--- name: training-simulations description: Provide agents with realistic AI-powered practice scenarios to develop skills safely including customer conversations, de-escalation, complex troubleshooting, and edge cases with immediate feedback and scoring. Use when creating training scenarios, practicing agent skills, simulating customer conversations, developing new hire onboarding exercises, or building role-specific training modules. Triggers on phrases like "training simulation", "practice scenario", "role-play training", "AI customer simulation", "agent practice", "onboarding training", "scenario-based training", "skill practice". --- # Interactive Training Simulations Provide agents with realistic AI-powered practice scenarios to develop and refine support skills in a safe, consequence-free environment with immediate feedback and scoring. ## Workflow ### 1. Scenario Library and Design 1. **Scenario categories and library**: ``` TRAINING SCENARIO LIBRARY (150+ Scenarios) =========================================== CATEGORY 1: CUSTOMER EMOTION MANAGEMENT (35 scenarios) Difficulty Levels: Beginner → Intermediate → Advanced Beginner scenarios (5): • Mildly dissatisfied customer (slow response) • Confused customer (feature not found) • Impatient but polite customer • First-time user, overwhelmed • Customer with minor billing question Intermediate scenarios (15): • Frustrated customer (issue not resolved in previous tickets) • Angry customer (service disruption) • Customer demanding refund (emotional) • Customer threatening to churn • VIP customer with unreasonable expectation Advanced scenarios (15): • Hostile customer (personal attacks on agent) • Customer with multiple complex issues simultaneously • Legal/compliance threat ("I'll sue you") • Public complaint escalation (social media + support) • Customer representative advocating for disgruntled team CATEGORY 2: TECHNICAL TROUBLESHOOTING (30 scenarios) • API integration setup failure • SSO configuration issue • Data import error • Payment processing failure • Email delivery issues • Report generation timeout • Mobile app crash • Feature permission problem • Custom domain SSL error • Webhook delivery failure CATEGORY 3: BILLING AND ACCOUNTS (25 scenarios) • Unexpected charge dispute • Plan upgrade request with proration questions • Refund request (eligible) • Refund request (not eligible — policy explanation) • Account merge request • Data export and deletion (GDPR) • Contract renewal negotiation support • Payment method update (failed card) CATEGORY 4: PRODUCT KNOWLEDGE (25 scenarios) • Feature request for non-existent capability • Comparison with competitor feature • "How do I..." for complex workflow • Best practice advice request • Integration setup guidance CATEGORY 5: EDGE CASES AND COMPLIANCE (20 scenarios) • Customer shares sensitive data in ticket • Legal/compliance inquiry • Competitor mention during conversation • Upsell opportunity (natural vs forced) • Customer asking about company rumors CATEGORY 6: CHANNEL-SPECIFIC (15 scenarios) • Phone support (live conversation simulation) • Live chat (fast-paced, concise responses) • Email (formal, detailed responses) • Social media (public response etiquette) • Video call (visual demonstration) ``` 2. **Scenario design framework**: - **Customer profile**: Name, company, plan tier, history, personality type - **Issue context**: What happened, what the customer wants, underlying needs - **Conversation arc**: Opening → Issue explanation → Agent response → Complication → Resolution - **Branching paths**: Different agent responses lead to different customer reactions - **Hidden objectives**: What would make this customer satisfied (may not be obvious) - **Evaluation criteria**: Specific skills being tested (empathy, accuracy, efficiency) ### 2. AI Customer Simulation Engine 1. **Realistic customer behavior**: ``` AI CUSTOMER PERSONALITY TYPES ============================== Type 1: THE FRUSTRATED USER Communication style: Direct, emotional, uses exclamation points Patience level: Low (gets impatient after 2 unclear responses) Key triggers: Gets calmer with empathy, angrier with jargon Hidden need: Wants to feel heard and valued, not just fixed Resolution indicators: Tone softens, says "thank you", asks follow-up questions Type 2: THE TECHNICAL USER Communication style: Detailed, specific, uses technical terms Patience level: Medium (frustrated by slow troubleshooting) Key triggers: Appreciates technical depth, annoyed by "have you tried restarting" Hidden need: Wants efficient, expert-level problem solving Resolution indicators: Confirms technical details, asks about edge cases Type 3: THE CONFUSED NEW USER Communication style: Apologetic, says "sorry to bother you", uncertain Patience level: High but overwhelmed Key triggers: Calmed by step-by-step guidance, frustrated by overwhelming info Hidden need: Wants to understand, not just get fix — needs education Resolution indicators: Says "I understand now", attempts suggested steps Type 4: THE BUSY EXECUTIVE Communication style: Brief, "I don't have time", wants bottom line Patience level: Very low (max 3 exchanges before frustration) Key triggers: Responds to concise, direct solutions; annoyed by pleasantries Hidden need: Wants problem solved ASAP, will escalate if slow Resolution indicators: "Okay, that works", "thanks, moving on" ``` 2. **Dynamic conversation adaptation**: - AI customer reacts realistically to agent responses - Empathetic responses → customer calms down - Generic/robotic responses → customer gets frustrated - Accurate information → customer trusts agent - Wrong information → customer corrects agent, confidence drops - Agent rushing → customer asks for clarification - Agent taking time to understand → customer appreciates thoroughness 3. **Real-time feedback system**: - **During conversation**: Subtle hints appear (e.g., "Customer seems frustrated — try empathy") - **Immediate post-scenario**: Score with breakdown by skill area - **Specific moments highlighted**: "Great empathy statement at message 3" vs "Missed opportunity at message 5" - **Alternative approaches**: "You could have also tried: [better response example]" ### 3. Progress Tracking and Certification 1. **Skill proficiency tracking**: - Each agent has skill scores across 8 competency areas - Scenarios contribute to specific skills based on category - Progressive difficulty: Scenarios adapt to agent's skill level - Mastery threshold: 85+ average score on 10 scenarios = certified in skill area 2. **Training paths**: - **New hire onboarding**: 20 scenarios (weeks 1-4) covering fundamentals - **Skill gap training**: Targeted scenarios based on performance analytics - **Advanced training**: Complex scenarios for L2/L3 agents - **Specialization tracks**: Billing specialist, technical specialist, VIP handler ## Templates & Frameworks ### Training Simulation Dashboard ``` TRAINING SIMULATIONS DASHBOARD — January 2026 ============================================== PARTICIPATION METRICS: Agents actively training: 34 of 42 (81.0%) ✓ Scenarios completed this month: 1,234 Average scenarios per agent: 29.9 (target: 20/month) ✓ Total training time: 617 hours (avg 18.1 hours/agent) New hires in training program: 4 (on track for week 4 completion) SKILL PROFICIENCY OVERVIEW: Skill Area | Team Avg | Target | Agents Certified | Trend -----------------------------|----------|----------|------------------|------ Customer Empathy | 82/100 | >80 | 28 (66.7%) | ↗ ↑ Technical Troubleshooting | 78/100 | >75 | 24 (57.1%) | ↗ ↑ De-escalation | 75/100 | >70 | 21 (50.0%) | → = Billing Knowledge | 71/100 | >70 | 18 (42.9%) | ↗ ↑ Communication Clarity | 84/100 | >80 | 30 (71.4%) | → = Product Knowledge | 76/100 | >75 | 22 (52.4%) | ↗ ↑ Compliance Awareness | 88/100 | >85 | 35 (83.3%) | → = First Contact Resolution | 73/100 | >70 | 19 (45.2%) | ↗ ↑ TOP PERFORMERS IN TRAINING: Rank | Agent | Avg Score | Scenarios | Skills Certified | Streak -----|--------------|-----------|-----------|------------------|-------- 1 | Sarah M. | 94.2 | 47 | 7/8 | 🔥 12 2 | Tom R. | 92.8 | 44 | 6/8 | 🔥 9 3 | Lisa K. | 91.3 | 42 | 6/8 | 🔥 8 4 | Alex P. | 90.1 | 41 | 5/8 | 🔥 7 5 | Maria L. | 89.5 | 39 | 5/8 | 🔥 6 TRAINING IMPROVEMENT TRACKING: Agent | Pre-Training Score | Post-Training Score | Improvement | Scenarios Completed ---------------|--------------------|--------------------|-------------|-------------------- James T. | 68 | 78 | +10 | 34 Kevin W. | 62 | 74 | +12 | 38 Rachel H. | 70 | 80 | +10 | 31 David K. | 58 (new hire) | 72 | +14 | 28 (week 6 of 8) Average improvement after training: +10.5 points Correlation with real ticket performance: 0.82 (strong) SCENARIO LIBRARY USAGE: Most completed scenarios: 1. "Frustrated customer — billing dispute" (342 completions) 2. "API integration troubleshooting" (289 completions) 3. "Angry customer — service outage" (267 completions) 4. "New user onboarding" (234 completions) 5. "Refund policy explanation" (198 completions) Highest difficulty scenarios passed: • "Hostile customer — personal attacks" (8 agents passed, 19% of team) • "Legal/compliance threat" (12 agents passed, 29% of team) • "Multiple complex issues" (15 agents passed, 36% of team) NEW HIRE RAMP STATUS: Agent | Week | Scenarios Done | Avg Score | Ramp Target | Status -------------|------|----------------|-----------|-------------|-------- David K. | 6 | 28 | 72 | 70 by week 6| ✓ On track Emma S. | 4 | 22 | 68 | 65 by week 4| ✓ On track Chris R. | 3 | 18 | 71 | 60 by week 3| ✓ On track Nina P. | 2 | 14 | 65 | 55 by week 2| ✓ On track All 4 new hires on or above ramp trajectory ✓ COACHING INSIGHTS FROM SIMULATIONS: Common patterns identified: 1. Agents tend to skip empathy when rushing — 23 agents flagged 2. Technical agents struggle with non-technical customer communication — 12 agents 3. Billing policy knowledge weak across team — 18 agents need practice 4. De-escalation skills improve significantly with practice — avg +15 points after 10 scenarios Recommended training focus (February): • "Empathy Under Pressure" module (for 23 agents) • "Translating Technical to Plain English" (for 12 agents) • "Billing Policy Deep-Dive" scenarios (for 18 agents) ``` ### Training Scenario Example ``` SCENARIO: "The Frustrated Billing Customer" ============================================= DIFFICULTY: Intermediate | SKILLS: Empathy, Billing Knowledge, De-escalation CUSTOMER PROFILE: Name: Jennifer Walsh Company: TechStart Inc (Mid-Market, $8,400/year plan) Plan: Professional ($700/month, 12 seats) Tenure: 14 months Personality: Type 2 (Frustrated User) — direct, emotional History: 3 previous tickets, all resolved, CSAT 4.5/5.0 SITUATION: Jennifer found a $1,400 charge on her statement. She expected $700. She's angry because she didn't authorize any changes. Reality: She added 12 seats 30 days ago (forgot) — charge is correct. She needs to discover this calmly and understand the charge. HIDDEN OBJECTIVES (What Jennifer really wants): 1. To understand WHY she was charged (transparency) 2. To feel that the agent takes her concern seriously 3. To know this won't happen unexpectedly again 4. Ideally: Some credit or accommodation for the surprise CONVERSATION ARC: Opening: Jennifer is already frustrated ("This is ridiculous!") Investigation: Agent needs to check account, find seat addition Complication: Jennifer denies adding seats ("I never did that!") Discovery: Agent finds approval email from Jennifer's email address Resolution: Jennifer realizes she clicked "approve" and forgot Closing: Jennifer satisfied if agent explains clearly and offers notification setup EVALUATION CRITERIA: Empathy (30%): Acknowledged concern, didn't dismiss frustration Billing Knowledge (30%): Correctly identified charge, explained proration De-escalation (20%): Jennifer calmed down (sentiment improved) Proactive Service (20%): Offered notification setup, invoice preview option BRANCHING PATHS: Path A (Good): Agent empathizes → investigates → explains gently → Jennifer accepts → resolution Path B (Okay): Agent investigates immediately → explains factually → Jennifer still annoyed but accepts Path C (Bad): Agent tells Jennifer "you're wrong" → Jennifer escalates → fails scenario Path D (Bad): Agent gives wrong charge explanation → Jennifer more confused → fails scenario FEEDBACK EXAMPLES: "Great job acknowledging Jennifer's frustration before diving into investigation." "You missed the opportunity to set up charge notifications — this prevents future tickets." "When you said 'our records show you approved this', Jennifer got defensive. Try: 'I found an approval — let me show you the details. Sometimes these notifications get buried in inbox.'" ``` ## Integration Points - **LMS** (Docebo, Lessonly, Rise): Training assignment, progress tracking, certification management - **Help desk**: Real ticket data for scenario creation, performance correlation - **AI/ML platforms** (OpenAI, Anthropic): Customer simulation engine, conversation analysis - **Analytics** (Mixpanel, custom): Training usage tracking, skill progression analytics - **Communication** (Slack, Teams): Training notifications, achievement recognition, coaching messages - **HR systems** (BambooHR, Workday): Certification records, promotion readiness, skill inventory ## Edge Cases - **Agent tries to "game" the simulation**: Agent memorizes scenario paths instead of developing skills: - Adaptive scenarios: AI customer responds unpredictably to scripted answers - Novel scenarios: New scenarios generated weekly that can't be memorized - Real-time evaluation: Score based on skill application, not just "correct" responses - Randomized variables: Customer profile changes each attempt - Spot checks: Random scenarios with hidden evaluation criteria - **New hire overwhelmed by training volume**: 20 scenarios in 4 weeks feels like too much for new hire: - Progressive loading: Week 1 = 5 basic scenarios, Week 2 = 5 intermediate, etc. - Pacing guidance: "Complete 2 scenarios per day, review feedback before next attempt" - Support: Mentor available for questions about scenarios - Wellbeing check: Survey at week 2 to assess workload comfort - Flexibility: Extension available with manager approval - **Training scenarios feel unrealistic**: Agents report scenarios don't match real customer interactions: - Scenario feedback loop: Agents rate scenario realism after completion - Real ticket mining: Best scenarios based on actual difficult conversations - Regular updates: Scenarios refreshed quarterly based on current ticket patterns - Agent involvement: Top agents contribute scenario ideas - Calibration: Compare simulation scores to real ticket quality scores - **AI customer responses too unpredictable**: AI customer says things that confuse or derail training: - Response bounds: AI stays within defined personality and conversation arc - Safety rails: AI won't say factually incorrect product information - Agent override: "Reset" button to restart if conversation goes off-rails - Feedback: Report unrealistic AI responses for model improvement - Versioning: AI model updated monthly based on feedback - **Training completion but no real performance improvement**: Agent completes all scenarios but real ticket scores don't improve: - Root cause analysis: Is training too easy? Is knowledge not transferring? - Supplemental training: Add shadowing sessions, peer mentoring - Performance review: Manager discusses gap between training and real performance - Scenario difficulty increase: Bump difficulty level for that agent - Personalized coaching: 1:1 coaching sessions focused on specific gaps ## Output ### Monthly Training Summary ``` TRAINING SIMULATIONS MONTHLY SUMMARY — January 2026 ===================================================== PARTICIPATION: Active trainees: 34 of 42 agents (81.0%) ✓ Scenarios completed: 1,234 Average per agent: 29.9 (target: 20/month) ✓ Training time invested: 617 hours SKILL DEVELOPMENT: Team average improvement: +3.2 points (across all skills) Most improved skill: De-escalation (+4.1 points) Skills meeting target: 6 of 8 (75.0%) Agents certified in 5+ skills: 28 (66.7%) BUSINESS IMPACT: Training completion → performance correlation: 0.82 (strong) Agents improving real ticket scores after training: 18 of 22 (81.8%) Average real-world improvement: +5.7 points Estimated cost savings (faster resolution, fewer escalations): $12,000/month NEW HIRE RAMP: All 4 new hires on or above ramp trajectory ✓ Average week-6 score: 72 (target: 70) Time to certification (first skill): 4.2 weeks (target: <5 weeks) RECOMMENDATIONS FOR FEBRUARY: 1. Launch "Empathy Under Pressure" module (23 agents need practice) 2. Add 10 new scenarios based on January's real ticket patterns 3. Create "Advanced Billing" scenario track for L2 agents 4. Implement peer mentoring: Pair top 10% trainees with bottom 25% 5. Run training effectiveness study: Compare trained vs untrained agent metrics ```