Help Center / How Consensus Works
How Consensus Works
How does Satcove work?
Satcove takes one question, sends it to several AI models in Consensus mode, compares their conclusions and sources, then presents agreements, disagreements and a practical next step. It is a decision aid, not a guarantee of truth.
How do I get the 6-AI consensus?
By default Cove answers fast with a single model — useful for everyday questions and follow-ups. For an important question, start Consensus: six providers answer independently, Satcove shows where their conclusions align or diverge, and the synthesis recommends a concrete next step. Agreement is not factual verification.
What are the 6 AI models?
Claude (Anthropic) — strong at reasoning and nuance. ChatGPT (OpenAI) — general-purpose, breadth. Gemini (Google) — multimodal, research. Mistral — European, multilingual. Perplexity — live web search, up-to-date data. Grok (xAI) — real-time internet, sharp reasoning. Each brings a distinct perspective.
Why six AI models and not more?
Six models provide useful perspective diversity while keeping latency, cost and the final report understandable. More models are not automatically better; the useful signal is independent agreement and clearly explained disagreement.
How long does consensus take?
Typically 8–15 seconds, depending on the models, source retrieval, attachments and network conditions. You see a live timeline as each model responds in real time. The synthesis is generated after all models have replied.
How is the consensus report generated?
The models answer independently first. Satcove extracts shared conclusions, separates material conflicts, checks whether sources or freshness matter, and writes a synthesis with the next verification step.
What is the agreement score?
The agreement score measures how closely the 6 AI answers aligned in conclusion and direction. It does not measure factual accuracy or the probability that the verdict is true. High agreement means the panel converged; low agreement means the models reached different positions and Satcove will show where. Check sources, freshness, and qualified expertise separately when the stakes are high.
What is the difference between agreement and synthesis?
Agreement describes how closely the models converge. Synthesis is the readable report that combines their answers, preserves important disagreement and turns it into a practical next step.
Is agreement the same as accuracy?
No. Agreement measures how closely the models align, not whether they are correct. Several models can repeat the same outdated or incorrect assumption, so important claims still need current sources or qualified human review.
What should I check before acting on a verdict?
Look at the divergences, the assumptions, the freshness of the information, and whether a primary source or qualified expert is needed. Satcove helps expose agreement and disagreement; it does not replace professional advice or primary evidence.
What should I do when AI models disagree?
Do not choose the most confident-sounding answer. Isolate the disputed claim, inspect the sources and check the current primary evidence. In Satcove, disagreement is a signal that tells you where to slow down and verify.
What is Board mode?
Board mode decomposes a complex project across 5 specialist roles: strategy, finance, technology, operations and research. It then integrates their contributions into one plan while preserving cross-role challenges.
Can I use Satcove offline?
No. Fresh model responses, source checks and consensus reports require a network connection. You can keep exported information locally, but Satcove cannot run a new multi-model query offline.