Sales Research Agent
An agent researches each new lead with a real browser, scores fit against your ICP, and drafts a personalized outreach sequence for rep review.
The problem
Reps spend the start of every deal cycle on manual research: company sites, news, and scattered public sources, with notes pasted into the CRM by hand. The research is inconsistent across reps and often skipped under quota pressure. Outreach ends up generic, and strong-fit leads get the same treatment as weak ones.
How it runs on Neblex
When a lead is created in Salesforce, a flow hands it to the research agent. The agent uses browser automation to visit the company website and other public pages that have no API, reads what it finds, and scores the lead against your ICP criteria. The score, the supporting evidence, and the research notes are written back to the lead record through the Salesforce connector.
The agent then drafts a personalized outreach sequence as Gmail drafts, not sent mail. Each draft goes to the rep through human-in-the-loop approval: the rep sees the reasoning, the evidence, and the proposed message, and approves or rejects. Nothing reaches a prospect without a human decision.
The pipeline around the judgment step is deterministic: webhooks trigger it, retries handle transient failures, and replay resumes from the exact failed step. Every action the agent takes, from browser steps to CRM writes, is recorded in the tamper-evident hash-chained audit trail.
Step by step
Lead triggers research
A new Salesforce lead fires a webhook that starts the flow.
Agent browses public sources
Browser automation visits the company site and other pages without APIs to gather context.
Score against your ICP
The agent scores fit using your criteria and writes the score and evidence to Salesforce.
Draft the outreach sequence
Personalized emails are created as Gmail drafts tied to the research notes.
Rep reviews and approves
The rep approves or rejects each draft from the task inbox before anything is sent.
Platform capabilities used
- Browser automation for pages without APIs
- Salesforce and Gmail connectors as tools
- Human-in-the-loop approval before outreach
- Agent as a step in a deterministic flow
- Hash-chained audit trail
Common questions
Will the agent email prospects on its own?
Not unless you remove the checkpoint. Outreach drafts route through human-in-the-loop approval, so a rep sees the reasoning and the message before anything sends, and every action is logged to the audit trail.
How does the agent research companies that have no API?
It drives a real browser. Browser automation lets the agent open public web pages, read them, and extract what it needs, the way a human researcher would.
Can we keep lead data on infrastructure we control?
Yes. You can point the agent at any OpenAI-compatible endpoint, including self-hosted models on your own infrastructure, or Anthropic. On-Prem Workers can also run the flow inside your network over an outbound-only connection.
Related use cases
Lead Routing and Enrichment
Enrich new Typeform submissions with Clearbit firmographics, score them, and route each lead to the right Salesforce owner within minutes.
Sales & RevOpsQuote-to-Cash Sync
Move closed-won opportunities from Salesforce through Neblex CPQ to NetSuite billing automatically, with no rekeying between systems.
MarketingCampaign-to-Pipeline Attribution
Sync Google Ads spend and HubSpot campaign engagement into BigQuery on a schedule so marketing and sales report from the same pipeline numbers.
Want this running on your stack?
Neblex Integration Fabric is in beta: full platform, free while in beta. Bring this workflow and we will map it to your systems.