Skip to main content
← All posts
Case Study11 min readFeb 2026

From 0 to 10 Cr ARR: The AI Systems Behind Onsite's Growth

4 LangGraph agents serving 20 sales reps. Daily pipeline scoring of 500 leads. Facebook Ads optimization. The AI infrastructure that scaled a construction SaaS to 10 Cr ARR.

Sales IntelligenceLangGraphCRMFacebook AdsB2B SaaS
D

Dhruv Tomar

AI Solutions Architect

Tech Stack

PythonLangGraphZoho CRMFacebook Ads APIMCPPostgreSQL

Architecture

7 departments (Sales, Marketing, Support, Analytics, Operations, Finance, Product) -> Each with dedicated CONTEXT.md + SKILL.md -> MCP servers for Zoho CRM + Facebook Ads -> LangGraph agents for pipeline scoring, lead routing, and performance analysis.
0 to 10 Cr ARR contribution
20 sales reps served
500 leads scored daily
31% reply rate vs 8% industry average

When I joined Onsite, they had a CRM full of unscored leads, sales reps manually checking pipelines, and no systematic way to prioritize deals. Four years later, AI systems touch every part of their sales operation.

The Agent Architecture: 4 LangGraph agents run daily: Pipeline Scorer analyzes 500 leads against historical win patterns. Lead Router assigns new leads based on rep capacity and specialization. Performance Analyzer generates weekly insights per rep. Campaign Optimizer adjusts Facebook Ads spend based on lead quality signals.

The 7-Department System: Each department (Sales, Marketing, Support, Analytics, Operations, Finance, Product) has its own CONTEXT.md and SKILL.md — structured knowledge that any AI agent can consume. Two MCP servers connect directly to Zoho CRM and Facebook Ads, giving agents real-time data access.

The 31% Reply Rate: Industry average for cold outreach is 8%. Our AI-crafted sequences hit 31%. The secret isn't better copy — it's better targeting. The Qualifier agent rejects 60% of leads before outreach begins. When you only email people who actually need your product, response rates go up dramatically.

Key Lesson: AI didn't replace the sales team. It removed the information asymmetry. Reps now walk into every call knowing the lead's score, their likely objections, and the best case study to reference. That's the real leverage.

Want to build something like this?

I architect and deploy end-to-end AI systems — from MVP to revenue.

Let's Talk