Sales research assistant for a SaaS startup
A Series-A startup needed a research assistant that could pre-brief reps before discovery calls, drawing only from publicly available data and the company's own CRM notes.
Reps were spending forty-five minutes per discovery call assembling context from LinkedIn, the prospect's website, the CRM, and old Slack threads. The founders wanted the prep automated, but draft notes about prospects in a hosted LLM were a deal-breaker for their privacy posture.
Llama 3.1 70B on a GPU-equipped server, fronted by an n8n workflow that takes a calendar invite, pulls public web data, queries the CRM API, and stitches the result into a one-page brief in Notion. The same workflow records the brief back into the CRM so the next rep on the account inherits the context.
Per-call prep time dropped from forty-five minutes to under five. Reps reported richer context in briefs than they used to assemble manually, and the founders kept the CRM data on their own infrastructure.
