Embedded Files
Executive Snapshot
This project addressed a conversion gap in customer service sales caused not by lack of product knowledge, but by agents’ inability to recognize and respond to unspoken psychological resistance during live calls. I designed a blended learning solution combining interactive branching video and an AI-powered simulator to help agents identify “silent barriers” (e.g., uncertainty, loss of autonomy) and respond with appropriate psychological strategies in real time.
Business Context & Performance Problem
Customer service agents were expected to convert fixed services during support calls. While agents understood sales scripts and offers, conversion stalled at ~4%.
Analysis revealed that:
Customers rarely verbalized hesitation directly
Agents either pushed too hard (triggering resistance) or hesitated (missing opportunity)
Psychological barriers—uncertainty, perceived effort, fear of loss of control—went unnoticed
The business risk was ongoing:
Missed revenue opportunities
Frustrated customers feeling pressured or misunderstood
Inconsistent sales behaviors across agents
Success was defined as observable behavioral change in live calls, reflected in improved conversion rates.
Action Mapping (Diagnostic Backbone)
I conducted Action Mapping with customer service SMEs to ensure the solution targeted decision-making in conversation, not abstract persuasion theory.
Business Goal
Increase conversion rate of fixed services in Customer Service from 4% to 6% by Q3.
Critical On-the-Job Behaviors
Agents must:
Detect cues signaling hidden resistance (hesitation, uncertainty, disengagement)
Identify which psychological barrier is present
Select an appropriate response strategy (e.g., autonomy-support, reframing effort)
Apply that strategy fluidly within a live call
Key Barriers Identified
Perceptual Blind Spots
Agents often did not recognize resistance unless explicitly stated.
Strategy Selection Uncertainty
Even when hesitation was noticed, agents lacked confidence in how to respond.
Risk of Real-World Experimentation
Trying new psychological approaches with real customers felt risky.
What Was Intentionally Excluded
Generic sales theory
Long persuasion models detached from call flow
Passive knowledge-only content
These would not change in-call decision-making.
Experience the AI Simulator
I have vibe-coded the AI simulator using Lovable AI. It is powered by an LLM with a pre-defined system prompt.
Experience Interactive video
As a part of the solution, I designed and developed an interactive, branching training video iin Synthesia IO that is set in a story and helps the learner practice the right behaviour.
Constraints & Friction
Key constraints shaped the solution:
Agents had limited tolerance for “sales training”
Learning needed to fit between operational shifts
The solution had to scale without facilitator dependence
Practice needed to feel safe and relevant
These constraints ruled out role-play-heavy, info-dump solutions.
Options Considered & Rejected
Several alternatives were considered:
Script-based objection handling
Rejected because it encouraged mechanical responses without diagnosis.
Traditional sales eLearning
Too abstract; poor transfer to live conversations.
Single-format training
Insufficient for discrimination between different psychological barriers.
A blended approach combining interactive video for recognition and AI simulation for application was selected. The modules were spaced over a period of 2 months.
Design Strategy & Key Decisions
Key design decisions included:
Teaching agents to identify the barrier before choosing a tactic
→ Reduced pushy or hesitant behavior.
Using branching video for pattern recognition
→ Allowed learners to see consequences of misdiagnosis.
Using AI simulation for open-ended practice
→ Enabled realistic dialogue and repeated experimentation.
Designing around choice and autonomy-support strategies
→ Aligned with customer psychology rather than persuasion pressure.
Learning Science
Constructive Alignment & Backward Design:
We started with what agents need to do differently in live calls (recognize barriers, adjust responses), then worked backwards to design assessments (branching scenarios, simulator), activities, and content that support those behaviors. These ensure that learning objectives, activities, and assessments are tightly aligned
Cognitive Load Management:
To avoid overwhelming learners, content is chunked into manageable modules;
Active Learning & Varied Practice: Instead of simply reading or listening, agents engage via:
Branching video scenarios where they make decisions.
AI simulator practice where responses matter and feedback is immediate.
Reflection/reflection prompts: what worked / what surprised me. These engage deeper processing and help retention.
Interleaving types of barriers and responses across modules, so agents practise discrimination (not just repeating a single kind).
Spacing & Reinforcement:
Content is spread over multiple modules, each revisiting previously introduced ideas in new contexts (scenarios / simulator). Job aids reinforce learning after eLearning sessions. Follow-ups (reminders, reflected practice) help retention.
Motivation & Psychological Safety:
Using realistic scenarios, safe practice (no risk of real customer unhappiness), and branching outcomes that allow failure / correction builds confidence. Agents see relevance by tying scenarios directly to their work.
Feedback & Reflection Loops:
Immediate feedback in simulator / branching scenarios helps correction. Self-assessment and reflection after practice help agents internalize lessons.
Evaluation methods
Quantitative Metrics:
Weekly tracking of total user adoption
Analysis of consistent user engagement patterns
Measurement of unique use case implementation
Qualitative Assessment:
Implementation of Kirkpatrick Level 1 surveys for immediate feedback
Follow-up Kirkpatrick Level 3 evaluations after two months
eLearning NPS evaluation
Assessment of Barriers in Adoption
In-depth interviews with 30 participants regarding adoption barriers and preferred formats of instructional activities
Use of AI as an Accelerator
AI was used as a conversation simulator, not as content filler.
AI enabled:
Dynamic customer responses based on agent input
Practice with multiple psychological barriers in varied combinations
Immediate and reflective feedback loops
Scalable practice without coaching resources
AI was intentionally not used to automate evaluation or prescribe rigid scripts.
Implementation & Delivery
The solution was delivered as a structured learning flow:
Modules 1–2: Rise 360 core content with interactive lessons, case studies, and branching video (Synthesia)
Module 3: AI-powered simulator for applied practice
Job Aid: Quick-reference guide linking barriers to strategies and sample phrasing
Evaluation was embedded throughout.
Outcomes, Signals & Learnings
The project is currently in progress.
Early signals include:
Strong engagement with interactive video and simulator
Increased learner confidence reported in reflection prompts
Observable improvement in barrier identification during monitored calls
What This Project Demonstrates About Me
I design training around decision-making, not scripts
I use psychology as a diagnostic lens, not theory
I select technology to reduce risk and increase realism
I am comfortable building solutions where outcomes are measured later
I align learning design tightly to business metrics
Page updated
Google Sites
Report abuse