Executive Snapshot

This project addressed a measurable decline in customer experience driven not by lack of process knowledge, but by empathy breakdown under operational pressure. I designed 4 modules spaced over time, two of them including an AI-powered conversation simulator that allowed senior customer support agents to practice emotionally difficult conversations in a safe, repeatable environment. The solution focused on behavior change under stress, not theoretical soft-skills training, and was explicitly tied to reducing the Negative Satisfaction Score.

Action Mapping (Diagnostic Backbone)

I led a structured Action Mapping process with SMEs and customer support leadership to ensure the solution targeted observable behaviors, not abstract competencies.

Business Goal

Reduce the Negative Satisfaction Score from 8% to 3% within three months by improving empathetic handling of difficult customer interactions.

Critical On-the-Job Behaviors

• Agents must consistently:

• Identify customer emotions early

• Allow customers to fully explain their issue

• Demonstrate active listening signals under time pressure

• Validate emotions without escalating or over-apologizing

• Set boundaries professionally when conversations become abusive

• Signal ownership and progress toward resolution

Key Barriers Identified

1. Skill Degradation Under Load

• Empathy techniques were known but not reliably executed during high-volume shifts.

2. Lack of Safe Practice Space

• Agents had no environment to practice emotionally complex conversations without real customer risk.

3. One-off Training Ineffectiveness

• Previous soft-skills training did not transfer into daily behavior.

What Was Intentionally Excluded

• Long theory-based empathy courses

• Passive video-only learning

• Generic role-play without feedback loops

• These approaches failed to simulate real emotional pressure.

Options Considered & Rejected

1. Refresher ILT workshops

   • Rejected due to low transfer and scheduling friction.

2. Script-based empathy job aids only

   • Helpful as support, but insufficient for skill execution under stress.

3. Classic branching-only scenarios

   • Too predictable; did not reflect real conversational complexity.

The AI simulator was selected because it allowed free-form dialogue, emotional variation, and scalable repetition.

Use of AI as an Accelerator

AI was used as a performance simulation engine, not content automation.

AI enabled:

• • Dynamic emotional responses from “customers”

  • Real-time conversational branching based on agent input

  • Scalable practice without facilitators

  • Detailed post-conversation analysis highlighting patterns and blind spots

AI was not used to:

• Replace human judgment

• Score agents rigidly

• Deliver generic coaching

Outcomes, Signals & Learnings

The program is currently ongoing.

Early signals:

• Strong engagement with the simulator

• High voluntary return usage for challenge mode

• Positive qualitative feedback on realism and usefulness

This simulator is currently the most complex AI-based learning solution deployed in the organization.

Learning Science

To create an engaging, effective training solution for customer-facing agents, I applied seven core principles from learning science. The result is a curriculum that promotes skill mastery, knowledge retention, and real-world transfer for measurable impact on performance.

1. Spaced Repetition

Over a 2–3-month period, the system automatically schedules practical activities at increasing intervals. This ensures that once-learned skills (e.g., empathic listening) are revisited just as they begin to fade, strengthening long-term retention and reducing the need for retraining.

2. Scaffolding: Gradual Increase in Complexity

• Phase 1 (Focused Drill): Learners practice conversational skills in Storyline scenario-based multiple-choice prompts (A, B, C) to reinforce decision-making under guided conditions.

• Phase 2 (Guided Choices): The learners practice simulated conversations with AI simulator in Training mode that gives them hints to ensure practical application of new knowledge. The choices are guided, but the learners themselves write replies and get immediate feedback as well as delayed instructional one.

• Phase 3 (Open Dialogue): Agents engage in free-form conversations with AI simulator in gamified Challenge mode, applying all previously learned sub-skills in an authentic context. To drive engagement, the simulator counts points, increases levels of  difficulty and tracks streaks. At the end the learner is given specific developmental feedback so they know what to focus on next.

3. Elaboration

• Before introducing new material, I activate learners’ prior knowledge:

• We deploy Microsoft Forms surveys with open-ended prompts (e.g., “How do you demonstrate empathy in your daily calls?”).

• By articulating what they already know, agents form richer connections to upcoming scenarios and theory.
  
  

4. Generation

• Learners tackle realistic tasks before theory.

• They make decisions in the Storyline experience or the AI simulator and immediately see consequences — driving curiosity and self-discovery.

• Only afterward do we introduce the underlying models (even in the form of a job aid) and best-practice guidelines, ensuring concepts are anchored in hands-on experience.
  
  

5. Reflection

• After each practical activity:

• Microsoft Forms prompts (e.g., “Why did you choose that response? What happened next? What did you learn?”) guide agents through structured reflection. This metacognitive step solidifies lessons learned and encourages continuous self-improvement.
  
  

6. Constructivism

• I prioritized active construction of knowledge over passive absorption:

• A suite of short, context-rich activities simulates real call-center challenges, allowing agents to practice and adapt in safe, controlled environments.

• The simulator provides immediate, emotionally rich feedback—visual cues of customer satisfaction or frustration, plus detailed, criterion-based scores and targeted “developmental feedback” with actionable improvement opportunities.

7. Examples & Non-Examples

• Throughout the program, every principle and technique is illustrated with both:

• Positive Example: A model interaction demonstrating best practices.

• Negative Example: A contrasting scenario highlighting common pitfalls.

This approach combines proven instructional strategies with advanced simulation technology — ensuring agents not only learn, but excel.

8. Evaluation

• KirkPatrick LVL 2 (Knowledge): 

  1. Learners take scenario-based test at the beginning of the learning program and at the end, which allows us to see if they have absorbed the knowledge.

  2. Learners send us the feedback analysis from the simulator after each module so we can see their progress.

• KirkPatrick LVL 3 (Job application):

  1. Learners fill out self-evaluation questionnaire where they evaluate to what extent they perform the desired behaviours (techniques of empathetic communication) on the job before and after the training program.

• Main KPI: Impact on the Negative Satisfaction Score

  1. After the 2-month training program we will evaluate whether the main goal of lowering the Negative Satisfaction Score in customers' surveys have been reached.