Internal product
Train for the conversation that counts.
The AI plays the real person. Pushes back. Stays in character. And after every session you get an exact readiness score — with what worked, what fell flat, and what you would say differently next time.
what you should say.
You've never said it to someone who actually pushes back. That is not a confidence problem. It is a practice problem.
How it works
Four steps. A real conversation.
01
Pick a scenario
18 real conversations: work, personal, and life. Each with three levels of pressure.
02
Describe the person
Name, style, dynamic. The AI becomes that person — and stays in character.
03
Run the session
No softballs. The AI pushes back the way they would. You negotiate for real.
04
Get your score
Readiness score 1-10, what worked, what fell flat, and what to say differently.
Scenarios
18 real conversations.
Each scenario maps to a conversation most people avoid. Three levels: Supportive, Realistic, Combative.
Work
- Ask for a raise
- Negotiate a job offer
- Nail a job interview
- Quit gracefully
- Push back on workload
- Deliver critical feedback
- Challenge a performance review
- Push back on a difficult client
Personal
- End a relationship
- End a friendship
- Set limits with a parent
- Talk about money
- Apologize sincerely
Life
- Negotiate with a landlord
- Dispute a charge
- Confront a roommate
- Negotiate with a contractor
- Ask for more time
Difficulty levels
Train at the pressure you will actually face.
●
Supportive
The other person is receptive. Good for learning basic framing and argument structure.
●●
Realistic
They push back moderately. They ask for justification. They test your logic.
●●●
Combative
They push hard, cut in, and turn your arguments back on you. The real pressure before the real conversation.
The debrief
Not a score. A plan.
Every session ends with a detailed debrief: readiness score 1-10, the exact moment you lost ground, and what you would say differently next time.
Readiness score
7
/10
Ask for a raise — Realistic
Readiness score
1-10 based on clarity, holding under pressure, and resolution.
What worked
The exact moment your argument landed — why it held.
What fell flat
The exact moment you lost ground — what happened.
Try this instead
The exact phrase you should have said at that moment.
One thing to change
The highest-impact change for your next session.
Architecture
Built for real pressure.
18
scenarios
Work, personal, and life — the conversations that actually happen.
3
levels
Supportive → Realistic → Combative. Train at the pressure you will actually face.
1–10
score
Not a participation ribbon. The exact moment you lost ground.
EN/FR/AR
trilingual
Product, UI, and AI responses in three languages.
Built differently
Not a therapist. Not a coach. Available at 11pm.
Therapy happens after. Coaching is expensive and rare. RoleplayAI sits between: always available, resistant, scored, and accessible before the real conversation.
×
Therapy / coaching
Retrospective, costly, rare
Help arrives after the conversation. The interval between sessions is empty space. And no one pushes back like the real person.
×
Role-play with a friend
Too soft, personal bias
Your friends do not push hard enough. They know your history. The real person will not go easy on you.
✓
RoleplayAI
Available, resistant, scored
The AI plays the real person. Pushes back the way they would. You have a score and a plan before you sleep.
Why Arabic
The only conversation training platform built for Arabic.
Hard conversations do not stop at language borders. RoleplayAI is RTL from the architecture up: interface, AI, and scenarios — in native Arabic, not machine translation.
01
Native RTL interface
Text direction, navigation, and logic designed for Arabic from the first commit.
02
AI in Arabic
The AI responds, pushes back, and holds character in Arabic — with a register matched to real conversation.
03
All 12 scenarios
Every scenario, difficulty level, and debrief available in Arabic simultaneously.
Arabic UI — aperçu
اختر المشهد
The platform
AI, scenarios, debrief, and progress.
RoleplayAI is not a chatbot. It is a full training system: 18 scenarios, three difficulty levels, a scored debrief after every session, and progress tracking over time.
AI Roleplay
The AI becomes the real person. Stays in character. Pushes back the way they would in real life.
Readiness Score
Readiness 1-10. What worked, what fell flat, what to say differently.
18 Scenarios
Work, personal, and life conversations. Each with three levels of pressure.
3 Difficulty Levels
Supportive, realistic, combative. Train at the pressure you will actually face.
History & progress
Readiness scores by scenario over time. Watch your preparation improve.
EN/FR/AR
Full product and AI responses in three languages.
Account & onboarding
Context-aware onboarding, dark/light theme, 3-step cancellation flow, and 30-day account recovery.
Technical architecture
Next.js 16
App Router, SSR, and trilingual routing. Every page available in EN, FR, and AR.
Claude API
The AI plays the person. Holds character, pushes back, and responds in real time.
Supabase
Auth, database, sessions, and history. Progress stored per user.
Stripe
3-session free trial, Pro at CAD $15/mo, Annual at CAD $119.99/yr.
next-intl
Trilingual EN/FR/AR routing and translations. Native RTL support for Arabic.
Render
Production deployment with long-function support for debrief generation.
Have a product to build?
Maison GR designs and builds SaaS products, AI tools, and platforms for real problems. Let's talk about yours.
Start a project