Introduction

Learn what AskVerdict is, how multi-agent AI debates work, and why structured reasoning leads to better decisions.

2 min read
Share

AskVerdict is a multi-agent AI debate engine that helps you make better decisions. Instead of getting a single AI opinion, AskVerdict deploys specialized AI agents that argue both sides of any question, cross-examine each other's arguments, and deliver a structured verdict.

How It Works

Every decision goes through a structured debate process:

  1. You ask a question — "Should we migrate to microservices?" or "Which candidate should we hire?"
  2. AI agents debate — A Pro agent and Con agent present arguments, backed by evidence and reasoning
  3. Cross-examination — Agents challenge each other's points, identifying weaknesses and blind spots
  4. Verdict delivered — A Judge agent synthesizes both sides into a clear recommendation with confidence score

Key Concepts

Debates

A debate is a single decision-making session. You provide a question, choose a mode, and the engine runs the full debate lifecycle. Each debate produces arguments, rebuttals, and a final verdict.

Verdicts

The output of a completed debate. A verdict includes:

  • Recommendation — The judge's conclusion (e.g., "Proceed", "Reconsider", "Needs more info")
  • One-liner — A single-sentence summary
  • Confidence score — 0-100 indicating how clear-cut the decision is
  • Full reasoning — Detailed analysis of both sides

Modes

AskVerdict offers three debate modes:

ModeCreditsSpeedDepth
Fast1 credit~15 secondsQuick take, good for simple decisions
Balanced3 credits~45 secondsThorough analysis with cross-examination
Thorough8 credits~2 minutesDeep research with multiple rounds of debate

BYOK (Bring Your Own Key)

Use your own AI provider API keys (OpenAI, Anthropic, Google) for unlimited debates. BYOK keys are never stored — they're used in-memory for the duration of the request only.

Agent Personas

Each debate involves specialized AI agents:

  • Pro Agent — Argues in favor, finds supporting evidence
  • Con Agent — Argues against, identifies risks and downsides
  • Judge Agent — Neutral arbiter, synthesizes arguments into a verdict

What You Can Build

  • Decision support tools — Integrate verdicts into your product's decision workflows
  • Automated analysis — Run batch debates via API for portfolio analysis, content moderation, etc.
  • Research pipelines — Chain debates for multi-step reasoning (e.g., "Is X true?" → "What are the implications?")
  • Slack/Discord bots — Let teams get quick verdicts in their chat tools

Next Steps

Was this page helpful?