The problem
Why this system exists
Standard lists identify companies that fit on paper but do not explain why an account may need help now or what evidence should shape the first conversation.
Signal-Based Prospecting
Public Company AI Prospecting
Finds smaller US-listed companies showing public evidence of AI need, scores the evidence, finds decision-makers, and drafts outreach for review.
The outcome
What it produces
A ranked, evidence-backed account list with decision-makers and drafted outreach, plus a feedback loop that shows which signals lead to replies and meetings.
System overview
How the playbook moves from input to outcome.
Each stage has one job. The sequence stays visible while the active state moves through the system.
01
Discover
Screen the SEC universe or ingest selected tickers.
02
Enrich
Collect filing and web evidence with cited sources.
03
Score
Rank intent, capability gap, timing, and commercial fit.
04
People
Find the right decision-makers at qualified accounts.
05
Messages
Draft evidence-grounded outreach without sending it.
06
Learn
Record outcomes and examine which signals convert.
Evidence score
Four components add up to 100.
Intent30
Capability gap25
Timing25
Commercial fit20
Decision bands
65–100 · Qualified
Requires a hard signal and fresh outreach angle.
45–64 · Human review
The system pauses instead of forcing a verdict.
0–44 · Disqualified
No further research budget is spent.
Inputs
What the system needs
- SEC company universe or ticker list
- Public filings and XBRL data
- Public web research
- Configurable ICP profile pack
Outputs
What comes out
- Qualified and tiered company list
- Cited evidence and outreach angles
- Decision-maker contacts
- Four-step message drafts
- Outcome and signal-attribution reports
Inside the repository
The important pieces, without the file-system noise.
The explainer shows the operating structure. The repository remains the source of truth for complete setup and implementation details.
GTM-Public-Traded-AI-Accounts/
- 01src/pipeline/: discovery through analytics modules
- 02config/: default ICP, services, personas, and voice
- 03profiles/: client-specific configuration overlays
- 04sql/schema.sql: Supabase data model
- 05docs/: architecture, signals, and operating runbook
- 06data/: local caches, queues, and exports
Guardrails
What the system deliberately protects.
- Nothing is sent automatically.
- Paid research calls obey configurable per-run caps.
- Borderline accounts remain in a human review band.
- Every meaningful signal keeps evidence and a source URL.
Quick start
From repository to first useful run.
- 01Install Python 3.12+, uv, and the project dependencies.
- 02Configure SEC identity and a Supabase project.
- 03Apply the database schema.
- 04Ingest one ticker and run EDGAR enrichment.
- 05Review funnel status before expanding the batch.
Adapt the system
Want a version built around your market, tools, and sales process?
Keep exploring
Related playbooks
Claude Code SkillsOpen source
GTM Flywheel
Fifteen structured Claude Code skills that turn ICP research, signal scoring, outbound copy, campaign analytics, and sales intelligence into one compounding GTM system.
ICP Research
Signal Scoring
Cold Email
A versioned set of operating skills with scoring models, decision trees, templates, and quality checks that Claude Code can apply consistently to your own data.
Sales EnablementOpen source
Microsite Pipeline
Turns one sales lead into a researched, personalized eight-page sales microsite and PDF using sourced company data.
Lead
Enrich
Synthesize
A repeatable local pipeline that researches a company, generates TAM, ICP, and sales-signal insights, then renders an on-brand HTML page and PDF without fabricating missing data.