Smashing Magazine: why your website should never stop changing
Pierre Burgy, co-founder and CEO of Strapi and Fimo, published this August 2026 piece to challenge a common assumption about autonomous websites: that “full autonomy” means removing humans from the loop entirely. The article makes a case for composable AI agents — systems where delegation is granular, adjustable per person and per task, and expanded only as agents prove reliable over time.
The central observation is practical: most website work is maintenance, not creativity. Burgy categorizes ongoing website tasks into three tiers. The first, which he estimates at roughly 80% of the total, consists of rule-based work — accessibility checks, propagating design system updates, broken-link detection, legal compliance sweeps. These require no judgment and are prime candidates for agent automation. The second tier covers brand and aesthetic decisions: tone of voice, visual direction, what “feels right” in context. These cannot be delegated, because they involve taste, institutional knowledge, and accumulated brand equity that agents do not carry. The third tier sits between the first two and varies by person — dark mode defaults, for example, might seem like routine configuration to a developer but like brand identity to a designer.
The practical implication is that an “autonomous website” should be a variable setting, not a binary switch. A composable agent setup means the legal team might fully delegate compliance checks while retaining control over all copy changes, while the design team keeps sign-off on any component that touches brand expression. Trust in agents develops through visible results on low-stakes tasks and extends gradually as those results accumulate.
Burgy is not arguing against human oversight — he is arguing that oversight should be proportional and task-specific rather than uniform. The framework is useful for design teams debating where to introduce AI automation without losing authorship over what their products communicate.
The piece is sponsored by Fimo, the platform Burgy co-founded, which is worth noting when evaluating the objectivity of specific product references. The underlying framework for categorizing task types is useful regardless of which tooling a team eventually adopts.
Useful for design leads and product managers who are building internal AI workflows and need a principled way to draw the line between what agents can decide and what humans must own.