(Rae · Ext. 108 · knowledge base)

Research Analyst.

Turns an ICP brief into verified, scored, evidence-annotated lead lists and competitor intel — every claim with a source URL.

Launch this role

Rae cross-checking several source packets at a quiet table in the Stavatar workspace — **Rae**Ext. 108 · Research Analyst

What it does from day one

(no ramp-up, no aspiration)

Targeted lead lists from an ICP brief: Apollo search filters, paginated pulls, CRM dedupe, clean Sheet delivery
Waterfall enrichment across providers (~80%+ email match vs 40-50% single-source) with verification on every address
Claygent-style agentic web research per account: pricing pages, job boards, press — answering questions like 'do they sell enterprise?'
ICP fit scoring on a weighted 0-100 rubric with A/B/C tiers and a 'why this account, why now' note per row
Buying-signal sweeps on a watchlist: job changes, funding, hiring spikes, tech-stack changes
Weekly competitor briefs: pricing-page diffs, feature launches, G2 themes, auto-updated battlecards

How the work flows

(brief to logged output)

Brief intake & ICP spec — structured intake plus 5 dream customers and 5 bad fits; agent converts to filters and a scoring rubric, plays it back for approval
Source & build — query Apollo and search APIs to assemble the raw universe; dedupe against CRM and suppression lists
Enrich & verify — waterfall enrichment, agentic research fills contextual fields, every email verified (<2% projected bounce)
Score & annotate — apply the rubric, tier the list, attach evidence notes with source URLs per row
Deliver & log — push to staging Sheet/CRM, Slack summary, run stats logged; recurring signal jobs feed weekly deltas back in

The stack it plugs into

(real tools, official APIs)

Apollo.io API — people/org search (free), enrichment credits, job-posting and news endpoints
Email verification (NeverBounce/ZeroBounce/MillionVerifier) — keeps bounces under 2%
Exa or Firecrawl — LLM-native search and scrape-to-markdown for agentic research and page monitoring
Apify LinkedIn actors or Bright Data — LinkedIn data post-Proxycurl-shutdown; vendor legal record matters
HubSpot/Salesforce/Pipedrive APIs — dedupe and approved upserts with research notes as activities
Google Sheets/Drive + Slack — the actual deliverable surface and digest channel
Crunchbase + BuiltWith/Wappalyzer + job-board feeds — the trigger-signal layer

Its machine

(8GB RAM · 4 vCores, dedicated)

One headless Chromium via Playwright for research and competitor page diffs, a Python/Node worker for API calls and dataframe work, a scheduler for signal sweeps, and SQLite/Postgres for watchlists and diff history. Enrichment is I/O-bound, so 4 vCPU is ample; 8GB holds one browser plus worker with hygiene (restart browser every N pages, reuse contexts).

Seat economics

heavy token load

Heaviest non-voice role: per-row agentic research means 3-10 page fetches through the model per prospect — a 1,000-row deep list can hit 50-150M input tokens uncapped — so fair use meters deep-researched rows per month (Clay-style credits), with structured APIs first and deep research reserved for Tier-A rows.

Flat seat from $499/agent/mo — one agent, one dedicated machine. See pricing.

Judged by

(the numbers that matter)

Data accuracy — verified-email bounce <2%, matching the industry-standard 95%+ accuracy-with-replacement guarantee
Coverage/match rate — % of target records completed with valid contact data (waterfall benchmark ~80%+)
Qualified volume and downstream lift — Tier-A accounts/week that convert, plus minutes-per-brief vs the ~18 hrs/week manual baseline

Tasks in this role

(each documented in the task library)

Lead List Building
An ICP brief goes in; a clean, deduped, verified lead list comes out.
Waterfall Data Enrichment
Multi-provider enrichment that finds emails a single source misses — roughly 80%+ match versus 40-50% single-source.
ICP Lead Scoring
Every account scored on a weighted 0-100 rubric with A/B/C tiers and a 'why this account, why now' note per row.
Buying Signal Tracking
A watchlist of target accounts swept continuously for the moments that make outreach land: job changes, funding rounds, hiring spikes, and tech-stack changes.
Competitor Monitoring
Weekly competitor briefs built from pricing-page diffs, feature launches, and G2 review themes, with battlecards that update themselves.

Never alone

(guardrails — how trust works here)

Never push directly into a client CRM or outreach tool — deliver to a staging sheet; human approves before upsert (bad merges are hard to unwind)
Compliance gate on sourcing: no scraping behind logins, no fake accounts, EU records flagged with a documented legitimate-interest basis, source-of-record kept per row
ICP rubric changes require sign-off — agent proposes recalibrations, human approves; silent rubric drift corrupts every downstream list
Every competitor claim carries a source URL; anything unverifiable is marked 'unverified,' and per-provider spend caps hard-stop with escalation, never auto-overage

Full policy: our AI disclosure.

Try it

(alpha)

"Watch it research a company live": paste any domain and watch a streaming activity feed ('reading pricing page… found 3 open SDR roles… detected HubSpot + Segment… Series B, Jan 2026') resolve into a scored one-page dossier with an ICP fit dial and three talk-track angles. Public data only, self-personalizing, and the dossier hands off naturally to the cold-calling role. Cold calling ships first — try that demo today.