Endex

Solution

Endex is building an AI Analyst for financial firms by combining several OpenAI models with domain-specific evaluation and retrieval workflows. Rather than leaning on conventional RAG, the system uses reasoning models to retrieve, synthesize, and cross-check financial information the way an analyst would, drawing from firm-specific data, public disclosures, trusted sources, investor decks, Excel models, and filings. OpenAI o1 and o-series models simplified multistep financial reasoning tasks such as precedent transaction analysis, due diligence, IC memo preparation, and model reconciliation, while o3-mini reduced latency enough to make detailed workflows practical. The platform can return outputs as emails, reports, slide decks, or Excel models, and every conclusion is designed to be traced back to its source. Endex pairs this with blind expert testing, FAR benchmark evaluations on tables and charts, and reinforcement fine-tuning so financial professionals receive more referenceable, analyst-style outputs than with non-reasoning models.

Data flow

Internal firm data, public disclosures, investor presentations, Excel models, and other trusted sources are ingested into AI Analyst workflows. Endex routes reasoning across OpenAI models to retrieve facts, detect discrepancies, generate deliverables, and then measures the outputs with FAR benchmarks, blind user tests, and production metrics before returning them to analysts.