FactSet lays out AI blueprint for discoverability, workflows, and innovation

FactSet is expanding its investment in artificial intelligence and looking to this year’s hottest commodity, generative AI, as the decisive technology that will modernize the way its clients interact with workflows.

The data provider has laid out its AI blueprint on the heels of heightened attention on generative AI and large language models across the capital markets. The blueprint will serve as a product roadmap for how FactSet will look 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.

Through discoverability, FactSet looks to include a new conversational experience for users within its FactSet Workstation offering. “Today, someone might log into their FactSet Workstation and they have a bunch of tabs that they can click on,” says Kristina Karnovsky, executive vice president and chief product officer at FactSet. “Maybe one is for research, another for news, and it’s an experience in navigating to the type of data or type of task you are trying to accomplish.”

Instead of navigating the platform to determine where a certain piece of content is, Karnovsky says the future experience will be simply to ask a question and have FactSet surface the answer or guide the user to the right place. If a user is looking for data that FactSet might not have, the interface can also answer with multiple datasets to give a more comprehensive answer.

Conversational interfaces on platforms are popping up more and more. S&P Global rolled out a new version of its Capital IQ Pro workstation this year, which integrates the wealth of data acquired via the vendor’s purchase of IHS Markit, along with new charting resulting from its acquisition of ChartIQ. The vendor plans to add a generative AI search interface called ChatIQ by the end of the year. Outside the data provider world, Broadridge’s LTX electronic bond trading platform is leaning heavily on its BondGPT offering for pre-trade data to set it apart among trading platforms.

Auditability is a core focus of the conversational experience, Karnovsky says. “It’s all great, it feels like magic, you’re asking your question and you’re getting an answer,” she says. But in regulated markets, not being able to validate information and prove its accuracy on the spot means no time was really saved. Karnovsky says users will be able to give turn-by-turn directions to how the answer was retrieved with additional source documents when the answer is surfaced. “We weren’t about to just release a chatbot with answers that no one could validate because that would have been a rookie move in this space,” she says.

Karnovsky says the conversational experience will “feel like a chatbot,” but it will also serve as the launching point for some of the workflow automations. After asking the question and receiving an answer, a user can then ask to have their pitchbook updated with the information, as an example. FactSet will use the knowledge it has about the personas that use its solutions (investment bankers, portfolio managers, wealth advisors, etc.) to simplify the laborious tasks that they face. That could be updating a pitchbook, getting a summary of portfolio performance, or generating investment proposals.

Through the last pillar, innovation acceleration, FactSet is looking to make integration more seamless with fewer barriers. A client could look to build their own conversational experience on top of their own data while leveraging pieces of what FactSet has built. That could be integrated via APIs, cloud platforms or CRMs. “Not all of our competitors are taking the same approach, and that’s what we have heard from clients,” she says.

The discoverability portion of FactSet’s strategy will be accessible to a select group of clients in December when the provider kicks off its FactSet Explorer product preview program.

The workflows are currently in development, with clients giving feedback to shape the offerings. Groups are working on workflows for investment banking, the buy side, wealth advisors and corporate users. The workflows and innovation acceleration components of the blueprint will then become part of the FactSet Explorer program.

While the hype around generative AI potential hasn’t appeared to die down, some are cautious about what firms should be considering when implementing strategies. Jeremy Stierwalt, senior partner and US head of data and analytics at Capco, says starting small might be the optimal road to take. “Much like any other new capability, fail fast [and] fail cheap because we’ve seen clients that are broadly going after it in a much larger capacity and not put the right controls in place,” he says.

To that point, data provider strategies around generative AI have so far largely started with small steps. In March, Bloomberg became the first data provider to announce its own large language model, BloombergGPT. Bloomberg is looking to have BloombergGPT assist—and provide more depth to—the Terminal’s sentiment analysis, named-entity recognition, news classification, charting and question-answering capabilities, among other functions. Or, more simply, Bloomberg hopes it will become the engine that will supercharge its Terminal.

Frank Tarsillo, chief technology officer at S&P Global Market Intelligence, told WatersTechnology in July that the vendor was looking at domain-specific LLMs. Others, like Moody’s, are looking to partner with Big Tech to power their strategies. Moody’s announced a partnership with Microsoft in June focused on generative AI that will look to enhance its current offerings, one of which is Moody’s CoPilot, an internal tool that will combine proprietary data and research under Moody’s Analytics with LLMs and AI technology that sits under Microsoft.

Karnovsky says FactSet has leveraged offerings from third-party providers, which she declined to name. However, she adds that FactSet is also considering training its own models. “Because of the vast repository of data that we have, we do see opportunities to train our own models and fine tune them on our own datasets,” she says. “That’s definitely part of our roadmap going forward.”