In development · Built for our own sales team first
Railgun Marketing
AI outbound prospecting that shows its work: every claim in every proposal is backed by a source, checked by a second model, and reviewed by a human before it ever reaches a prospect.
Status: Internal use — we are running this on our own pipeline today. A packaged, self-serve version is next.
How it works
Six steps, one pipeline, one rule: nothing sends without your approval.
Research
Point it at a business and a campaign. It pulls the website, reviews, locations, and public signals into a prospect record.
Evidence
Every claim it wants to make gets a source: a URL, a quote, a screenshot. Nothing goes in a proposal without a citation.
Draft
An LLM writes the proposal and a short outreach email from the evidence — not from a template with the name swapped in.
Fact-check
A second, independent pass checks the draft against the evidence and flags contradictions, unsupported claims, and research gaps before a human ever sees it.
Your review
You read the proposal and the email, edit anything, and approve. Nothing sends itself. Test mode reroutes everything to your own inbox until you trust it.
Send & track
Approved sends go out under a daily cap, with replies, close probability, and estimated deal value tracked per prospect through won/lost.
Evidence-first research
Reviews, ratings, locations, and site content — pulled automatically and attached as evidence rows, not buried in a chat transcript.
Independent fact-check pass
A second model checks the first model. Contradictions and unsupported claims get flagged before anything reaches a prospect.
Short, human outreach
One tight email per prospect, not a drip sequence. Written from evidence, reviewed by a person, sent from a real address.
Deterministic lead scoring
A transparent scoring function — not a black-box "AI score" — ranks prospects by contact quality, evidence strength, and signals you define.
A real pipeline, not a spreadsheet
Research → Qualified → Drafted → Approved → Sent → Replied → Won/Lost. Every prospect has a status and a next step.
Test mode by default
New campaigns start rerouted to a test inbox with a daily send cap. You turn on real sending when you are ready, not before.
Why we're building it this way
Most AI outbound tools optimize for volume: generate hundreds of emails, spray them out, see what sticks. That produces the generic, half-true AI spam everyone already ignores.
Railgun Marketing was built for our own team's outbound first — where a bad proposal costs us a real relationship, not just an unsubscribe. Every claim has to trace back to a source. A second model checks the first one's work. And a human always reads the final draft before it goes out. That discipline is the product.
We're running it internally today. Once it's proven on our own pipeline, we'll package it as a standalone product — bring your own campaigns, evidence sources, and sending domain.
Want early access when it ships?
We're not taking sign-ups yet — the product is still internal-only. Reach out and we'll let you know the moment a design-partner beta opens.
Email hello@railgun.chat