LeadRight

• 

  The conversation: say what you would really say. The counterpart pushes back and takes the space you leave open.

• 

  Where it turned: your exact line, a stronger one you could actually say, and Retry from this line.

• 

  Second attempt: the debriefing compares your behaviour with your last try, competency by competency.

• 

  Home: ten hand-written scenarios for new managers (feedback, boundaries, saying no, accountability) plus your own situation.

• 

  The debriefing opens with an honest summary and asks how ready you feel now, to compare with before.

• 

  LeadRight. Kind, not nice.

Inspiration

New managers fail for a specific reason: they are too nice. They bury the message between two compliments, apologise for raising it, and retreat the moment the other person stiffens. Every leadership book says "have the conversation"; nobody lets you practise it. We built the simulator we wished we'd had.

What it does

Briefing → conversation → debriefing → retry. You type (or dictate) what you would really say to a direct report, a peer or your boss. The AI counterpart holds its ground and, like a real person, takes the space you leave open when you people-please.

The debriefing scores four competencies and four phases (opening, facts, listening, agreement), runs two detectors (people-pleasing and facts-over-vague), quotes the exact line where the conversation was won or lost, hands you a stronger line, and lets you rewind to that moment and replay it. A second run gets a behavioural comparison. Self-rated confidence before and after is the number that matters.

Ten hand-written scenarios across feedback, boundaries, saying no and accountability. Pro adds the live AI counterpart, every scenario, and your own situation: describe the real conversation and rehearse it before it happens.

On the App Store for iPhone: free to start (four scenarios with a scripted counterpart, five conversations a week), Pro with a 7-day free trial ($6.99 a month or $39.99 a year).

How we built it

Flutter (iOS) on a Next.js backend on Cloud Run; Genkit + Gemini 3.7 Flash; one model call per turn (reply + outcome detection + momentum) and one judgement call at the debriefing; a stateless API, so rewind is a client-side truncation of the transcript; a scripted counterpart for the free tier (zero model cost); RevenueCat SDK + remote Paywall for Pro, RevenueCat webhook → Postgres for entitlements. No account: history stays on the phone.

Challenges we ran into

Making the counterpart realistic without being cruel. Making feedback specific (a line, not a lecture) and confidence-building (correct, never humiliate). Keeping every scenario hand-written and grounded in the literature (Radical Candor, Crucial Conversations, The Manager's Path), because anyone who coaches managers for a living will notice.

Accomplishments that we're proud of

The rewind: not "try again", but retry from the line that lost it, with a line you could actually say. And the confidence delta on the home screen.

What we learned

People-pleasing is detectable from a transcript, and showing someone their own hedges, verbatim, changes the next attempt more than any advice does.

What's next for LeadRight

Voice mode, more scenarios per track, Android.

Built With

• dart
• flutter
• gemini
• genkit
• google-cloud-run
• ios
• next.js
• postgresql
• prisma
• revenuecat
• typescript