ForgeGTM

Our owned answer to Explee / Clay / Apollo. We build the engine — we don't rent it.

Vision Board Status: Draft v1 Updated 2026-07-22 Origin: Explee teardown brainstorm
The durable primitive isn't a 500M-row database. It's "describe your buyer in plain English → get a scored, enriched, ranked list with a one-line reason for each fit." We don't need mass volume — we need precision on the verticals we actually sell into. ForgeGTM is a lean, owned pipeline: per-campaign data pulls + AI fit-scoring + enrichment + human-approved outreach. No lock-in, no credit meter, no autopilot spam grenade.
"Point it at a vertical, describe the buyer, get a ranked hit-list with reasons — then approve the outreach batch. We own the engine, not rent it."

The Doctrine

  1. We own the engine, never rent it. No third-party lead platform in the critical path.
  2. Precision over volume. 100 perfectly-scored prospects beat 10,000 scraped rows.
  3. Per-campaign pulls, not a mega-DB. Fresh, per-vertical, per-run — cheaper and legally cleaner.
  4. AI fit-scoring is the magic. The part Explee charges $329/mo for is one prompt loop for us.
  5. Draft-and-queue, never autopilot-send. Human-in-the-loop on all outbound. The hard divergence from Explee.
  6. Deliverability is sacred. Warmed infra, isolated cold domains, never our primary domains for volume.
  7. The engine is portable. Same scoring aims at grants, partnerships, acquisition targets, hiring.

The Toolkit

01

Vertical Data Puller

Per-campaign sourcing: Google Places, LinkedIn public, Shopify / vertical registries. Raw list with domain + metadata. No licensed mega-DB.

02

Semantic Fit-Scorer THE WEDGE

Website + metadata vs. ICP → fit score (0–100) + one-line "why they fit." Build this first; prove quality on 25 real prospects before anything else.

03

Enrichment Agent

Multi-page site read → decision-maker, tech stack, buying signals, contact paths. Our deep-research subagent pattern, aimed at a lead list.

04

Personalization Composer

Personalized opening hooks per prospect from enrichment. Output is a draft batch — not a send.

05

Approve-and-Send Queue

Dwight reviews → approves → sends via warmed infra on an isolated cold domain. Reply-handling stays human-gated. Explicitly not autopilot.

06

Campaign Ledger

Sent / opened / replied / booked per campaign + vertical. Measure reply-rate vs. our current manual approach.

Roadmap

0

Prove the wedge NOW

Build the Semantic Fit-Scorer standalone. Run against 25 real board-game retailers. Eyeball scores + reasons with Dwight. Quality there → greenlight the pipeline.

1

Minimal pipeline

Data Puller (Google Places) → Fit-Scorer → CSV out. One vertical end-to-end, no outreach yet.

2

Enrichment + Composer

Add enrichment agent + personalization drafts. Output an approve-ready batch.

3

Guarded send

Warmed cold domain + approve-and-send queue + campaign ledger. Measure reply-rate on one small campaign.

4

Aim it elsewhere

Point the same engine at grants, partnerships, acquisition targets. Prove portability.

Boundaries & First Targets

We explicitly do NOT copy
  • The credit-metering business model
  • The fully-autonomous cold-emailer (domain-reputation liability)
  • The 500M-row licensed dataset (legal + cost baggage)
First ICPs to target
  • Board-game retailers (High Nooniverse + Game Changer)
  • OpenGym gym prospects
  • MMA / LEGO-flip adjacent buyers
  • (Portable) grants & partnership targets