Franklin Templeton bets on OpenAI's Codex to rebuild investment research

Franklin Templeton is using OpenAI's Codex to build AI workflows that help portfolio managers compare research views and examine the risks behind investment ideas, starting with equity research. The firm frames the shift as more than automation. It says AI has moved from supporting a human-led process to surfacing insights that portfolio managers act on.

The stakes are real for asset management. Research, risk, compliance, and trading each hold a different piece of the picture on any investment, and firms that can fuse those inputs faster can pressure-test ideas before markets move. Franklin Templeton is betting that a governed AI layer, not a single chatbot, is what makes that fusion work at scale.

What changed at Franklin Templeton?

The first attempt didn't work. The team loaded its data into a database and connected the models to it, but the models failed to answer the teams' questions. The reason, as the firm tells it, was structure: the models didn't understand how the data fit together.

The fix came in two parts — agents plus a semantic layer, meaning a structured model of what the firm's data means and how it connects. Once those were in place, the results changed. The team now uses Codex and other tools to build on that foundation.

"We've been able to build what we think of as a connected intelligence layer," a Franklin Templeton leader identified in the source as Dudley says. "We're bringing the data into the tools, and we're surfacing insights that make a difference."

The takeaway: plugging a model into raw data wasn't enough; the firm had to teach the system what the data means before it produced anything useful.

Why the 'intelligence layer' matters

Dudley describes the payoff in blunt terms. "The biggest unlock from using OpenAI's tools is the ability to merge continuous intelligence with human judgment," she says. "It has really transitioned because, beforehand, AI was automating our human-led process. Now AI is actually automating the ability to surface intelligent insights, and that intelligence layer plus human judgment is just a fantastic unlock."

The earlier equity-research workflow was human-led and sequential. An analyst did first-pass stock analysis, reviewed earnings calls, and studied competitors in each business line. The goal of the new setup is to surface useful information as analysts prepare ideas to pitch to portfolio managers, or as they update existing positions in the portfolio.

"It's just continuous intelligence combined with that human judgment," Dudley says.

The ambition goes further than faster analyst notes. Franklin Templeton wants its investment teams to work from a governed, reusable intelligence foundation that combines trusted data connectors, portfolio and workflow context, and agentic capabilities that surface actionable insights for analysts and portfolio managers. Research, risk, compliance, and trading inputs would all flow into one place.

"We're able to bring that all in and then use that to surface insights to the people who're making decisions, which are our portfolio managers," Dudley says.

How did the firm get from experiment to production?

Franklin Templeton has run hackathons with Codex, pairing experts alongside investment practitioners as they build their own tools. A group the firm calls its Investment AI champions is working to turn those experiments into production-ready workflows — with security review, expert support, and data controls built into the path.

Dudley frames the transition as a rethinking, not a bolt-on. "We have moved from augmenting our research process to reimagining the investment research workflow," she says.

The firm's leadership has set explicit expectations for what experimentation should deliver. "Our message from leadership is consistent: we encourage innovation and experimentation with AI within clearly defined risk guardrails. Not every initiative will succeed, but each should generate insights that strengthen our capabilities, inform future decisions, and drive continuous improvement," Dudley says.

Franklin Templeton's distilled best practices:

• Get new tools into people's hands. Dudley says the firm has brought practical labs to teams quickly after new technology launches, so they can test how it fits their work.
• Help teams learn the tools. "The goal is getting them to actually understand the power of the technology," Dudley says.
• Build with appropriate controls. Security review, expert support, and data agreements are part of the firm's path from experimentation to production-grade tools.

In short: fast access, real training, and hard controls — the same discipline the firm applies to investments applied to its own tooling.

Why personalization is a problem here

Many AI tools learn a user's preferences and show more of the same. If an analyst favors one research provider, the tool surfaces more of that provider's reports. In consumer settings that's a feature. In asset management, that narrowing risks baking a biased perspective directly into the investment process.

Franklin Templeton wants the opposite behavior from its systems. "We want to make sure that they're accessing that diversity of research and diversity of opinion," Dudley says.

The firm also wants its AI to foreground risk, not just the bull case. "We need AI to understand that a great investment idea is one that is well reasoned, but it also has analyzed the risks and makes our portfolio managers aware of the risk," Dudley says. "If you're looking at a research note, focusing on the fact someone has identified the investment risks is not a bad thing from our perspective. It makes for a more informed investment decision."

The point matters beyond one firm. Recommendation engines optimized for engagement can quietly shrink the set of viewpoints an investor sees, and Franklin Templeton is explicitly engineering against that failure mode.

What comes next?

Dudley's longer-term goal is to make the connected intelligence layer accessible across investment teams — from analysts up to the leaders of each investment group.

She also wants the technology to push people back out into the field. Analysts should spend more time speaking with different experts and professionals, including executives at private companies, and bring what they learn back to their teams. For Dudley, that human perspective remains essential as the tools develop.

"With OpenAI, our teams feel like they're at the forefront of technology," she says.

The direction of travel is clear: Franklin Templeton treats AI as a persistent intelligence substrate under human judgment, with risk awareness and viewpoint diversity engineered in — and it expects analysts to spend the time they save doing the human work models can't.