Build an AI Lead Research and Prospecting Agent
Build a capped evidence-first prospecting agent that researches approved public signals, preserves raw records, explains uncertainty, remembers prior events, and drafts without sending.
Define the Offer and Research Decision
Choose one offer, customer, geography, exclusion set, and decision the research should improve.
Research and Select Observable Signals
Compare source-backed event hypotheses and choose one or two signals with date and evidence rules.
Approve Sources, Caps, Cost, and Privacy
Allowlist sources and define canary size, full-run limits, cost ceiling, timeout, retries, privacy, and secrets.
Preserve Raw Evidence and Normalize Records
Create a run manifest, save immutable raw records, assign stable IDs, and preserve field-level evidence.
Deduplicate Entities, Events, and Repeat Runs
Use stable keys, retain duplicate links, separate company and event identity, and preserve review state across runs.
Classify Evidence Without Inventing Intent
Separate facts, derived values, hypotheses, recommendations, unknowns, and visible classification rules.
Build the Review Report and Grounded Drafts
Render scope, reconciliation, evidence, unknowns, safe links, exports, and editable drafts from one normalized run.
Run the Canary, Repeat, Recover, and Hand Off
Test success, stale, duplicate, uncertain, hostile, empty, partial, and failed-source behaviour before any future automation.
What you will work towards
Work through this course to prepare an evidence-first research and review pack. The curriculum takes you through source/cost caps, preserved records, entity and event deduplication, evidence classification and editable drafts.
Start with one offer, approved public sources and one or two observable signals. Use synthetic information while learning when permission to use live records is unresolved.
A practical case to carry through the lessons
A new location mentioned on a company website is an observation. It does not prove that the business wants to buy your service. Retain the dated source beside the interpretation.
Run the same input twice and check whether the agent recognizes the existing entity and event. Keep drafts separate from sending.
How to know the work is ready for review
Before calling the exercise complete, check that a reviewer can verify a signal, see uncertainty and distinguish a newly observed event from a repeated one.