DavidGadyan

AI ERP & SaaS · AI & Agents · AI Marketing & Analytics · Databases & Data Engineering

AI Lead Scoring & Churn Detection Engine (S1mple.ai)

By David Gadyan · Published · Updated

Next.js dashboard showing AI lead scores and churn risk by account

AI engine that scores leads by repeat-conversion likelihood and flags churn risk inside S1mple.ai, with a Next.js dashboard, a FastAPI and LangChain back end, and MCP connectors into HubSpot, Monday, Zoho and four databases, scaled on Kubernetes.

The problem

Marketing budgets get spread evenly across a CRM because nobody can say which accounts deserve attention. The client wanted every contact scored for repeat-conversion likelihood and churn risk, using the data already sitting in their CRM and database, so campaigns could concentrate on the few accounts where effort actually changes the outcome.

What I built

  • A Next.js dashboard that surfaces scores, churn flags and the reasoning behind each one.
  • A FastAPI back end that orchestrates scoring runs and serves results to the dashboard and back to the CRM.
  • LangChain-based scoring and churn classification, with MCP servers as the integration layer so the agent reads and writes through tools instead of bespoke glue code.
  • CRM connectors for HubSpot, Monday and Zoho.
  • Database connectors for PostgreSQL, MySQL, Supabase and MongoDB, so the engine runs on whatever the client already has.
  • Container orchestration on Kubernetes for scaling and management.

Outcome

Marketing effort concentrated on high-propensity and at-risk accounts, one scoring engine that is independent of the CRM and of the database, and a product module that now lives inside the S1mple.ai platform.

Related projects

← All projects