← All case studies
Enterprise Learning

The Assessment Debrief System

We took multi-hour advisor debriefs entirely off the calendar — replacing them with personalized AI-avatar reports that generate, review and deliver themselves at scale.

ClientDeliberate Breakthroughs
SectorWorkplace Assessment
Year2025
What we didAI Avatar Pipeline · Workflow Automation · Secure Delivery
Admin Dashboard
1,000+
Personalized debriefs
18 min
Unique video / person
12
AI avatars
Zero
Manual production
Overview

Deliberate Breakthroughs runs workplace personality assessments, then delivers the results in long, one-to-one advisor calls. The insight was valuable — but it lived inside an advisor's calendar, and it didn't scale.

We proposed taking the debrief off the advisor's plate entirely. We built an end-to-end platform that ingests each assessment, writes a personalized script, generates up to 18 minutes of unique AI-avatar video, routes it through a two-stage human review, and delivers it to the candidate on a secure microsite — with zero manual effort from submission to delivery.

The challenge

One-to-one debriefs are excellent and completely unscalable. When 250+ people need feedback at once, the model breaks.

Scale limits

Personal debriefs can't happen simultaneously for hundreds of employees.

Inconsistency

Quality varied with whichever facilitator happened to deliver the session.

Scheduling cost

Booking busy professionals into hour-long calls was expensive and slow.

Fixed depth

A single session never let people dive deeper where it mattered most to them.

The solution

An end-to-end automated debrief platform — personalized video, an intelligent admin workflow, and secure candidate delivery.

  • Webhook-driven ingestion matches each submission to a report template and scaffolds every section instantly — no data entry.
  • Template engine defines structure, default avatars, voices and captions, with a smart merge that preserves customizations.
  • HeyGen avatars deliver personalized content per result, narrated with ElevenLabs voice cloning.
  • Two-stage human review: admins approve scripts first, then verify rendered video before anything publishes.
  • Password-protected candidate microsites backed by signed URLs for private, time-limited delivery.
Core features

What we built, in detail.

01

Assessment-triggered automation

Submit → template match → generate → two-stage QA → deliver. The whole pipeline runs untouched when auto-submit is on.

02

Multi-avatar video generation

12 unique avatars deliver up to 18 minutes of personalized content, staged into batches with per-item retry for full cost and quality control.

03

Multi-module candidate experience

A guided, video-first debrief — Welcome, Drives Deep Dive, Mindset Under Pressure, Final Words — with on-page captions and progressive navigation.

04

Admin dashboard & templates

Full report lifecycle tracking, client-grouped branding, section-level template defaults, and live render-batch monitoring.

05

Bidirectional integration

Incoming webhooks from Zapier/Airtable create reports; outgoing webhooks sync published URLs back — closing the loop with existing workflows.

06

Engagement analytics

Play, pause, completion, scroll depth and device tracked per section — without ever exposing individual results.

Interface gallery

Inside the product.

Admin DashboardViews, status distribution, recent activity and top-viewed candidates
Candidates ListGroup, cohort and client filtering with per-report action controls
Render BatchesLive render queue with auto-submit toggle and per-module status
Report TemplatesSection layouts, default avatars, voices and intro/closing videos
Candidate Debrief PortalModule-based experience with avatar video and color-coded drives
Personality Profile GraphFull profile across all six personality dimensions
Results & impact

What it changed.

250+
Initial debriefs delivered
1,000+
Scaled on renewal
Zero
Manual production work
Rave
Client reviews
Built with
React 18TypeScriptNode.jsPostgreSQLHeyGenElevenLabsSendGridAWS S3AirtableFramer Motion
Next case study

Render Intelligent Training

→