Embedded Files
AI-Augmented Design & Development:
Diagnosing and Redesigning a Remote Patient Monitoring Call Center Agent Onboarding Curriculum
The Problem:
A remote patient monitoring (RPM) training program was producing a steady stream of downstream errors — chart corrections, coaching interventions, and early trainee attrition — despite a comprehensive job aid already existing. The job aid ("the QuickBook") documented clinical protocols, escalation procedures, and operational workflows in detail. The problem wasn't missing content. It was that trainees didn't know where to look and the training itself never built the habit of self-service lookup.
A sample exchange from the team's Slack channel captures it:
A trainee asked where to find unenrollment documentation. The trainer's reply: "Please bookmark this — it is linked in the QuickBook," followed by a link to a separate doc.
This pattern repeated across escalation protocols, alert-clearing procedures, and edge-case documentation: the canonical answer existed, but nothing in the training built the muscle of finding it independently.
The Process:
Rather than starting from assumptions about what training needed fixing, I used Claude — connected to our Slack workspace and Google Drive — to build an evidence-based gap analysis directly from real operational data:
Pulled a month of chart-correction and coaching signal from a private leadership Slack channel, extracting 71 chart-correction/send-back instances and 16 distinct coaching interventions over a 30-day window.
Categorized errors inductively into six recurring themes (time allocation, template usage, documentation quality, escalation compliance, alert-clearing, and process confusion), then mapped each theme against the existing curriculum to classify it as a coverage gap (content existed but didn't transfer) versus a content gap (no training existed at all).
Ran a second, targeted pass specifically searching for QuickBook-reference friction across three related teams' Slack channels, surfacing seven additional navigation and documentation gaps.
Cross-referenced every finding against the job aid itself, marking each as Documented or Partial, and assigned a priority and recommended fix type (SME content update, process clarification, eLearning module, scenario-based simulation, job aid, or vILT).
The Findings:
The analysis produced ten prioritized findings, split roughly into three categories:
Process gaps — escalation protocol failures, alert-clearing errors, and concurrent message management, where the fix was procedural clarity, not more content.
Content that existed but didn't transfer — template selection errors and documentation quality issues, where trainees had been exposed to the material but hadn't built fluency.
Genuine content gaps — specialty clinic escalation rules and a rarely-used protocol, which had never been formally addressed.
The highest-frequency findings (18+ instances of template selection errors, 16+ escalation protocol failures) were flagged HIGH priority and prioritized for both eLearning coverage and scenario-based practice reps.
The Redesign:
The findings drove concrete structural decisions in the rebuilt 20-day blended curriculum:
AI-avatar scenario practice (built in TaleSpin) was sequenced to land before live trainee calls began — a direct response to the finding that documentation and escalation errors stemmed from lack of practice reps, not lack of content exposure.
A three-tier lookup hierarchy (AI knowledge assistant → job aid → trainer) was woven into the curriculum at five separate touchpoints, explicitly training the self-service habit the gap analysis showed was missing. By the second week, trainers were coached to respond to lookup questions with "what did the assistant say?" rather than answering directly.
A gamified navigation activity was built specifically to drill the job-aid lookup gaps the targeted Slack search had surfaced (there is a mock up example below).
Quick-reference job aids covered the net-new content gaps that didn't warrant a full eLearning module (there is a mock up example below that would be deployed in Google Apps Scripts to learners).
The full curriculum — vILT storyboards, facilitator guides, slide decks, LMS scaffolding, and trainer coordination — was designed, built, and launched end-to-end once the redesign requirements were finalized.
Deliverable Samples:
Validation and Next Steps
The redesign was SME reviewed before launch and then cross-checked against an independent class audit conducted separately by a training team lead, and validated on launch by leadership. Ongoing daily trainer debriefs and direct session observation were built into the rollout to catch friction early rather than waiting for a formal review cycle.
If you're building something where accuracy and speed both matter, let's connect.
Page updated
Google Sites
Report abuse