The agent could not access AFI's full website content due to JavaScript rendering and instead assembled an explanation from web search results and prior knowledge about the Indonesian facade contracting industry. It successfully identified AFI as a curtain wall contractor backed by an aluminum extrusion manufacturer and explained their value engineering positioning, but could not retrieve pricing, detailed service tiers, project portfolio, or explicit competitive differentiation from the site itself—all information came from external searches and domain knowledge inference.

• ›All seven successful steps were web searches [16, 17, 19, 20, 22, 23, 25], not direct fetches from afi.id. The agent tried to visit afi.id/ and its subpages (/en, /en/services, /en/about, /layanan, /tentang) but received only HTML skeletons with JavaScript bundles—no rendered content with business information, clients, services, or pricing.
• ›The agent inferred AFI's core business (curtain walls, ACP cladding, unitized systems) and backing (PT. Damai Abadi aluminum extrusion) from web search snippets and external industry sources, not from afi.id's own pages. The search query results explicitly mentioned Damai Abadi as an aluminum manufacturer and linked AFI to value engineering practices found in construction industry articles.
• ›The site is hostile to static content discovery: it uses Next.js with client-side rendering, no embedded metadata about services/pricing in the HTML head, and no accessible static alternative. The agent had to bridge the gap using prior knowledge (60% of fetches sourced from prior knowledge per the run metadata) and external web searches (40% from previous artifact/context), leaving critical details like pricing, service tiers, and project examples completely unavailable.
• ›The agent explicitly acknowledged three major information gaps in its final response: (1) no pricing model disclosed, (2) no explicit service tier breakdown, (3) unclear competitive positioning vs. named competitors like PT. Polareka Fasadindo. These gaps exist because the website does not publish them in a machine-readable or static form.