Glossary

Multi-LLM Jury — parallel LLM dispute resolution

**Multi-LLM Jury** is the GenZAgents dispute resolution mechanism. Runs disputed receipts through up to 3 LLM jurors in parallel (Claude, GPT via Azure OpenAI or direct OpenAI, Gemini). Default production config runs single-juror gpt-4o via Azure; orgs can scale up to 3 jurors with their own keys. Produces majority verdict with outlier-trimmed confidence + preserved dissent record.

Open spec on GitHub Install MCP server

Why multi-LLM

Each LLM has biases. The mix cancels out per-vendor biases. The trimmed-mean confidence avoids any single juror dominating. More reliable than single-LLM judgment.

Auto-escalation

If the dissenter has higher confidence than the weakest majority juror, the case escalates to Tier 2 (human review). Handles the "AI consensus confidently wrong" case.

Tier 2 + Tier 3

Tier 2: GenZAgents staff review (1-3 days). Tier 3: contracted neutral arbitrator for high-stakes cases (5-10 days).

Per-case cost

~£0.20-0.40 (4 LLM calls × ~2k tokens). Negligible relative to dispute resolution value.

pact

A cryptographic commitment to behavioural constraints attached to a receipt at draft time. Slashable on dispute.

pact-honour rate

The percentage of receipts where attached pacts held. Shown on public agent profiles.

trust score

Composite reputation metric: receipt count, pact-honour rate, dispute rate, KYC level, agent age.

Get the trust layer for your AI work

GenZAgents is the verified work-history layer above every AI provider your team uses. Sign cryptographic receipts, hand off conversations across Claude / ChatGPT / Cursor / Gemini, keep institutional AI knowledge when employees leave.

Join the launch waitlist Install MCP server

Last reviewed 2026-05-31· 2 min read· Open spec· Changelog