference between receiving a colleague’s thinking and an AI construction of their thoughts.
On this hand
Human instantiation of thought ranges from bare words on a slide to a lawless flyer layout, from a nicely formatted doc to a haphazard note, from a message with references to a quick call. Across the spectrum of human-to-human communication of ideas, intention is forefront and center regardless of brevity or thoroughness. We pick up on what someone aims to accomplish when we have direct access to their thinking.
On the other hand
AI generation of thought ranges from intricate to abstractive content, from information-dense flyers to indigestible data-rich slides. The through line is our non-design colleagues are explaining and sharing their thinking and intentions with machines and then relaying those outputs to us for conversion into human understandable artifacts. An AI-generated slide, flyer, prototype, etc. effectively functions as the completed version of their thinking and ideas — a machine has already decided how it all compiles and renders.
My opinion: AI can't really design; it can only predict and then produce legible (but flattened or averaged) design outputs based on its training data and the prompts of its user. Whereas we actually experience the efficacy of a design as we design it, evolve throughout the course of our lives and careers, and develop sensibilities and taste — as it stands, AI by contrast cannot.
What can be done
If you work in marketing, an agency or any workplace where colleagues are beginning to or already have begun using AI, there are some practical things you and your team can do to preserve design and brand integrity, and establish yourselves as the necessary curators and protectors of human thought and brand presence.
1. Guidelines and guardrails
Compose machine-readable guidelines/guardrails that
Cover brand or campaign style, fonts and typographic hierarchy, color palette and usage ratios, etc. (boil down your brand and style guides);
Cap the max amount of content allowed per screen or page;
Delineate voice and tone for your brand or per campaign;
Require an accompanying plain text or word doc version of the content with annotations of user intent.
2. Chats and projects
Ask your colleagues to share their ChatGPT or Claude chats with you. Establish it as a prerequisite when submitting a Wrike or Asana request. They’re instrumental context for understanding what your coworker’s aim is and how they arrived at the generated piece. A team project everyone has access to builds trust and accessibility into a workflow.
3. Consult the context
Check out the chat(s) that led to the generated piece. By following the train of thought you’ll be much better prepared to comprehend the content. Read the plain text, avoid absorbing the generated material until you have a firm grasp on what the information is and can parse it. If it’s too dense, try trimming or reworking it until it’s human readable/understandable.
4. Deconstruct generated designs (optional)
Import generated designs into Canva, open them in Word, edit them in Illustrator — whatever tool you can use to literally deconstruct the generated design, spend at least half an hour taking apart the layout, restructuring the flow of information, and for god’s sake tear through all of Claude’s content blocks.
5. Then design.
AI is here and its use is unlikely to recede. But that’s not a cause for alarm. For any business or organization, communication remains key; AI may generate in any language and construct layouts in any configuration, but it can’t humanly communicate or design — only we can.
Stay awesome!
~ Murphy