Skip to main content

Admin MCP — Governed Pricing

This is the most important flow to show. It proves the AI is governed: it reads real data, refuses to invent an opportunity that isn’t there, and routes any change through approval rather than acting silently.

Prompt​

Paste into the chat

“Find revenue opportunities at Grand Meridian for this coming Friday. If there are peak slots underpriced, propose rate changes.”

Tool chain​

find_revenue_opportunities → peak-rate scan, then balanced scan

get_availability → Friday availability check

propose_rate_changes → pending action (only if justified)

Expected output​

  • The peak-rate scan returns zero opportunities — Friday is wide open, so there is no demand pressure to justify a markup.

  • The agent declines to raise rates, correctly refusing to manufacture an opportunity.

  • Any change it does surface comes back as a proposal for approval — never a silent edit.

The whole thesis in one moment

A novelty chatbot would have happily produced a confident-sounding rate increase. ZenSpace’s AI looked at the real data, found no justification, and said so — then routed anything actionable through approval. The AI proposes; the platform decides; everything is logged. That restraint is the product.

Verify

Confirm no rate was silently changed. A proposed change must appear as a pending action with before/after rate, expected impact, and an approver — not as a completed edit. Check the server log to be sure.

RUN IT IN ORDER