Arbiter: Blind AI Deliberation for Multi-Party Decisions

Group decisions go in circles. Email threads, meetings that run long, the loudest voice winning. Arbiter collects arguments independently, then lets AI deliberate blindly and deliver a definitive verdict.

How it works

Define the question and the evaluation criteria (AI can suggest these). Invite participants via private links. Each person argues their position against each criterion without seeing anyone else's submission. When everyone has submitted, two AI models deliberate independently and produce a ranking with percentages. Always a winner, even if it's 51/49.

AI that can't be sycophantic

The most common criticism of AI is sycophancy: it agrees with you, defers to authority, tells you what you want to hear. Arbiter sidesteps this entirely. The models never see who submitted which argument. They don't know if it came from the CEO or an intern. There is no identity to defer to, no authority to mirror. All they have is substance weighed against criteria.

This is one of the few setups where AI sycophancy is structurally impossible, not just discouraged. The blind architecture removes the input that causes the problem in the first place. Two independent models reasoning separately add another layer, so the verdict reflects genuine analytical consensus rather than one algorithm's quirks.

What you get

A clear verdict with a percentage split, plus a criterion-by-criterion breakdown from each model. You can see which side won on cost, which won on risk, and how the models weighed competing strengths. The reasoning is transparent and auditable. Useful for office decisions, budget allocation, vendor selection, or any situation where multiple stakeholders need to present their case fairly.

Arbiter is also available through MCP, so you can manage decisions from Launchpad or any MCP-compatible assistant. AI usage is included with every plan.

Start your free trial and try it out. No credit card required.