AI-Powered CRM Agent
An AI agent that plans, approves, executes, and audits Pipedrive workflow changes instantly via Slack prompts.
My Role
Challenge
Managing CRM configurations required manual intervention for every field addition, pipeline update, or permission modification. This created significant friction for the Revenue Operations workflow.
Bottlenecked Operations: Ops personnel became the sole gatekeepers and executioners for simple system updates.
Lack of Governance: Configuration changes were frequently undocumented or executed without formal approval tracking.
Fragmented Workflows: Sales managers and operations staff had to leave their primary communication hubs to navigate complex CRM admin panels.
Information Silos: No centralized audit log existed to track the history of system modifications or data exports in real time.
Results
I built Pipefie, an AI-driven Slack agent deployed on Render that acts as an approval-gated middleware between the team and the CRM infrastructure.
Conversational UI: Users request structural CRM changes or data exports using plain English within standard Slack channels.
Intelligent Parsing: Claude AI evaluates requests against a live Pipedrive snapshot to generate structured JSON execution plans.
Secure Execution: Designated administrators can approve or reject the AI generated plan with a single click via Slack Block Kit buttons.
Automated Auditing: Every executed change automatically logs the exact timestamp, original requester, specific action, and authorized approver to a secure Confluence database.
< 1 Min
Change Turnaround
0
Unauthorized Changes
10+
Hours Saved Weekly
Process
Discovery & Architecture: Mapped the administrative bottleneck to design a secure Node.js application utilizing Slack Socket Mode for real-time responsiveness behind firewalls.
Prompt Engineering: Crafted strict system instructions for Claude to accurately map fuzzy human language to exact Pipedrive entity IDs and stages.
API Orchestration: Built robust REST clients to seamlessly interact with Pipedrive endpoints for fields, user permissions, pipelines, and CSV data exports.
Cloud Deployment: Configured the application environment on Render to ensure high availability and continuous uptime.
Access Control & Logging: Engineered a strict authorization gate evaluating Slack user IDs and routed all execution logs to a centralized Confluence workspace.
