Back to work
BlueBox Sales Manager
AI Sales Manager Assistant
A review-first coaching system that connects call activity, a business-specific rubric, CRM outcomes, and manager corrections.
OpenAISalesforceSales CoachingHuman Review
Focused tests
37
Passing tests across the safe case-study package.
My role
Workflow design, integration, testing, and verification — built with AI as a development tool
System state
Live internal system
What I built
A review-first coaching loop connecting call activity, a visible rubric, CRM outcomes, and manager corrections.
Problem
- Calls, texts, recordings, and CRM outcomes lived in separate places.
- Managers needed clear coaching priorities without handing decision-making entirely to AI.
- Customer recordings and corrections could not be exposed in a public portfolio.
What I Built
- OpenPhone/Quo webhook intake
- Business-specific call scoring
- Salesforce outcome matching
- Manager correction controls
- Daily and weekly reports
- Calibration workflow
Solution
- Connected call intake, a business-specific eight-checkpoint rubric, Salesforce matching, and reviewed manager corrections.
- Added rule-based fallback behavior so the workflow stays useful when AI scoring is unavailable.
- Built daily and weekly reports plus a calibration loop around approved corrections.
Business Impact
- Turns scattered call activity into a focused coaching queue.
- Keeps manager judgment visible and reversible instead of hiding it behind an AI score.
- Creates a repeatable path from real sales conversations to reviewed team learning.
Lessons
- AI coaching is more trustworthy when the rubric and manager override are both visible.
- The useful system is the full feedback loop, not the score by itself.