Step-by-Step Study Material

Step 1: Why Traditional Sales Role Play Fails

Before building a solution, understand the problem it solves.

Traditional role play has four critical weaknesses:

Key Insight: When a rep worries about looking bad in front of their manager, their attention is on the room, not on learning. Removing the audience enables real repetition.

---

Step 2: The Solution — An Always-On AI GTM Coach

The answer is an AI coach that is:

• Always available (no scheduling needed)
• Consistent (same quality every session)
• Private (no social risk for the rep)
• Trained on real buyer conversations (not invented scenarios)

The Two-Layer Architecture

┌─────────────────────────────┐
│        FRONT END            │
│  (Lovable)                  │
│  - Role play studio         │
│  - Transcript upload        │
│  - Live scorecard           │
│  - Certification paths      │
│  - Drill + completion track │
└────────────┬────────────────┘
             │
┌────────────▼────────────────┐
│        AGENT LAYER          │
│  (ElevenAgents)             │
│  - System prompt            │
│  - Voice + language config  │
│  - Knowledge base           │
│  - Scoring rubric           │
└─────────────────────────────┘

Think of it this way: The agent is the brain and voice of the buyer. The front end is everything the rep sees and touches.

---

Step 3: The Three Design Decisions That Make It Work

These choices separate a useful AI coach from a generic chatbot.

Decision 1: Tiered Disclosure

• The AI buyer does not volunteer information freely
• The rep must earn details by asking sharp questions or leading with a point of view
• This prevents reps from passing with three generic questions

Example in practice:

Rep asks a vague question → buyer gives minimal response Rep leads with insight → buyer opens up and reveals a lost deal

---

Decision 2: Structured Scoring

• The agent breaks character at the end and scores the rep
• Scoring is against defined criteria, for example:
  • Technical accuracy
  • Buyer credibility
  • Whether the rep pitched at the buyer's level
• A live scorecard fills in during the call, not just after

---

Decision 3: Teaching Mode

• If a rep is stuck, they say so
• The agent pauses the exercise, teaches the skill with a concrete example, then resumes
• This makes repeated use feel safe rather than exposing

Why this matters: Reps will only return to a tool that doesn't make them feel punished for not knowing something.

---

Step 4: Best Practices for Building Your Own AI Role Play Coach

Follow these six principles:

1. Write the Rubric Before the Prompt

• Define what success looks like first
• Then build the agent to test for those criteria
• Doing it backwards produces a scorecard reps ignore

2. Mine Real Calls for Scenarios

• Do not invent scenarios from scratch
• Point an AI tool at your library of historical sales calls
• Have it identify the most common conversation types and build replicas

3. Keep Agent Responses Short

• Limit the agent's responses to 2–4 sentences in the prompt
• Longer instructions cause the agent to behave inconsistently

4. Match the Model to Conversation Complexity

Simple role play  →  Light, fast model  →  Lower cost, lower latency
Complex scenario  →  More capable model →  Higher accuracy, higher cost

5. Tune Prompts for Regional Differences

• A US enterprise buyer pushes back differently than one in EMEA, India, or Japan
• Work with local reps to understand regional conventions
• Write those differences explicitly into the prompt

6. Treat Adoption as Change Management

Deployment is the easy part. Bringing people along takes more work:

Remember: The build takes an afternoon. Changing rep behavior takes longer.

---

Summary: The Full Picture

PROBLEM → Traditional role play is inconsistent, hard to schedule,
           and socially risky for reps

SOLUTION → AI GTM Coach: always-on, objective, private

ARCHITECTURE → Agent layer (ElevenAgents) + Front end (Lovable)

KEY DESIGN → Tiered disclosure + Structured scoring + Teaching mode

BUILD RULES → Rubric first → Real call scenarios → Short prompts
              → Right model → Regional tuning → Change management

---

Quick Self-Check Questions

1. What are the four limitations of traditional sales role play?
2. What is the difference between the agent layer and the front end?
3. Why does tiered disclosure make the exercise harder to game?
4. What happens when a rep says they are stuck in teaching mode?
5. Why should you write the scoring rubric before writing the system prompt?
6. What is the biggest challenge in rolling out an AI coach — the build or the adoption?