Skip to main content
FirstCoast.ai
All case studies
Case study: Agency lead pipeline

Automated Site Analysis for Inbound Leads

When a prospect asks an agency about their existing site, the honest answer takes hours of manual review, so it usually becomes a generic pitch instead. We built a pipeline that runs the real review in minutes and drafts the response for human approval.

How the automated site analysis pipeline worksA prospect asks the agency's website chat what it can tell them about their site, sharing a URL and an email on a validated webhook. Their contact is filed in HubSpot the moment they ask, independent of everything that follows, so the lead is captured even if the analysis never finishes. The request then fans out to four checks that run in parallel: AI visibility, meaning whether AI engines cite the business, mention it, or leave it out entirely when asked about its category and city; search fundamentals; page speed; and email deliverability. Every module works from passive public sources, page fetches, headers, DNS lookups, and public APIs, with no scanning and no probing, and the modules are failure tolerant, so one slow API cannot stall the run. The results fan back in to an AI compose step bound by strict content rules: nothing sensitive goes in writing until the requester has proven the site is theirs. Two documents come out. A full internal report for call preparation, and a short client-facing summary saved as a draft in the agency's own Gmail, which a human reviews before anything sends.INBOUNDWEBSITE CHAT“What can you tell me about my site?”URLEmailValidated webhookTHE MOMENT THEY ASKCRMHUBSPOTContact filed firstbefore any analysis runsTHE BATTERYREAD-ONLY, NO PROBINGAI visibilitycited, mentioned, or absentSearch fundamentalsthe technical posturePage speedhow the pages performEmail deliverabilitywhether their mail landsONE SLOW MODULE CANNOT STALL ITAI compose stepstrict content rulesNOTHING SENSITIVE IN WRITINGuntil they prove the site is theirsTWO DOCUMENTSFull internal reportinternal, for call preparationGMAILNEVER AUTO-SENTClient-facing summarya human reviews every word

The challenge

  • "What can you tell me about my site?" is the highest-intent question a web prospect asks, and the hardest to answer well on the spot
  • A proper review of AI visibility, search fundamentals, performance, and technical posture took hours nobody had for an unqualified lead
  • Generic responses wasted the moment: the prospect asked for specifics and got a brochure

What we built

An automated analysis pipeline triggered from the agency's website chat. The visitor shares their URL and email; the system validates the site and runs a battery of read-only checks while the lead is still warm.

  • Checks how the business shows up when AI engines are asked about its category and city (cited, mentioned, or absent), plus search fundamentals, page speed, and email deliverability
  • Everything is passive and public: page fetches, headers, DNS lookups, and public APIs. No scanning, no probing
  • Two documents come out: a full internal report as call preparation, and a short client-facing summary saved as a Gmail draft
  • A human reviews every draft before it sends, and security observations never go in writing to someone who hasn't proven they own the site

How it's built

A validated webhook, failure-tolerant analysis modules (one slow API can't kill the run), an AI compose step with strict content rules, and drafts created in the agency's own Gmail, never auto-sent. The requester's contact flows into HubSpot the moment they ask, independent of the analysis.

Results

  • A question that used to cost hours of manual review now costs a review-and-send click
  • Prospects get specific, verifiable observations about their own site while their interest is hot
  • The full findings arrive as call preparation, so the follow-up conversation starts ahead

Want to see it from the other side?

Ask our assistant about your existing site. You'll have the summary within one business day.