Investment firms using AI for due diligence and research face a specific problem: a single AI model's output reflects one training dataset's worldview, one set of recency biases, and one interpretation of contested data — presented with uniform confidence. When that output informs a $10M–$500M decision, the cost of invisible analytical blind spots is not theoretical. AI Consensus addresses this by running due diligence questions through seven leading AI models independently, forcing cross-review that surfaces factual disagreements, and producing a confidence-scored research brief with a full conflict audit trail.
Single-model AI output introduces two compounding problems for investment research. The first is confirmation bias amplification: if a thesis is plausible, most AI models will confirm it, because they are optimized to produce helpful, coherent responses rather than adversarial scrutiny. The second is recency bias — each model's training data has a cutoff, and the model cannot flag when its information on a specific market, regulatory environment, or company is outdated without prompting.
The more fundamental problem is that investment teams cannot distinguish a high-consensus finding from a contested one. A single AI model produces a confident-sounding answer regardless of whether the underlying facts are settled or actively debated across different knowledge bases. The confidence score AI Consensus produces is the signal that makes this distinction visible.
Upload a thesis document, market analysis, or due diligence question. In Phase 1, all seven models generate independent assessments — stress-testing the thesis from seven distinct analytical starting points simultaneously, with no anchoring. In Phase 2, each model evaluates the others' Phase 1 positions and must explicitly state where it agrees or disagrees. Factual conflicts are automatically flagged and attributed. In Phase 3, Gemini synthesizes all positions into a BLUF-formatted research brief with a confidence score and dissenting notes.
For document-intensive work, Precision Mode requires all models to cite specific passages from uploaded materials using [Ref: "exact quote"] format, with any claims not supported by the document labeled [External Knowledge]. This produces a research brief where every claim is traceable to its source.
The Decision Memo export contains: a BLUF conclusion (the recommendation in the first sentence), the confidence score with the computation basis, dissenting model positions with attribution, supporting evidence with citations, and recommended next steps. This format is designed for direct inclusion in board materials, LP letters, and investment committee packages — it reads like an analyst memo, not a chatbot response.
Individual analysts can run the free tier (5 models, 3-phase consensus, TXT export) to evaluate the platform on real questions. Professional ($299/seat/month) unlocks all 7 models, Council Mode, Precision Mode, and the Decision Memo export. Enterprise (from $2,500/month) adds API access for programmatic integration with research workflows, SSO, and compliance documentation. Full pricing details →
Run a live analysis on a real investment question from your portfolio. 30-day pilot, scoped to your team.
Request Enterprise Pilot