Pilothouse / DTC Newsletter / Senior Website and Email Operations Manager / 2021 – 2024
AI-powered lead enrichment and routing
The CFO wanted to enrich leads. The real problem was that sales could not tell which signups were worth calling the moment they came in.
The read
Newsletter signups were flowing into HubSpot as raw contact records. A company name and an email address. No company context, no qualification signal, no urgency indicator for sales.
The request that came in was simple: enrich leads automatically when they sign up. On the surface it looked like a data problem. I sat with the CFO and sales team to understand what they actually needed. The real problem was not missing data. It was that by the time sales manually researched a lead and decided whether to reach out, the moment had passed.
The read: this was a speed-to-contact problem, not an enrichment problem. The system needed to tell sales who was worth calling and do it in real time, not after a manual research pass.
The decision
Reframe the goal from lead enrichment to speed-to-contact infrastructure. That changed everything downstream - the tools selected, the routing logic, the definition of done.
The system needed to filter, enrich, qualify, and surface high-value leads to sales the moment they signed up. Not in a daily batch. The moment they came in.
The metric I chose
Lead data pre-population rate. What percentage of fields did sales have before they touched a record? If the system was working, a salesperson should be able to open a new lead and already know who they were looking at.
Secondary: enriched lead volume per month. The system had to perform at scale, not just in demo conditions.
The build
Mapped the full workflow before selecting any tools. Defined exactly what a sales-ready record needed to contain, then built backward from that output.
The workflow ran in sequence: Zapier caught the new signup, filtered out personal email domains immediately, pulled company-level data through StoreLeads, enriched individual contact data using LindyAI and LinkedIn, passed both to ChatGPT for company fit classification and persona identification, structured the output into a CRM-ready record format, wrote it to HubSpot, and routed a Slack alert to sales and leadership when a high-value company came through.
Iterated on LLM behavior across multiple versions. Early versions produced inconsistent classification on edge cases. Added structured output requirements between steps and a validation layer to catch unreliable data before it reached CRM fields. The system only became reliable once outputs were forced into a defined schema rather than freeform text.
Extended the same workflow logic to Pilothouse B2B client accounts after validating it on DTC signups.
~70%
Lead data pre-populated before sales contact
1K–2K
Enriched leads generated per month
+30%
Lead quality improvement vs. unenriched baseline
The reframe was the whole thing. Enrichment as a goal produces a richer spreadsheet. Speed-to-contact as a goal produces a system sales actually uses. The difference was not in the tools. It was in asking what problem the data was actually supposed to solve.
Tech
Zapier, workflow orchestration and trigger logic
LindyAI, individual contact enrichment
ChatGPT API, company classification and fit scoring
StoreLeads, company-level data by domain
HubSpot, CRM destination and pipeline management
Slack, real-time high-value lead routing
Miguel N. Monzones
Vancouver, BC, Canada
