The Guardrail Conversation
Pitch a new AI guardrail to a skeptical exec and watch a hidden trust meter move with every line you choose. Cite real incident data or propose a bounded pilot and trust climbs; dismiss their concern, promise zero risk, or bury them in jargon and it craters. Land above 70 and the guardrail ships — stall in the middle and it's a "maybe," bottom out and you've lost the room. The moat is the X-ray scrubber: it rewinds the exec's invisible internal reaction line by line, so you can find the exact sentence that tanked trust — the "what they're actually thinking" track a real conversation never shows. Runs entirely in your browser (0 uploads, works offline).
How to use it
- Start from a preset — Win them over or Lose the room — or build your own pitch from scratch.
- Click lines from the Build trust and Erode trust palettes to add them, in order, to the conversation.
- Read the transcript: each line shows its trust swing and the running trust score, with the biggest drop flagged.
- Drag the X-ray scrubber in the signature panel to rewind the exec's felt trust and read their internal reaction at each line.
- Copy or share your result — the "I lost the exec at line 3" hook is built to be argued about.
What this clears up (the fundamentals)
- Trust is spent and earned per sentence, not at the end — the meter moves on every line, so a single dismissive reply can undo three good ones. Persuasion is a running balance, not a closing argument.
- Concrete beats confident — citing a logged incident or a reversible pilot moves a skeptic more than a bold promise, because a skeptic can argue with a feeling but not with data or a small, safe test.
- Over-promising reads as dishonest — "zero risk, ever" doesn't reassure an exec; it tells them you don't understand the system, and trust in everything else you said drops with it.
- The line that tanks trust is findable — the biggest drop is a specific sentence at a specific moment. Naming it turns "that went badly" into "here's the line to rewrite."
Where it's used
This is the top-of-funnel playable for the Leadership L2 "The Guardrail Conversation" module — the fastest way to feel why risk communication to a non-technical stakeholder is a trust curve, not a slide deck. It's a miatz build-lab concept playable: play it here, then learn to build the trust-meter model and the rewind-scrubber reveal yourself, and it feeds the reverse-question trivia item bank.
FAQ
What is the trust meter actually modelling?
A skeptical exec's confidence in you and your proposal, on a 0–100 scale starting at 50. Each conversational move carries a hand-set signed weight; the engine applies them in order and clamps the running total to 0–100. It's a deterministic teaching rubric — no model call, no randomness.
Why does over-promising lower trust?
Because an experienced exec knows no system is risk-free. An absolute guarantee reads as naive or dishonest, so instead of reassuring them it makes them re-doubt the rest of your pitch. Honest, bounded claims (a pilot, real data) are what actually build trust.
What is the "rewind scrubber" showing me?
The exec's invisible internal reaction. A real conversation only shows you their face; the scrubber replays their felt-trust value and the reason it moved at each line, so you can pinpoint the exact sentence that turned the room — the Demystify X-ray move.
Is anything uploaded?
No. Your choices are scored entirely in your browser — nothing is transmitted, stored, or logged. Turn off your Wi-Fi and it still works.
Can I lose the room and recover?
Sometimes. If trust hits 0 at any point the room "collapses" and that's flagged, but as long as it stays above the floor a few strong lines can pull a stalled pitch back toward approval. The order you play the lines in matters as much as which ones you pick.
Limits
A teaching rubric with six fixed moves and hand-set weights — it's a fast, opinionated read on trust dynamics, not a script for your actual board meeting. Real personas (a CFO, legal counsel, an impatient CEO) weight risk, jargon, and pilots differently, and timing and tone carry meaning the meter can't. The lesson it drills — concrete data and reversible pilots build trust while dismissal and over-promising destroy it — is exactly the real practice.
Related
Part of the Demystify Playgrounds. Explore the rest from the Playgrounds home.
Bookmark this page (Ctrl+D, or ⌘D on Mac) — it works offline the next time you need it.