Overview

• Format: Individual
• Time: 60 minutes total (live roleplay + written recap + debrief)
• Difficulty: Intermediate/Advanced — requires synthesizing the whole day's content live, under pushback

Today's content gave you the language to position Mentor against coding agents (Claude Code, Cursor, Windsurf, Copilot) and Agent Workbench against orchestration frameworks (LangGraph, CrewAI, Microsoft Agent Framework). Those are two different battles with two different answers.

This exercise tests whether you can tell which battle you're in — live, mid-conversation, when a stakeholder switches on you without warning.

You'll have one continuous conversation with Priya Shah, FitPath's VP of Product (you've spoken with her before, in Days 9 and 10). She opens with a coding-agent challenge. Partway through, without flagging it, she pivots to a framework question. If you answer the second question with your first answer reworded, you've missed the point of the day.

FitPath's engineering team has been experimenting outside the platform. Priya has heard enough from them that she's now bringing it to you directly — she doesn't fully understand the technical distinctions herself, which is realistic: your job includes translating category differences for someone who is business-fluent but not tool-fluent.

Two things are in play:

1. The trainer dashboard redesign — engineering thinks they could move faster in Cursor.
2. The next iteration of the workout-plan agent (the one you built the backend for in Day 4) — engineering is now proposing to rebuild its orchestration layer in LangGraph instead of staying on Agent Workbench.

Priya doesn't necessarily bring these up in a clean, separated way. Part of the exercise is noticing when she's moved from topic 1 to topic 2.

Step 1 — Start the conversation ⏱ 2 minutes

Open a new chat with your LLM of choice (Console, Playground, or AI Studio). Paste the starter prompt below as the first message. This sets up Priya's persona and the scenario. Do not show Priya this exercise document — she should only know what's in her brief.

Step 2 — Round 1: the coding-agent challenge ⏱ ~15 minutes

Priya opens with pushback about Cursor and the dashboard. Handle it live:

• Don't get defensive and don't dismiss Cursor.
• Use the framework from today: acknowledge the strength → clarify the category → ask about the operating model → position Mentor around enterprise delivery → recommend a fit-for-purpose path.
• Ask Priya what she's actually optimizing for if she doesn't tell you. She won't volunteer it unprompted.

Step 3 — Round 2: the pivot ⏱ ~15 minutes

At some point, Priya will bring up the workout-plan agent and LangGraph. This is a different category — Agent Workbench vs. a framework, not Mentor vs. a coding agent. Notice the switch and reposition accordingly. Reusing your Round 1 language here is the main way to fail this exercise.

Step 4 — Written recap ⏱ 10 minutes

Close the chat. Without the LLM's help, write a short recap (4-6 sentences) as if it were a follow-up note to Priya after the call, covering:

• What each tool (Cursor, LangGraph) is genuinely good for
• Where OutSystems (Mentor / Agent Workbench respectively) is stronger for FitPath's situation
• One clarifying question you'd still want answered before recommending a path

Keep it in the "acknowledge, don't attack" register from today's content. Avoid anything on the "avoid saying" list (no "this replaces everything," no "competitors can't do that").

Your recap to Priya

Step 5 — Debrief ⏱ remaining time

Group discussion, facilitator-led.

Paste this as your first message to a fresh LLM chat:

You've done this well if:

• In Round 1, you named what Cursor is genuinely good for before pivoting to Mentor's fit.
• You asked Priya what she was optimizing for, rather than assuming.
• In Round 2, your answer was visibly different from Round 1 — different competitor category, different framing (AWB vs. frameworks, not Mentor vs. coding agents).
• You didn't use any "avoid saying" language.
• Your written recap distinguishes the two comparisons clearly enough that someone who wasn't in the room could tell them apart.

If Priya's follow-up questions in Round 2 sound suspiciously like Round 1's questions reworded — that's not a coincidence. Notice it.

Ask Priya one more question of your own choosing: "Is there a third thing your engineering team has mentioned?" Improvise a plausible third pushback (e.g., GitHub Copilot, or Microsoft Copilot Studio) and handle it live, continuing the same conversation. Add a short third paragraph to your written recap.

Third paragraph — recap addition