The problem

A small sales team needs a steady flow of genuinely qualified prospects but has no appetite for a custom app or dashboard, and can't afford to burn enrichment credits scoring companies that were never a fit.

Off-the-shelf lead lists don't reflect the client's actual ideal-customer profile or current buying signals, and generic AI scoring hallucinates fit instead of grounding it in real firmographics and evidence.

The approach

A portable, spreadsheet-native pipeline (no bespoke software by client requirement) sources a company universe, applies a cost-gated qualification filter before any expensive enrichment, then enriches survivors with firmographics and buying signals.

Each company is scored on a four-dimension 0-100 model: three deterministic formula dimensions plus one grounded LLM step that returns JSON situation-fit with a summary and a suggested outreach angle, so the AI read is evidence-based and auditable rather than a black-box number.

Every scored row lands in a weekly Google Sheet as 'awaiting approval' with a verified decision-maker attached — assisted, human-in-the-loop, nothing auto-sends — and the whole spine ports to a new vertical by swapping configuration rather than rebuilding.

What I built

  • Four-dimension weighted score (0.35 situation-fit + 0.25 profile-fit + 0.25 signal-strength + 0.15 decision-maker-identified) mapping to Tier 1/2/3 with a suppression cutoff, all weights and tier bands owner-calibrated rather than hard-coded
  • A single grounded LLM column returning structured JSON situation-fit, one-line summary, and outreach angle per company
  • Signal-strength scoring as a weighted sum of buying signals with recency decay, plus a cost-gate stage that filters the universe before spending enrichment credits
  • Decision-maker email waterfall and CSV-based suppression of already-contacted accounts, delivered via a native weekly Google Sheets export
  • A verify-on-fixture build discipline: every pipeline stage is authored as a copy-paste artifact and validated on a small sample before any full-list run to control credit spend
  • Portability proven across two distinct verticals from the same spine, plus an alternate trade-show-exhibitor source harvested into a multi-thousand-row prospect CSV
  • Clay
  • Claude (grounded JSON situation-fit)
  • Google Sheets
  • Apify
  • Python

Result

A configuration-not-a-rebuild lead engine that turns a raw company universe into a calibrated, tiered, decision-maker-attached prospect sheet each week, with the AI fit-read grounded in real signals and every row gated on human approval.

Confidential engagement — reference on request← Back to all work

Want something like this for your business?

Book a free consult