The agent gathered enough information to provide a coherent explanation of First Concepts' purpose, target market, pricing tier, and differentiation, but had to rely heavily on web search and prior knowledge because the website itself lacked detailed feature documentation and public pricing specifics. The site's minimal navigation (404s on /features and /docs) forced the agent to piece together a narrative from external press coverage and funding announcements rather than authoritative product pages.

• ›The agent's core claims about First Concepts being a 'taste-aware AI orchestration layer' for creative agencies, the 70% speed improvement claim, client names (Mother, R/GA), and the £50/seat/month Team tier all came from web search results [8, 12, 14, 16], not from firstconcepts.co itself. The homepage and pricing page fetches [1, 3, 5] returned only HTML scaffolding without accessible text content, suggesting a client-side rendered site that the agent could not parse.
• ›The agent discovered that standard subpaths (/about, /features, /docs) returned 404s [6, 10, 11], indicating either a single-page app with no public documentation or a very minimal public site. No internal navigation links were extractable from the homepage fetch to guide discovery.
• ›The agent relied on external journalism (UKTN article [15], Tech.eu, StartupRise) to validate product claims and pricing rather than finding this information on the domain itself. The site surfaces almost no machine-readable product detail, forcing agents to either know the product beforehand (63% prior knowledge sourcing) or search externally (25% web search).
• ›Key gaps the agent explicitly called out—exact feature set, differentiators vs. Frame.io or Brandfolder, Agency plan pricing, specific tool integration depth—remain unanswered because the site does not publish them in a discoverable form. The agent had to admit 'Limited public product details' and note that understanding required 'visiting the live site or contacting them directly.'
• ›The site is poorly navigable for agent-led discovery: no linked docs, no FAQ, no help section, and the homepage itself is not directly parseable as plain text. The agent succeeded only by pivoting to external sources, which is a failure of the site's own information architecture, not the agent's search capability.