AI Safety & Governance

Two agents disagree on the fix,
the arbiter picks the one with the receipts

Consensus Arbiter resolves conflicts between competing agent proposals. When the postgres specialist says "kill the slow query" and the cache specialist says "warm the cache" for the same incident, the arbiter compares evidence quality, blast radius, agent trust scores, and cost. It picks one. The other waits as a fallback. Disagreement no longer freezes the response.

Get Started Talk to Sales
app.novaaiops.com / consensus
● LIVE
4
Tie-break dimensions
Sub-second
Resolution time
Fallback
Loser stays armed
Logged
Every arbitration
Four Tie-Break Dimensions

Picked deterministically, not by vibe

When two proposals conflict, the arbiter scores each on four dimensions: evidence quality (does the proposal cite live signals?), blast radius (smaller is preferred for ambiguous incidents), agent trust (the agent's historical track record on similar incidents), and reversibility (a fix you can undo beats one you cannot). The proposal with the higher composite score wins.

  • Evidence quality: live SLIs, traces, or log spans cited > general reasoning > nothing
  • Blast radius: reversible single-service fix > multi-service > destructive
  • Agent trust: recent verified-success rate on this incident type, not global trust
  • Reversibility: idempotent or auto-revertable fixes get a tiebreak bonus
app.novaaiops.com / consensus · scoring
Loser Stays Armed

Second-best is still useful

The losing proposal does not get thrown away. It is held as a fallback. If the winner's ground-truth verifier reports a regression, the arbiter activates the fallback automatically. This is why we want competing specialists: when one is wrong, the second guess is already prepared and reviewed.

  • Fallback held for 24h: losing proposal stays armed until the winner is verified at T+24h
  • Auto-activated on regression: verifier failure triggers the fallback path without waiting for a human
  • Agent ledger captures both: both proposals are bundled, so a postmortem reads "we tried A, A regressed, we fell back to B"
app.novaaiops.com / consensus · fallback
Operator Override

You can always pick the loser

Sometimes a senior engineer knows the second proposal is right based on context the agents do not have. Override is a single click on the conflict view: pick the other proposal, write a one-line reason, the arbiter records the override and uses your decision. Future arbitrations on the same conflict pattern weight your override into the score.

  • One-click pick the other: no escalation chain, no approval gate, your override is final inside the incident window
  • Reason captured: override requires a free-text reason that lands in the Decision Bundle
  • Learned from over time: patterns where humans repeatedly override become tuning signals for the arbiter
app.novaaiops.com / consensus · override
When the Arbiter Refuses

Two close scores plus high blast radius equals human

If the composite scores are within 10 points and at least one proposal has a high blast radius, the arbiter does not pick. It escalates to a human via Approval Queue. Speed is good; speed on a coin-flip with destructive proposals is not. The escalate-on-tie rule prevents agent-on-agent lottery moments.

  • Tie threshold: composite scores within 10 points triggers tie-break review, not arbitrary pick
  • Plus blast guard: tie + destructive proposal = always escalate, never auto-pick
  • Times out cleanly: if no human approves within 15 min, agents fall back to suggest mode and write a runbook
app.novaaiops.com / consensus · escalate
Video walkthrough coming soon

Subscribe to Nova AI Ops on YouTube for demos, tutorials, and feature deep-dives.

Multi-agent without the deadlock

Specialists disagreeing is healthy. Specialists deadlocking on every incident is not. The arbiter is what makes the multi-agent fleet shippable.

Get Started Request a Demo