How It Works
The Question You Bring
You begin with a problem you want stress-tested — a business idea, a go-to-market strategy, a procurement decision, a project proposal. You don't need a polished presentation. You need candid input from perspectives that aren't already invested in the outcome.
You submit your question through the Candid Sail portal. Along with the question itself, you can provide background facts you consider reliable and any additional context you want the panel to work from. That package becomes your Brief — the starting point for every model in the discussion.
What Happens Next
Once your Brief is submitted, the Candid Sail system routes it to a panel of AI models from different providers. These are not chatbots running in a browser. They are frontier models — the same premium systems available through individual subscriptions such as Claude, ChatGPT, Gemini, and comparable services — accessed through their professional APIs.
Each panelist reads your Brief independently and provides an opening position. No panelist sees another's response at this stage. The goal is genuine first-impression analysis, not consensus-building.
After each model has spoken, the discussion phase begins. Panelists read what the others have said and respond. Depending on the format you select, this may take the form of structured critique, open debate, or a pro-and-con exchange. Each response goes on record. Nothing is paraphrased or summarized away.
What the Moderator Produces
A separate AI model — the Moderator — does not participate in the debate. Its role is to read the full transcript and produce the report you receive.
The Moderator leads with a summary of the discussion: where the panel agreed, where it diverged, and what questions the disagreement raises. This is followed by the full transcript — every panelist's opening statement, every round of response — so you can read the reasoning behind the conclusions, not just the conclusions themselves.
The report does not tell you what to do. It shows you what a set of genuinely independent, well-informed analytical systems said about your question when given the same facts and asked to reason in the open.
Why It Takes Time
Candid Sail is not a chat interface. Each report involves multiple sequential calls to multiple external AI providers, moderation, and structured assembly of the transcript. A typical report takes between fifteen minutes and a few hours depending on format and provider response times.
When your report is ready, you receive a notification and can retrieve it from your account.
The Outcome
You get a document you can read, share, and return to. Not a chat session that disappears. Not a single model's opinion formatted to look like a debate. A structured record of how multiple independent AI systems reasoned through your question — including where they disagreed and why.
For investors, that means a due diligence input that surfaces assumptions your own analysis may have shared. For program managers, it means a second opinion that didn't come from the same team that wrote the plan. For anyone making a decision with real consequences, it means the kind of friction that makes decisions better.
The Process
1. Input
Upload your problem statement and question. Select the AIs for your panel.
2. Planner
The planner performs grounding research to prevent AIs from fabricating sources.
3. Director
The director runs the panel and moderator manages the flow of guest AI interaction.
4. Producer
The producer and moderator perform analysis and prepare the PDF report.