The agent partially fulfilled the task by assembling a description of Saegus's business (AI consulting, enterprise focus, independent positioning) from the homepage and a web search result, but could not retrieve critical information about pricing, detailed service offerings, or competitive differentiation because those pages either returned 404 errors or were not published on the site. The site's client-side rendering, broken links, and deliberate information gatekeeping made it difficult to understand Saegus fully without external research.

• ›Step [2] (saegus.com/en homepage) was the only successful full-page fetch that yielded actual content, and the agent relied on it to establish baseline facts about company positioning ('partner for business reinvention through AI'). The response was returned but the agent could not extract rendered text from the Next.js-heavy HTML.
• ›Step [8] (web search) was the most content-bearing step and supplied critical facts: the company's mission statement, its leadership announcement, and proof that service pages exist (/en/offres/ai-for-business, /en/offres/data-ai-at-scale). Search results appeared in the snippet metadata, not from crawling the site itself.
• ›Steps [4], [6], [9], [10] all returned 404 errors for pages the search results claimed existed (/en/pricing, /en/services, /en/offres/ai-for-business, /en/offres/data-ai-at-scale), indicating either a broken site structure, a recent redesign, or URL mismatches between search indexes and live URLs. The agent could not retrieve pricing or service details.
• ›The agent explicitly noted three major information gaps it could not close: pricing models, competitive positioning, and detailed service specs—all core requirements of the task. These were not published in fetch-able form anywhere.
• ›The site's Next.js client-side rendering meant the HTML responses contained minimal text content; the agent had to rely on prior knowledge and web search results (57% from prior_knowledge and web_search combined per sourcing metrics) rather than on-site HTML, indicating poor machine readability.