Deep research on every prospect. Zero fabrication.
SDRs spend 60 to 70% of their time researching accounts before one message gets written. Our system runs four data sources per prospect in parallel, cross-checks everything against a whitelist, and rejects quotes that do not byte-match an approved source.
How it works
-
Four sources, all independent.
Apollo enrichment (firmographics, funding, headcount). Clay MCP (recent news, open jobs, tech stack, funding, headcount growth). Brave web search (live web content). A direct website crawl of the homepage and four interior pages. Website content is ranked as the most authoritative source on what a company actually does.
-
Apollo-only claims are flagged, not cited.
Apollo data is useful but not verified. Revenue, funding, ARR, customer counts, and employee counts coming from Apollo alone get logged in an apollo_only_claims field and the research confidence is capped at medium. Data aggregators do not count as independent confirmation. Only journalism, press releases, or the company's own site.
-
Company status check.
Before any sequence writing, the research agent checks for acquisition, merger, shutdown, or rename signals. If a company is no longer active, the output reflects current reality. Nobody gets an email addressed to a company that got bought last year.
-
Stats come from a Kyle-curated whitelist.
Any cited statistic in any sequence must be selected from config/approved_stats.json, where every entry has a verbatim quote, a live URL, and a Kyle-verified date. The generator byte-matches the rendered quote against the whitelist row. If it does not match, the sequence is rejected and regenerated.
Want to see a real research brief?
Send us your target personas on the discovery call. We will walk through a sample brief from start to finish, sources and all.