How Hedge Funds Are Using Agentic AI: Insights from the Unique x Kadoa Dinner

• Portfolio managers are inundated with data and research from a growing number of sources – internal systems, public data, sell-side research, and more.
• Agentic AI and platforms like Unique allow users to deploy autonomous agents that synthesize and surface key insights – transforming hours of manual scanning into actionable summaries.
• This kind of AI acts as a research analyst that never sleeps, accelerating the path from raw information to decision-making.

• Broad LLM access or generic chat interfaces are not practical for portfolio managers operating in high-stakes, time-constrained environments.
• What resonates are tailored workflows – e.g., parsing 10-Ks for ESG signals, pulling comps from internal models, or generating earnings summaries aligned to house style.
• The future is AI that molds to the firm’s existing process – not the other way around.

• Hedge funds are experimenting with AI agents that extract insights from the web – but signal quality is a limiting factor.
• Participants emphasized that poor data leads to poor outcomes – so model outputs are only as strong as the underlying content.
• There’s growing interest in curated pipelines, clean web scraping, and dynamic context management to ensure fidelity in outputs.

• No firm is comfortable putting sensitive strategy docs or investor reports into models that they don’t control.
• There was strong alignment around the need for enterprise-grade safeguards, including:
  • Full control over which documents can be read or stored
  • Assurances that data isn’t used for model training
  • Options for private model hosting or API key isolation
• Trust is the foundation of AI adoption in finance.

• With major model releases happening weekly (e.g., GPT-4o, Claude, Gemini), funds want optionality – not lock-in.
• The idea of model routing and interchangeability resonated: dynamically choosing the right model for the right task, based on cost, performance, or latency.
• Forward-looking teams are designing their AI infrastructure to be modular and flexible, not tied to a single LLM vendor.

• Agent Collaboration: Some teams are testing “agent chains” (multi-step agents working together), e.g., a model that ingests earnings transcripts, identifies macro themes, and generates visual dashboards automatically.
• Internal Data Integration: Connecting LLMs to firm-specific data (internal notes, research memos, CRM entries) is seen as the next frontier – especially when permissions and context can be managed precisely.
• Change Management & Adoption: A few participants highlighted the cultural challenge of AI adoption – PMs are skeptical by nature and need demonstrable ROI before trusting a new tool.

To Sum It Up

What stood out most in our conversation with hedge fund and asset management leaders was this: AI in finance isn’t about chasing the latest trend, it’s about solving real problems in real workflows. Portfolio managers don’t need another flashy interface; they need tools that think like their teams, speak their language, and fit into their daily rhythm.

The future of AI in this space will belong to those who get the details right: clean data, secure systems, tailored use cases, and the flexibility to evolve. Get that right, and AI becomes more than a tool, it becomes a trusted partner in the investment process.