AI is producing more design assets than ever. Here's why brand teams in Southeast Asia must raise the quality bar — not just the output volume.
Ninety percent of everything is mediocre. Sturgeon’s Law — coined by sci-fi writer Theodore Sturgeon in the 1950s — has never been more relevant to a marketing team than it is right now.
As AI tools flood the pipeline with generated UI components, ad variations, and branded assets, the actual constraint has shifted. It is no longer can we produce enough design? It is do we have the systems to filter out the 90% that shouldn’t ship? For growth and marketing teams across Southeast Asia managing multi-platform campaigns across Shopee, LINE, TikTok Shop, and their own web properties simultaneously, that distinction has real revenue consequences.
Sturgeon’s Law Is Now a Design Infrastructure Problem
Chris Becker’s analysis on UX Collective frames the AI content explosion through Sturgeon’s Law precisely: more generation capacity does not produce more quality — it produces more volume to sort through. The problem isn’t the AI. The problem is that most organisations have invested in generation pipelines without investing in quality gates.
From a data architecture perspective, this mirrors a pattern I see constantly with analytics stacks: companies build ingestion pipelines before they build validation layers. The result is dashboards full of data that nobody trusts. The same logic applies to design pipelines running on AI tooling. Without defined quality criteria baked into the workflow — brand token compliance, accessibility contrast ratios, mobile viewport testing — you’re shipping the equivalent of unvalidated data to production.
The fix isn’t slowing down generation. It’s treating quality filtering as infrastructure, not a final review step someone does in a hurry before a campaign goes live.
Perception of Value Is a Design Decision, Not a Budget One
A photography project by Adriana Mora and Celia Pladevall, highlighted by It’s Nice That, makes an elegant counterpoint to the volume argument. Their Commercial Candy series takes corner-shop sweets — genuinely mass-market objects — and reframes them through lighting, composition, and art direction as luxury items. The product doesn’t change. The perception does.
This is directly instructive for UI design. The same product page on Lazada, photographed and laid out with intent versus assembled from a generic template, can communicate entirely different brand positioning to the same shopper. Brands like Pomelo Fashion and Naiise have demonstrated this in Southeast Asian e-commerce: considered visual hierarchy and deliberate negative space on product pages don’t just look better, they convert better. Pomelo’s mobile-first design approach, with its high-contrast typography and restrained colour palette, reflects a brand decision as much as a UX one.
The lesson: perceived quality is a multiplier on whatever product you’re actually selling. AI can generate the assets. It cannot make the brand decision about what those assets should communicate.
Mobile-First Means Quality Constraints Are Stricter, Not Looser
Southeast Asia’s mobile penetration context makes this more urgent, not less. With over 70% of e-commerce traffic in markets like Indonesia, Thailand, and the Philippines coming from mobile devices, UI quality issues that might be tolerable on desktop become conversion killers on a 6-inch screen with variable network conditions.
Specifically: AI-generated UI components frequently fail on two fronts in mobile contexts. First, touch target sizing — buttons and interactive elements generated without explicit mobile constraints often fall below the 44x44px minimum that both Apple’s HIG and Google’s Material Design guidelines specify. Second, visual hierarchy collapses on smaller viewports when components are designed at desktop scale and scaled down rather than designed mobile-first.
Platform-specific considerations compound this. Shopee and Lazada both have their own UI conventions that shoppers have internalised. Brand assets that feel visually jarring relative to the platform context — even if technically correct — create friction. The implication for teams using AI generation at scale: mobile viewport testing and platform-context review need to be mandatory gates in the pipeline, not optional QA steps.
Building the Quality Filter Into the Design System
The practical question is: what does a quality gate for AI-generated design actually look like in a mid-size brand team? Three components matter most.
First, brand token enforcement. If your design system doesn’t have a defined, machine-readable set of colour, typography, and spacing tokens, AI generation tools have nothing to check against. Establishing these tokens in a tool like Figma’s Variables or a Storybook configuration means automated checks can flag non-compliant outputs before a human reviews them. This is the equivalent of schema validation in a data pipeline — catch bad structure early.
Second, a tiered review process. Not every asset needs senior creative review. Classify outputs by exposure and risk: a hero banner on a homepage warrants different scrutiny than a social story variant. Build a triage layer so human attention goes where it has most impact.
Third, a defined feedback loop from performance data back to design criteria. If a particular layout pattern consistently underperforms on mobile click-through across three campaigns, that’s a signal to update the generation prompt constraints and the quality checklist — not just to pause that one campaign. Connecting design output metrics to the criteria you’re generating against closes the loop and makes the system smarter over time.
Forward-Looking Close
The brands that will differentiate on design in the next 24 months won’t be those with the fastest AI generation pipelines. They’ll be the ones who built the quality infrastructure to make those pipelines trustworthy. In a region where consumer attention is split across a dozen platforms and brand switching costs are low, the margin between a UI that communicates quality and one that doesn’t is also the margin between retention and churn. The real question is: who in your organisation owns that quality layer — and do they have the tools and authority to enforce it?
At grzzly, we work with growth and marketing teams across Southeast Asia to build design systems and campaign pipelines that don’t just produce assets fast — they produce assets that hold up under scrutiny, across platforms, and at scale. If you’re rethinking how your team manages design quality as AI tooling becomes standard, we’d like that conversation. Let’s talk
Sources
Written by
Chunky GrizzlyDesigning the foundational plumbing — data warehouses, lakehouse models, and ETL pipelines — that separates organisations with genuine intelligence from those drowning in dashboards.