No single AI model is the best choice for every task in a hedge fund's workflow. Sonnet may reason best today. Gemini may handle structured data best tomorrow. An open source model may be the only option that meets your data residency and compliance rules next quarter. Unique AI Conduct is built around that reality.
Conduct works like a conductor for your AI agents. It runs curated skills powered by any code execution model, and it swaps the model behind each skill based on the task, the model's current capabilities, and your environment. It can run in the public cloud, in a regionally restricted setup, or fully on-premise.
Take a portfolio manager at a long/short equity fund who logs on at 7:15am and needs a full picture of the portfolio before the market opens. Normally that means switching between FactSet, Bloomberg, an inbox, and broker PDFs. With Conduct, it happens in one place, through five skills.
- Portfolio snapshot. A portfolio comparison skill pulls long and short positions from the internal holdings database and maps them against FactSet consensus. The output is a table showing positions against the street, earnings surprises, and consensus shifts since yesterday. Sonnet handles this step, it works well with structured data and tabular reasoning, and only market-level consensus data is involved, so no proprietary position data leaves the environment.
- News and events scan. A news and events skill pulls from FactSet as its primary source and scans for headlines, guidance changes, M&A activity, and regulatory filings across the portfolio, with web scanning as a secondary layer. It also surfaces upcoming earnings dates, investor days, and regulatory decisions. Scanning this volume of public information calls for a token-efficient model, and an API-based model like Sonnet filters signal from noise well here.
- Broker research ingestion. A research ingestion skill connects to email or a research management system and scans overnight broker notes, earnings previews, initiation reports, and price target changes, extracting the key takeaways. This step draws on Conduct's connector capability: email, Slack, and semantic document search, pulling research from wherever it lives.
- Report generation. A reports generation skill consolidates everything into a single output: portfolio delta summaries, news and event highlights, broker research, and flagged headlines, formatted as an Excel workbook or a PDF in the fund's template. This step aggregates proprietary position data with everything collected so far, and it's structured so the PM can scan it in under two minutes.
- Draft communication. A communication drafting skill turns the enriched research into a briefing email for a team or an investor. For a straightforward task on non-sensitive content, Conduct uses a lighter, faster model, Haiku 4-5, which keeps cost down without slowing the workflow.
In under ten minutes, the PM goes from log-on to fully briefed: portfolio analytics, news and event screening, broker research, a formatted morning report, and a draft communication.
Speed matters, but the real value is control. At every step, the model was chosen deliberately: Sonnet for structured reasoning and token-efficient news scanning, an on-premise model when proprietary data was involved, Haiku for fast, low-cost drafting. The skills stay the same. The workflow stays the same. The model adapts. When a better model appears or your compliance requirements change, you swap the model and keep working.
Unique AI Conduct: Swappable Intelligence for Hedge Funds
No single AI model is the best choice for every task in a hedge fund's workflow. Sonnet may reason best today. Gemini may handle structured data best tomorrow. An open source model may be the only option that meets your data residency and compliance rules next quarter. Unique AI Conduct is built around that reality.
Conduct works like a conductor for your AI agents. It runs curated skills powered by any code execution model, and it swaps the model behind each skill based on the task, the model's current capabilities, and your environment. It can run in the public cloud, in a regionally restricted setup, or fully on-premise.
Take a portfolio manager at a long/short equity fund who logs on at 7:15am and needs a full picture of the portfolio before the market opens. Normally that means switching between FactSet, Bloomberg, an inbox, and broker PDFs. With Conduct, it happens in one place, through five skills.
In under ten minutes, the PM goes from log-on to fully briefed: portfolio analytics, news and event screening, broker research, a formatted morning report, and a draft communication.
Speed matters, but the real value is control. At every step, the model was chosen deliberately: Sonnet for structured reasoning and token-efficient news scanning, an on-premise model when proprietary data was involved, Haiku for fast, low-cost drafting. The skills stay the same. The workflow stays the same. The model adapts. When a better model appears or your compliance requirements change, you swap the model and keep working.