Key Takeaways

  • Open text prompt boxes fail for visual creation because users cannot describe the exact aesthetic they want before seeing it.
  • Canva anchors its AI generation in a library of over 165 million templates, graphic assets, and images instead of starting from raw text.
  • The company pays human designers directly whenever their templates or visual styles serve as reference inputs for AI generations.
  • Brand custodians inside companies are more necessary than ever to filter out generic "AI slop" and preserve visual identity.

The Blank Box Trap

Every AI startup rushed to put an open text box on their home screen. You type a sentence, wait five seconds, and get an image. Canva co-founder and COO Cliff Obrecht thinks that interface is a dead end for visual work.

The issue is human psychology. Most people do not speak the language of art directors. They do not know color palettes, aspect ratios, or typography terms. When faced with a blinking cursor, they freeze or type something bland.

“People don't know they want a psychedelic retro '80s themed birthday party invitation,” Obrecht told John Collison. “But when they see it they're like, 'Oh jeez, I really like that.'”

Visual taste relies on recognition, not recall. Giving someone a text box asks them to invent an aesthetic from scratch. Giving them visual options lets them select, remix, and refine.

Paying Humans for AI Inputs

To make generative tools useful, Canva ties AI directly to its library of curated assets. Instead of generating pixels out of thin air from web scrapes, the platform points models at structured templates built by professional designers.

“We've got over 165 million templates, images, and graphic assets on the library,” Obrecht explained, “and we really don't think the prompt box is the right vector for creating visual content.”

Canva also built a creator payout model into this workflow. When an end user builds on top of a specific layout, the original designer gets paid.

“And we are now one of the only platforms, if not the only platform, where you can generate with AI and reference templates created by real humans that we then pay those real humans for those input images, rather than just training on the internet,” Obrecht said. “So, if you spot that psychedelic template, etc., etc., then if you import that into your creation, then we also give the attribution and payment to that creator.”

This structure creates an incentive loop. Professional designers continue submitting fresh cultural trends because they earn money when users build on their work.

Brand Custodians Beat AI Slop

Unconstrained generation creates an explosion of low-quality, generic graphics. When anyone on a team can generate a poster in three seconds, company visual identity falls apart fast.

Obrecht argues that the rise of synthetic media makes professional designers more valuable to organizations, not less.

“There is so much AI slop out there that you really need custodians of the brand within organizations,” Obrecht noted, “and so I still think the role of designers, brand designers, and anyone in that space really has a lot of value to add.”

Instead of letting employees generate unbranded assets from scratch, software must enforce guardrails: approved fonts, exact color values, and pre-set component layouts. The AI handles tedious variations while human designers define the rules.

What to Do With This

Audit your product's AI entry points this week. If you force users to type raw prompts into an empty input field, replace that box with three curated presets and a structured gallery. Give users visual examples to react to, then let the model modify the selection rather than generating from zero.