AI is changing investment research! Franklin Templeton is redesigning its workflow with Codex
Hi everyone, it's Shii-chan! Today I'm sharing a customer story from OpenAI about an asset manager.
OpenAI News
openai.com
What was announced?
OpenAI News published a story about Franklin Templeton. Starting with equity research, the firm is building AI workflows that help portfolio managers compare research views and examine investment risks. Codex is one of several tools its investment teams use.
Why it matters
It shows how to design AI in a field where decisions carry a lot of responsibility.
What changes
- Equity research used to be human-led and sequential: first-pass analysis, earnings calls, competitor studies
- The goal is for AI to keep surfacing insights and combine them with human judgment
- Separate views from research, risk, compliance, and trading are brought together for portfolio managers
Dive Deep
- At first the team just put data in a database and connected models, but the models didn't understand how the data fit together
- Adding agents and a semantic layer (a structured model of what the data means and how it connects) changed the results
- Hackathons with Codex let practitioners build their own tools
- Moving experiments to production goes through security review, expert support, and data controls
- Tools that learn preferences can narrow what people see, so the firm stresses access to diverse research and opinions
- Its view is that a good idea is one whose risks have been analyzed, so the systems are designed to emphasize risk
Wrap-up
- AI workflows started with equity research
- The semantic layer and agents made the difference
- The design avoids bias and puts risk up front
- Production use goes through proper controls
This is for anyone struggling to connect business data to AI!
