FactSet looks to build on portfolio commentary with AI

Chris Ellis has seen the waves of innovation in portfolio analytics up close. In the near 30 years he’s spent at FactSet, he’s witnessed offerings in the space evolve as asset managers and wealth managers, too, have broadened the range of asset classes in which they invest, and their portfolio considerations.

“We started doing portfolio analytics in roughly 1997 to 1998, and one of the foundational tenets from the very beginning was achieving a deeper understanding, a more holistic understanding, of the portfolio,” Ellis tells WatersTechnology. “We do absolute return analysis, relative return, which is performance attribution, risk analysis, ESG analysis, and weights and composition.” What started life as offerings in equities have since expanded to fixed income and private capital, with tools and applications that Ellis says provide the flexibility to bend, twist, and contort the analysis to meet clients’ needs.

Interpreting charts and tables is one thing, but most people take in information best when reading it. This is where portfolio commentary comes into play.

“Portfolio commentary is a fundamental thing that every asset manager needs to do,” says Ellis, who serves as executive vice president and head of strategic initiatives at FactSet. “You need to do it for your clients, you need to do it for your consultants, you need to do it for senior management. Every asset owner needs to explain the performance of their internal strategies and their external managers to senior management and to the boards that they are beholden to.”

Traditionally, in order to write that commentary, a platform like FactSet could offer the charts, tables, and numbers required. But with artificial intelligence, a user can go one step further in the process.

We’ve seen so many commentaries come through from different clients, we were able to piece together a sort of collective view of what we thought a good standard model would be

Chris Ellis, FactSet

FactSet is now rolling out a new solution centered around making the attribution summary workflow in portfolio analysis more effective. The Portfolio Commentary tool aims to make the process of writing attribution summaries less time consuming and challenging for users. Using generative artificial intelligence (GenAI) and large language models (LLMs), users can now generate these summaries at the click of a button in a time considerably shorter than before. With traceability a top concern among developers of GenAI systems, all numbers referenced in the summaries are linked to their original source. Additionally, users can tweak sentences and add lines to match their own style.

Users can create four types of insight. An executive summary gives general information about how the portfolio has performed, providing information around benchmarks and sectors from a wider perspective. The “stepping in” summary offers information around portfolio performance drivers, including individual companies and their results. “Stepping back” provides information around market movements and can look across sectors included in the portfolio to analyze returns. Finally, the “sub-period analysis of trends” summary measures how selected stocks perform month-to-month and tracks their performances relative to chosen benchmarks, such as Russell 1000 Growth benchmark.

Something that once took 25 minutes to write and compile has now been whittled down to 30 seconds, according to client feedback received during the solution’s beta-testing phase.

Last November, FactSet laid out its AI blueprint on the heels of the heightened attention being paid to GenAI and LLMs across the capital markets. The blueprint will serve as a product roadmap for how FactSet aims to build on its artificial intelligence capabilities across current offerings, workflows, and new products. It also lays out the provider’s three-pillar approach to AI: mile-wide discoverability, mile-deep workflow automation, and mile-high innovation acceleration.

Ellis says Portfolio Commentary adheres to that blueprint. “We’re really building the foundational tools to be able to do things like this,” he says. “We’re in the business of enhancing the productivity of our users, helping them to be more efficient, and freeing them up to focus on improving their performance, and building stronger relationships with their clients.”

To build the new tool, developers took attribution reports from dummy portfolios and wrote commentaries by hand. These were then used to train a model to replicate and improve on them. “We’ve seen so many commentaries come through from different clients, we were able to piece together a sort of collective view of what we thought a good standard model would be,” Ellis says.

Portfolio Commentary is accessible through FactSet’s Portfolio Analysis solution within the FactSet Workstation. 

Kristina Karnovsky, FactSet’s chief product officer, told WatersTechnology in November that groups across the company were working on workflows for investment banking, the buy side, wealth advisors, and corporate users. Those workflows then become part of the FactSet Explorer program, where select groups of clients can preview forthcoming solutions in beta.

In December, the vendor launched FactSet Mercury, an LLM-based knowledge agent, into its FactSet Explorer product preview program. The beta offering allows users to access FactSet data as well as regulatory information through a chat interface. As an example, the interface could be used to prepare for earnings call season by pulling up year-by-year commentary on companies.

For Ellis, an incremental approach to AI is key: “To use a Risk analogy—as in the board game—I don’t want to parachute into no man’s land, and then work my way back to what we do.” 

That incremental approach appears to be developing into a common gospel across the industry. Symphony CEO Brad Levy told WatersTechnology earlier this month that despite the significant hype that exists around AI right now, it will be the incremental steps taken with the technology that will prove most fruitful for the industry. “It doesn’t have to do that much,” he said. “It just has to solve for incremental problems over time, which is finding things that are hard to find.”

But this is merely the beginning for companies making strategic decisions on where to invest in AI. “After doing this 29 years, you’re never done,” Ellis says. “If you think you’re done, it’s a mirage, and you just haven't figured out what you need to do next. So the journey keeps going, and I think that’s especially true with the technology here.”