All projects
AI052025

AI CRM Platform

Lead scoring, sentiment analysis, and automated follow-ups in a modern CRM.

Full Stack Developer 3 months 3 engineers 2025

Scored

every inbound lead ranked by predicted fit

Sentiment

conversation analysis feeding follow-up priority

Dockerized

reproducible deploys across environments

Tech Stack

Next.jsNode.jsPostgreSQLOpenAIDocker

The Challenge

Sales teams drown in unqualified leads and stale follow-ups. The goal was a CRM that ranks attention for you: scoring leads from behaviour and conversation quality, then recommending the next action — without becoming a black box the team refuses to trust.

Approach

  1. 1

    Built lead scoring from explicit signals (source, engagement, deal stage) blended with an LLM pass over conversation history for intent and sentiment.

  2. 2

    Surfaced the why: every score rendered its contributing factors, so reps could disagree with the model instead of ignoring it.

  3. 3

    Automated follow-up recommendations as drafts tied to sentiment — a frustrated lead gets a different tone than a warm one.

  4. 4

    Containerised services with Docker so API, worker, and database environments matched from laptop to production.

Technical Highlights

Explainable scores

Scores are stored with their factor breakdown, making regressions debuggable and giving reps a reason to trust (or override) the ranking.

Async AI pipeline

Sentiment and scoring run in background workers against PostgreSQL state, keeping the CRUD UI fast while AI enrichment catches up.

What I Learned

AI features get adopted when they explain themselves; opaque rankings get worked around.
Separating the interactive path from the enrichment path is the difference between a snappy CRM and a sluggish one.

More Case Studies

Online now

Usually replies within an hour

Chat on WhatsApp1