OutreachIQ is a five-workflow agentic AI system that autonomously sources B2B prospects, enriches them through a six-API waterfall, verifies the decision maker, produces consulting-grade diagnostic reports through Claude Opus, delivers those reports via a tracked link in a personalised email, and classifies every reply — all with human involvement at exactly two points: reviewing the diagnostic and personally sending the email.

The system is not a tool that assists with outreach. It is a pipeline that replaces the research entirely and makes generic outreach structurally impossible. Claude Opus cannot produce a generic diagnostic because it has to browse the actual website and find something specific. If it cannot find something specific, it flags it and asks for direction.

Five Layers of the System

• Layer 1 — Autonomous Sourcing (W0): Thirty-six search queries across two prospect categories fire daily. Six APIs run in waterfall sequence to find company data, decision maker names, tech stacks, and email addresses. Geography filtering, ICP validation, and large-company blocking run at the source.
• Layer 2 — AI Enrichment (W2): Groq processes every sourced prospect at speed — producing ICP scores, enrichment summaries, tech gap analysis, UX flaw detection, and preliminary bait findings. First-pass filter before Claude’s deeper work.
• Layer 3 — Deep Diagnostic (Claude Opus Project): Claude Opus browses the prospect’s actual website, verifies the decision maker, identifies four severity-classified findings, and produces a structured JSON diagnostic that maps every finding to a specific ANR service.
• Layer 4 — Report Delivery Infrastructure: Cloudflare Pages renders the prospect-facing report at report.anrdigitalstudio.com. A Cloudflare Worker logs every click before redirecting. The report uses a bait-hook structure — one finding visible, the rest locked behind a CTA.
• Layer 5 — Reply Intelligence (W4): Gmail API watches for replies. Groq classifies every reply — interested, not interested, out of office, referral, request for more information.

Technical Stack