AI-powered lead enrichment and routing

DTC Newsletter signups were reaching the shared Pilothouse HubSpot instance with basic contact information but little company context, qualification data, or indication of which leads deserved immediate attention.

The CFO wanted to enrich those records automatically. Before building anything, I spoke with the stakeholders involved to understand what information would help Sales. Enrichment was useful, but it wasn’t the end goal. The larger need was faster qualification and real-time visibility when someone from a promising company signed up.

The opportunity was to turn each signup into a structured, sales-ready record and surface the strongest leads while their interest was still fresh.

I designed the workflow around speed-to-contact rather than enrichment volume. That changed what the system needed to produce: not simply more data, but a reliable record that helped Sales understand who had signed up, where they worked, how well they fit, and whether the lead warranted immediate attention.

The system needed to filter, research, classify, validate, populate the CRM, and alert the right people in one connected flow.

Lead-data pre-population rate was the clearest measure of whether the record was useful. By the time Sales opened a new lead, most of the research and qualification context should already be there.

I also tracked the number of signups processed and enriched leads produced each month to make sure the workflow remained useful at production volume, not only in testing.

I began by defining what a sales-ready HubSpot record needed to contain, then mapped the research, classification, validation, and routing steps required to produce it.

LindyAI orchestrated the workflow, detected new signups, and filtered out generic email domains. StoreLeads supplied company-level data, while LinkedIn and web sources supported contact research. Clay was added in a later iteration to improve the depth and accuracy of company and contact enrichment. The ChatGPT API assessed company fit and persona, and returned the information in a defined structure before the validated record was written to HubSpot. When a high-value lead appeared, Slack alerted Sales and leadership in real time.

The workflow went through several iterations. Early versions produced inconsistent classifications on edge cases, so I tightened the prompts, required structured outputs, and added validation before anything could populate the CRM. Once the system was reliable, we adapted the same workflow pattern for other Pilothouse B2B client accounts.