Graphic design, by its longest-running definition, is a creative and technical activity that transmits ideas through images. It’s more than aesthetics: it’s the visual layer through which strategy gets communicated, and it deserves the same care as any other discipline in the mix, whatever the toolkit looks like this year.
What has changed, fast, is what the toolkit can do. AI image generation, AI-native design platforms, and AI features inside professional software have made design more accessible than at any point in marketing’s history. Marketers are leaning on these tools because they’re fast, cheap, and good enough for most day-to-day work. A real trend, and worth using thoughtfully.
AI is a powerful part of the toolkit, but it isn’t the centre of design. The craft still sits underneath all of it, requiring the same things it always has.
The principles haven’t moved
Good design still requires a trained eye and the judgement to know which visual choices communicate the message versus just look pretty. It still requires the rigour to follow a brief, respect a brand, and execute with technical care. Colour theory, typography, hierarchy, composition, contrast, white space: these are still the foundations, no matter what generated the first draft.
Everything still starts with the brief
The most common cause of design problems is unclear thinking on the marketing side. Stakeholders who can’t articulate what they need, fill in a brief poorly, or change their minds every meeting are still the bottleneck. Specific briefs produce specific work. Vague briefs produce 10 rounds of “can we try something different.” This is older than design itself, and no technology has fixed it.
What AI has changed for the brief is previsualisation. When a stakeholder can’t describe what they want, you can generate 3 or 4 directions in 15 minutes, put them side by side, and let the brief sharpen against real choices. An hour-long meeting trying to extract a vision that doesn’t yet exist becomes a fast conversation against concrete options. The final piece comes later, with the brief now in focus.
Where AI fits: three kinds of tools
The first kind is AI-native design platforms like Canva and Adobe Express. These have evolved from template galleries into full design suites with AI baked in: layout generation from a prompt, brand-aware design that pulls your colours and logos automatically, one-click resizing, bulk variants from a spreadsheet. Both have generous free tiers, work for non-designers, and produce output good enough for social, presentations, email graphics, and most everyday assets.
The second kind is conversational AI with image capability: ChatGPT, Claude, Google Gemini, and others. These generate images, draft layouts, render text accurately, and are increasingly integrated with the platforms above. The choice depends on what the output needs to be: a finished image, an editable layout, or something living inside an existing workspace.
The third kind is professional design software: Adobe Creative Cloud, Affinity, Figma. These remain the gold standard, now AI-augmented with tools like Generative Fill and Firefly, keeping the depth and precision that high-end work requires. Creative Cloud is still the industry default for professional production; Affinity, now owned by Canva, is completely free and a credible alternative.
The brand foundation is what makes any of this work
This is the part most marketers skip and then wonder why their AI-generated designs look generic. To get useful output from any of these tools, the brand needs 3 things in writing.
First, a tone-of-voice document: how the brand sounds, with examples to use and avoid. Second, a visual identity guideline: logos in every variant, exact hex codes, fonts, image style, the anti-patterns that would make the brand look wrong. Third, a brand kit configured inside the platform itself, so AI generation pulls from those assets automatically. In Canva, that’s Brand Kit; Adobe Express has the equivalent.
Once it’s set, every output starts from the brand instead of a generic template. The difference between “this looks like our brand” and “this looks like a Canva template” is almost entirely whether the foundation was configured first. The brand asset library is the foundation the tools build on. The richer it is (clean logo files, approved photography, past designs that worked), the better the AI output.
Human designers are still essential
This is the part that gets misunderstood when people talk about AI replacing design. The list of work where a human designer is the right call has narrowed, but the remaining work is real, and the cost of getting it wrong is high.
Foundational brand identity still happens in Illustrator with a designer’s hand; AI produces concept ideas, not final marks. High-quality print and large-format work (packaging, billboards, anything where bleed and ink behaviour matter) carries too many technical pitfalls for generated output. Complex compositing and photographic art direction remain a Photoshop-and-experience job. Accessibility needs human review. Brand systems across many languages and channels need human governance.
There’s also something harder to measure: a designer’s point of view. A good designer brings a perspective on what the work should feel like that AI can’t replicate, because AI averages across what’s worked before. For brands trying to stand out, that distinction matters more than any feature comparison.
AI gives you the average. A designer gives you a take.
Agencies are adapting too
Where small agencies once charged for the production work of design, that work has compressed dramatically. The agencies thriving now charge for strategy, brand thinking, and the senior judgement that decides whether the AI output is the right thing in the first place. Production volume has stopped being what sets an agency apart. Judgement is what’s still scarce.
The human verification loop
AI design is good-enough output. It still produces strange artefacts: bad hands, malformed text, off-brand colours despite the brand kit. Every output needs a human eye before it goes live. Logos that haven’t deformed. Colours that match in print. Resolution adequate for the use. Type that hasn’t been squeezed to fit a frame. Image rights clear for commercial use.
The marketer sending design to production still needs basic technical literacy: file formats, colour modes, what a clean export looks like. The scarce skill now is recognising when the tool got it wrong.
The takeaway
Graphic design is still graphic design. The work still demands taste, judgement, a clear brief, and a strong brand foundation. The tools have changed dramatically and will keep changing fast. That gives marketers more leverage than they’ve ever had, and it puts more of the responsibility for quality on the marketer’s side of the table. When an AI tool gets a design 80% right, the gap between 80% and 100% is what separates work that looks branded from work that just looks AI-made.
Brand foundation. Clear brief. Fluent tool use. A designer for the work that needs one. That’s the model that holds up no matter how fast the tools change.