Exa
Back to Demo

How It Works

One domain becomes a cited account list, scored prospects, decision makers, and always-fresh monitoring.

This is just one simple approach — with savvy Exa queries, Exa can uncover all people and company data that exists on the web.


This demo runs a 5-step pipeline that goes from a single domain to a fully enriched prospect list with decision makers. Every step uses Exa's deep search with output_schema to extract structured data directly — no LLM post-processing needed.

  1. Identify a company from its domain
  2. Discover a sample of Total Addressable Market (TAM) with parallel searches
  3. Enrich & score every prospect with structured deep search
  4. Find decision makers at top prospects
  5. Set up always-on monitoring with Exa Monitors for always fresh data

The Pipeline

1

Identify the Company

A single exa.search() call with type="deep" and an output_schema extracts a full company profile — name, description, ICP, target customer, known customers with domains, and recent news. No LLM needed.

The output_schema tells Exa exactly what fields to extract. The response comes back as structured JSON — no parsing or LLM extraction needed. Known customers are returned as structured objects with domains for favicon display.

2

Discover the TAM

Using the company profile from Step 1, we build multiple search angles and run them in parallel. Each angle uses a system_prompt that includes seller context and previously-seen domains to avoid duplicates.

All 4 angles run in parallel. The system_prompt includes seen domains so Exa deduplicates across angles automatically.

3

Enrich & Score

One deep search call per company extracts 8 structured fields — funding, headcount, growth signals, tech stack, competitive position, recent news, and a personalized fit score (0-100). The seller context lives in both the system_prompt and the output_schema field descriptions, so Exa evaluates fit directly.

All enrichment calls run in parallel, staggered at ~1 req/sec. Each call extracts structured data directly — no LLM post-processing. The fit rating is a 0-100 score evaluated by Exa based on the seller context embedded in both the system prompt and the schema field descriptions.

4

Find Decision Makers

One type="deep" search per company with category="people" and an output_schema returns structured decision-maker profiles — names, titles, LinkedIn URLs, and relevance — directly from Exa. No separate LLM vetting call needed.

Every step in this pipeline — including people search — uses zero LLM calls. Exa's deep search with output_schema handles all structured data extraction end-to-end.

5

Always Enrich with Monitors

Enrichment data goes stale. Exa Monitors re-run your queries on a schedule and push fresh data to your webhook — so when a prospect raises funding or doubles headcount, you know before your competitors.


That's the entire pipeline — from domain to decision makers, powered entirely by Exa search. The key insight is that output_schema lets you skip the LLM extraction step entirely. Exa returns structured data directly.