Frequently Asked Questions

It places multiple language models into a structured discussion. Models answer, inspect one another’s positions, revise when warranted, and produce a synthesis plus an inspectable record.

Available families may include OpenAI, Google Gemini, Anthropic Claude, xAI Grok, DeepSeek, and others. The exact models shown in the workspace depend on provider availability and configuration.

It is a structured inventory of distinct points raised in the discussion. Each accepted point includes transcript evidence, source attribution, materiality, confidence, and a verification status. It describes the decision space; it does not prove the final recommendation is correct.

Yes. A follow-up supplies the prior synthesis and the new information to another council run, then generates a fresh map reflecting the latest discussion.

Yes. Estimates are calculated from configured token assumptions and published model pricing without calling an LLM. Estimates are approximate rather than fixed quotes.

AI Deliberator reserves prepaid credits before a run and settles the charge from measured token usage. The launch screen shows the estimated customer price, including the application markup. One credit currently represents one U.S. dollar of spending power.

The run may finish, but results can remain unavailable until the outstanding balance is paid. You do not have to approve each small call when your available balance covers its estimate.

The original experiment creator is always the billing owner. An invited collaborator is not charged for that experiment. If the collaborator creates a separate experiment, that person needs their own credits.

The Research area includes datasets only when their identified source terms have been reviewed for the displayed use. You can choose question counts, categories, sampling methods, and seeds. Source licenses still control permitted use.

No AI output should be treated as guaranteed. Models can make factual, mathematical, interpretive, or citation errors—even when several agree. Verify important claims and use qualified professionals for high-stakes decisions.

No. The service is not medical, mental-health, legal, financial, employment, or emergency advice. If someone may be in immediate danger, contact local emergency services or an appropriate crisis resource.

Do not enter passwords, API keys, full payment-card data, protected health information, classified information, highly sensitive personal information, or material you lack permission to share.

Council runs, transcripts, usage records, synthesis, and maps are scheduled to expire after 90 days. Owners can delete an entire experiment sooner from its Data controls.

It goes to the model providers selected for the run, along with relevant discussion context. Clerk handles authentication, Stripe handles checkout, Vercel hosts the application, and the configured database stores workspace records.

Only if you have the rights and permissions needed for the prompts, source material, personal information, benchmark content, and intended publication. You remain responsible for verification, attribution, and legal or ethical review.

Email info@systemanalytics.net. For an experiment you own, the fastest deletion method is the Data controls section in the experiment workspace.