The problem
A B2B software company generating 300–500 inbound leads per month had a painful bottleneck: their two SDRs were spending 60% of their time on first-pass qualification — reading websites, checking LinkedIn, scoring against ICP criteria, and writing the first personalised email. High-value leads were sitting for 2–3 days before first contact.
The solution — what we built
1
Lead ingestion pipeline (n8n)
n8n workflow triggered on new CRM entries from any source — HubSpot form, LinkedIn ad, Typeform, or direct API. Lead data normalised and pushed to the enrichment queue.
2
Data enrichment layer
Apollo.io and Clearbit APIs pull company size, revenue, tech stack, funding stage, and LinkedIn headcount. The company website is scraped and summarised by an LLM. All enriched data stored in PostgreSQL.
3
LLM ICP scoring engine
A structured LLM prompt evaluates each lead against the client's ICP criteria: company size, vertical, tech stack, pain signals, funding stage. Returns a 0–100 score with written reasoning. Prompts tuned iteratively with the sales team over the first two weeks.
4
Personalised outreach drafting
For leads scoring 70+, the LLM drafts a personalised first email referencing the lead's specific company context — their tech stack, recent funding, or a pain point inferred from their website. SDRs review and send with one click.
5
Next.js dashboard
A clean web dashboard shows the scored lead queue, enrichment data, AI reasoning, and draft email. SDRs work from a single interface — no tab-switching between CRM, LinkedIn, and email. Full-text search, filters by score/vertical/stage, and one-click CRM sync.
Our SDRs went from spending 3 hours on 20 leads to processing 80 leads in that same time. The quality of outreach improved because every email has a specific hook. And the leads that were previously falling through the cracks are now getting responded to same day.
— VP Sales, B2B SaaS Company, US
