EIKO CRM

Overview

A CRM built from scratch with AI at the core, not bolted on. It auto-syncs Gmail, Google Calendar, and Google Drive, then uses Claude to classify email relevance, match contacts, and surface what needs attention.

The Challenge

Generic CRMs are either too complex (Salesforce) or too simple (spreadsheets). A small consultancy needs something in between: automatic activity capture from existing tools, intelligent prioritization, and zero manual data entry. We built what we actually wanted to use.

How It Works

OAuth integration syncs Gmail, Calendar, and Drive automatically

Claude Haiku classifies every email for business relevance (conservative: uncertain items go to a review queue)

Multi-step contact matching: exact email match (100% confidence) > domain match (80%) > AI fuzzy matching. Uncertain matches are queued for human review.

Kanban pipeline board for deal tracking (lead > contacted > discovery > proposal > closed)

Company profiles auto-link to Google Drive folders via fuzzy matching

Activity feed aggregates emails, meetings, and file changes per contact/company

Tech Stack

Lessons Learned

• AI classification needs a human escape hatch. Binary "relevant/not relevant" classifiers will always have edge cases. Building a review queue for uncertain items (rather than forcing a decision) kept the system trustworthy without sacrificing automation.
• Multi-step matching beats single-model matching. Running cheap deterministic checks first (exact email, domain) before calling the AI model reduced API costs by ~80% while maintaining accuracy. Only genuinely ambiguous cases hit the LLM.

Status & What's Next

New. Core features built: pipeline, contacts, companies, activities, Google sync, AI classification.

Dashboard refinements, automated follow-up reminders, deeper Drive integration.