How Ecogen Services made quotes 5x faster with an AI-assisted quoting platform
A solar sales team closing deals in the field with full quoting power and zero access to the margins behind every number.
The quoting system that had a 500-customer expiry date
Ecogen Services is a residential solar installation company, selling and installing solar systems for homeowners. Sales reps quote jobs directly at the customer's home, based on site details and engineering drawings.
Ecogen's founder built the quoting system on Google Sheets. It worked, but only for a while.
Reps were spending 15 to 20 minutes per site visit manually reading engineering drawings before they could quote anything. Every pricing update needed the founder personally. Every new customer added another tab to a sheet already close to breaking.
The spreadsheet also had a hard ceiling. Past 500 customers, the whole thing was projected to degrade. Cross-rep reporting was nearly impossible. And the pricing formulas, margins, and commission splits were all sitting in a sheet that any rep with basic spreadsheet skills could get into.
The founder knew it needed replacing. The question was what to replace it with.
A system built around how solar sales actually work
Agent Loopr rebuilt Ecogen's quoting infrastructure from the ground up as a secure, full-stack SaaS web application. The entire stack, backend and frontend, was built using Claude Code. Every architectural decision was made to separate what reps interact with from what drives the numbers behind their quotes.
Server-side pricing engine
Every pricing formula, equipment margin, install redline, and commission split was rebuilt inside a Python FastAPI backend. We developed the pricing model from the client's requirements and industry practices for deriving financing quotes. Claude analyzed 202 historical financing quotes and found the pricing wasn't price-per-watt-based at all. That figure was an output, not an input. The real mechanism is a 25-year power-purchase present-value calculation; against 163 complete quotes the model reached 94% explained variance with a median error under 5%. The web client transmits site inputs only. It receives back exactly what each role is permitted to see. The pricing logic itself never reaches the browser.
The browser never sees a formula, margin, or commission split.
Three-tier role-based access control
Role-based access control returns three distinct data shapes from the same quote endpoint. Reps see commissions. Managers see pipelines. Owners see everything, including equipment costs, labor splits, and net profit.
Reps, managers, and owners each see only what their role permits.
Versioned price book
The price book was version-controlled inside a PostgreSQL schema. Every time an admin updates an adder cost, the change applies only to new quotes going forward. Every historical quote locks in the exact price snapshot active at the moment it was created.
Adder updates apply forward only; historical quotes never shift.
AI Planset extraction
A rep uploads an engineering planset PDF in the field. Claude (claude-sonnet-4-6, vision) reads the drawing and returns structured data, including module model, wattage, panel count, array configuration, roof pitch, and azimuth. The quote form pre-fills automatically. On a multi-array reference planset with mixed orientation, Claude scored 100% on core spec fields.
On a multi-array reference planset with mixed orientation.
Monday.com CRM integration
Once a quote is finalized, a single click sends the complete data set directly to Ecogen's Monday.com board. This integration allows the quoting platform and CRM to function as a unified system, eliminating the need for reps to double-enter information.
Finalized quotes flow straight onto the Monday.com board.
Full-stack cloud deployment
The complete application was deployed with continuous deployment across Vercel (frontend) and Railway (backend and managed PostgreSQL). The codebase features documented architecture, enabling future developers to contribute without needing context from the original build team.
Vercel + Railway · Documented architecture, zero-downtime deploys.
What did replacing the spreadsheet actually deliver for Ecogen Services?
5x faster quotes
Quote preparation time dropped from 15–20 minutes to under 4 minutes per site visit. Sales reps now spend more time engaging with homeowners, rather than reading engineering plans during meetings.
88% reduction in calculation time
Latency dropped from 2,500ms to under 300ms. The pricing engine now returns accurate, role-appropriate figures in under a third of a second, fast enough that it never interrupts a sales conversation.
100% of pricing locked server-side
Every formula, margin, and commission split lives in the backend. No rep can access the numbers behind their quotes, regardless of their spreadsheet skills.
Unlimited customer scale
The previous 500-customer limit is eliminated. The platform's relational database offers unrestricted scalability, robust cross-rep reporting, and full historical data access.
Our reps can quote anything in the field now, and I'm still the only one who ever sees the margins. That was the whole point and Shivam’s team ensured that. I really appreciated their iterative feedback loop every week as it helped me stay fully involved in the process even as a non tech person
Tools and technologies
used on this project
Frontend & UI
A fast, role-aware quoting interface reps use in the field.
Backend & Data
Server-side pricing, a versioned price book, and vision extraction.
Integrations & Deploy
CRM sync and continuous deployment as one flow.
One coherent stack: pricing, access, and CRM, secure and production-ready at launch.