AI-powered sales coaching platform dashboard showing performance analytics and video coaching interface for field sales teams
Case Study: AI & Cloud Platform

Fragmented Performance Data Transformed Into Automated Coaching at Scale

Unified data from four systems. AI-powered personalized coaching. 104,000 coaching videos annually.

Company Overview

The client is a national residential flooring retailer founded in 2007, specializing in direct-to-consumer shop-at-home flooring sales and fast installation services. Instead of traditional retail stores, they bring material samples directly to customers' homes for consultations and often complete installations in a single day.

Operating across multiple U.S. states including Georgia, Texas, Florida, North Carolina, and Tennessee, the company serves the residential repair and remodel markets with approximately 2,000 field sales representatives conducting in-home consultations across multiple branches. They sell and install carpet, hardwood, luxury vinyl, laminate, and tile through this large distributed sales organization.

Despite having performance data across four disconnected systems—lead scoring, virtual ride-alongs, presentation timings, and job outcomes—the organization had no unified view and no systematic mechanism to identify what high-performing reps did differently or deliver data-driven coaching at scale.

description

Executive Summary

JBS Dev built a cloud-native AI-powered sales coaching platform on AWS that unified performance data from four disconnected systems and generated personalized video-based coaching delivered to every sales rep on a recurring cadence—scaling data-driven feedback to 2,000 field reps without requiring proportional increase in management effort.

info Without a unified view of sales performance and no systematic coaching mechanism, the organization could not scale what good looked like, could not systematically improve win rates, and relied on manual manager effort across thousands of reps.

Business Challenge

Performance data fragmented across four disconnected systems with no practical way to identify what high performers did differently or deliver personalized coaching at scale.

Fragmented Performance Data:

  • Four disconnected systems with no integration
  • No unified view across lead scoring, virtual ride-alongs, presentation timing, and outcomes
  • No systematic mechanism to identify high-performer patterns
  • Limited real-time visibility into rep-level trends

Manual Coaching Limitations:

  • Inconsistent performance delivery across branches
  • No structured, data-driven feedback loop
  • Manual effort required that could not scale
  • No historical baseline for performance comparison

Operational Scalability:

  • Manual coaching could not scale to 2,000 reps
  • No standardization around high-value sales stages
  • Institutional knowledge trapped in managers' heads
  • Could not systematically improve win rates at scale

JBS Dev's Approach

JBS Dev designed, built, and deployed a production-ready cloud-native Sales Coaching Engine on AWS that ingests data from four sources, stores it in a unified database, uses large language models to generate personalized coaching insights, and delivers video-based coaching to every rep on a configured cadence.

Data Pipeline and Unification

  • Built scheduled ETL pipeline ingesting data from the client's existing data lake
  • Integrated four data domains: Think Unlimited lead scoring, Rilla Virtual Ride Along section scores, iPad presentation slide timing data, and lead/job outcome data
  • Loaded up to 5 years of historical sales and outcome data to establish performance baselines for trend analysis and rep benchmarking
  • Transformed and loaded data into dedicated PostgreSQL coaching database optimized for coaching analysis
  • Created unified schema linking rep identity, presentation quality scores, slide timing behavior, lead quality, and outcomes
  • Validated referential integrity across data sources

AI-Powered Coaching Engine

  • Implemented LLM-based analysis engine evaluating each sales rep's recent performance across all data dimensions
  • Generated personalized coaching insights per sales rep based on unified multi-source data analysis
  • Integrated HeyGen video generation for automated video-based coaching delivery using AI avatars
  • Delivered coaching via video format on configurable cadence (daily or weekly)
  • Coaching videos reference real rep-specific metrics drawn from at least one of the four data domains
  • Enabled systematic identification of what high-performing sales reps do differently

Management Interfaces and Scalability

  • Built sales rep view for accessing personalized coaching videos
  • Built sales manager view for accessing their team's coaching videos
  • Automated video-based coaching generation targeting 2,000 sales reps with weekly coaching cadence
  • Infrastructure designed to generate approximately 104,000 AI-powered coaching videos annually
  • End-to-end smoke tests validated: data ingestion → coaching video generation → video delivery → rep/manager interface display

The platform leverages AWS infrastructure, PostgreSQL database, ETL pipeline infrastructure, LLM-based coaching engine, HeyGen video generation integration, and web application hosting to deliver a cloud-native platform designed for production reliability and operational scale. Phase 2 is planned to replace third-party sales call analysis tools and expand avatar configuration capabilities.

Cloud-native sales coaching platform architecture showing AWS infrastructure, PostgreSQL database, ETL pipeline ingesting data from four sources, LLM-based coaching engine, HeyGen video generation, and sales rep/manager interfaces

AI-powered sales coaching platform unifies four data sources and delivers personalized video coaching to 2,000 field sales reps

Business Outcomes

Before: Fragmented Manual Coaching

  • close
    Performance data existed across four disconnected systems with no unified view
  • close
    No systematic mechanism to identify what high-performing reps did differently
  • close
    Manual effort required to generate individual performance feedback
  • close
    Managers lacked real-time visibility into rep-level trends and team benchmarks
  • close
    No structured, data-driven feedback loop to help reps continuously improve
  • close
    Inconsistent sales delivery across branches with no standardization
  • close
    Years of historical data sitting unused with no practical way to extract insights
  • After: AI-Powered Automated Coaching

  • check_circle
    Unified data platform consolidated performance data from four previously disconnected systems into single coaching platform
  • check_circle
    AI-powered pattern recognition systematically identified what high-performing sales reps do differently
  • check_circle
    Automated coaching at scale delivered personalized video-based coaching to 2,000 reps with no manual generation effort
  • check_circle
    Real-time visibility enabled through sales rep and sales manager interfaces for accessing coaching videos
  • check_circle
    Data-driven feedback loop helping reps continuously improve with coaching videos referencing real rep-specific metrics
  • check_circle
    Standardized sales delivery focused on high-value stages of the sales pitch identified through data analysis
  • check_circle
    104,000 annual coaching videos generated through platform designed for approximately 2,000 active reps with weekly cadence
  • verified_user Scaling Coaching Without Scaling Management: 5-year historical data backfill enabled baseline benchmarking and trend analysis, while automated video delivery ensured consistent coaching cadence without manager intervention.

    What Made This Work

    integration_instructions

    Data Unification Across Fragmented Systems

    JBS Dev unified four disconnected data sources (lead scoring, virtual ride-along evaluations, presentation timing, and outcomes) into a single coaching platform with referential integrity, 5-year historical baseline data, and a unified schema linking rep identity, behavior, and results.

    psychology

    AI at Scale for Field Sales Performance

    The platform uses large language models to analyze unified performance data and generate personalized feedback—evaluating each rep's recent performance across multiple data dimensions, generating personalized coaching insights, and integrating HeyGen for AI avatar-based video delivery at enterprise scale.

    cloud

    Cloud-Native AWS Architecture for Operational Reliability

    Built on AWS infrastructure designed for production reliability: scheduled ETL pipelines, scalable database operations, HeyGen video generation orchestration, and web application hosting for sales rep and sales manager interfaces. This is production infrastructure supporting coaching delivery for thousands of field sales reps, not a prototype.

    speed

    Automated Video Coaching Without Manual Effort

    Every active sales rep (up to 2,000) receives personalized video-based coaching on the configured cadence (weekly by default), with no manual generation effort from management. Automated video delivery freed managers from repetitive performance review tasks to focus on high-value coaching conversations.

    insights

    Historical Data Baseline for Pattern Recognition

    Loading 5 years of historical sales and outcome data established performance baselines for trend analysis and rep benchmarking. The AI engine answers "What do top reps do differently?" by correlating behaviors across lead quality, presentation timing, virtual ride-along scores, and successful outcomes.

    groups

    Field Sales Operations Understanding

    Deep understanding of distributed field sales operations, in-home consultation workflows, multi-branch operations, presentation delivery, and the operational realities of managing large field sales teams informed the platform's architecture, coaching logic, and scalability requirements.