GTM Engineering, Not Another Tool

SUMMARY

Three scheduled jobs, morning driver, response watcher, weekly digest, that pick accounts, research them, find contacts, write personalized outreach, and stage it in Apollo for human review.

PROBLEM

Reps were manually deciding which accounts to work each day and drafting outreach from scratch, eating into actual selling time.

SOLUTION ARCHITECTURE

A two-axis Fit + Signal scoring engine assigns each account a verdict: STRIKE, NURTURE, QUALIFY, or KILL. The morning job pulls the highest-priority accounts, a research agent builds context per account, a drafting agent writes personalized content, and everything stages in Apollo for a human to review and send. A response-watcher job tracks replies; a weekly digest reports pipeline health.

END STATE

Live and running daily. AEs review pre-drafted, prioritized outreach every morning instead of building it from scratch. The system never sends without human confirmation.

SUMMARY

An AI-driven pipeline that finds ICP accounts, identifies the buying committee, enriches contacts, and produces a reasoned SDR account plan automatically, at scale.

PROBLEM

SDRs spent 2 to 3 hours per account on manual research before they could even start outreach, with inconsistent coverage across verticals.

SOLUTION ARCHITECTURE

Programmatic ICP account discovery across 5 verticals feeds an AI research layer that identifies the actual buying party per account. A separate enrichment step fills in verified contact data, with a zero-result remediation loop that re-attempts "dead" accounts instead of dropping them. Output is a structured account plan, not just a contact list.

END STATE

11,000 accounts enriched, 93.9% verified contacts. Zero-result remediation recovered 23 of 64 initially dead accounts, adding 246 contacts.

SUMMARY

Bulk extraction and analysis of sales call transcripts to answer what buyers actually object to, and which answers lead to closed-won.

PROBLEM

Objection handling varied rep to rep with no data behind it, and leadership had no way to know why deals were really being lost beyond anecdote.

SOLUTION ARCHITECTURE

Built a custom extraction pipeline against Gong's POST-only API, since there's no native bulk export, to pull 300+ call transcripts. Classified discovery questions and objections across 318 calls, tagging 1,139 rep questions by type and outcome, then cross-referenced against closed-won and closed-lost data.

END STATE

90.8% of buyer pains raised in calls went unprobed by reps, a fact leadership didn't have access to before, replacing anecdote with a callable answer.

SUMMARY

A lead scoring and routing system that replaces a workshop-built MQL definition with one backed by actual revenue data.

PROBLEM

Sales didn't trust marketing's leads, marketing couldn't prove lead quality, and there was no shared definition of "qualified."

SOLUTION ARCHITECTURE

Clay handles the data layer, enrichment across 11 fields per lead. HubSpot runs the scoring logic, a formula-driven 0 to 100 ICP score with A/B/C/D tiering, built from patterns in past deals rather than assumption. Separation of concerns means data sources can be swapped without touching the scoring model.

END STATE

Every lead is scored before a rep sees it, against criteria tied to what's actually closed before. No more "is this lead any good" argument between teams.