Inbound lead enrichment agent | GTM AI agents

What it does

A form fill usually gives you an email address and nothing else. Someone on the team then tabs between LinkedIn, the company website and an enrichment tool to figure out whether this lead is worth ten minutes or ten seconds. On a good week that research happens; on a busy week it does not.

This agent does the research the moment the form is submitted. It classifies the email before spending anything, reverse-looks-up the person, resolves their company through a chain of fallbacks, checks the match against independent signals, scores the lead against your ICP, and posts one card per lead to Slack with the tier and the evidence.

It is a read-only agent by design: it posts intelligence to your team, it never messages the lead. And it is honest about uncertainty. When it cannot identify someone, the card says so instead of presenting a guess as a fact.

Before and after

Before

A form fill is an email address in a list
Leads get googled when someone has time
Qualification depends on who happened to look
Weak leads get the same attention as great ones

With the agent

A Slack card with person, company and tier, minutes after submit
Enrichment spend gated so junk and internal emails cost nothing
One scoring rubric applied to every single lead
The team opens the CRM already knowing which leads matter

How it works

Six gates, and the ordering is the point: classification before spend, corroboration before claims, honesty before polish.

1
Classify the email before spending anything
Internal and test addresses are skipped silently. Personal mailboxes and professional domains take different paths, so an enrichment miss on a throwaway address costs nothing.
2
Reverse-lookup the person
The email goes through a reverse enrichment lookup: name, role, employment history, and the company domain when one can be established.
3
Resolve the company through ordered fallbacks
The company is resolved from the enrichment result, the professional email domain, or a self-reported domain from the form, in that order. The resolved domain is then run through store-level intelligence: commerce platform, catalog size, vertical, region.
4
Corroborate before claiming
A company match is only presented as a match when at least two independent signals agree. Anything weaker is labeled low confidence and shown as candidates, not conclusions.
5
Score against your ICP
A deterministic scoring pass turns role, company size and vertical into a tier. Same rubric for every lead, no mood-dependent qualification.
6
Post one card per lead, honestly
Full profile when the evidence supports it, low-confidence candidates when it partially does, and a plain "couldn't identify this one" when it does not. The card never fabricates a match to look useful.