Amanda Stent dived into natural language processing (NLP) when there were barely 20 people in the field. Now, she leads AI initiatives at Bloomberg.
After earning her PhD in NLP from the University of Rochester in 2001, she focused on natural language generation long before it became trendy. “Someone once joked, ‘You’ve been doing AI for a very, very, very long time,’” she recalls with a laugh.
Stent began her career in research at AT&T, working on speech and language processing within AI applications. One of her projects aimed to automate news story generation from structured data, like stock price changes—this was in the early days of what we now call generative AI.
After her time at AT&T, she joined Yahoo Labs until the company was sold to Verizon in 2016. Then, she stepped into Bloomberg’s tech office, focusing again on NLP. Stent later spent three years at Colby College, where she helped launch the nation’s first undergraduate AI program before returning to Bloomberg to head AI strategy and research.
Stent emphasizes the need for professionals informed about AI beyond engineers. “The world needs AI-savvy product managers, journalists, and salespeople,” she states.
Today, she heads AI strategy at Bloomberg, which has been integrating AI since 2009. Initially, they offered clients market sentiment models. “Now, our major focus is on generative AI and using it to tackle real client challenges,” she explains.
Bloomberg processes three data types: structured data (like price time series), unstructured data (news and company reports), and communications data. “Generative AI helps clients extract actionable insights from this data,” Stent says. One of their early successes was summarizing earnings calls. Instead of generic summaries, they deliver tailored content focusing on 13 key categories that matter to clients. Users can click through for specifics within the call, a method Stent calls transparent attribution.
Bloomberg also uses GenAI to distill lengthy news stories into bullet points and allows clients to query documents for relevant information. “All Bloomberg terminal users have access to GenAI, and some are crafting their own solutions,” Stent shares. “There are clients eager to dive deep into GenAI and even collaborate with us, while others are more cautious.”
As generative AI reshapes society, Stent stresses the importance of understanding its risks, particularly AI hallucinations. “In finance, it’s crucial to ensure your model stays updated,” she warns. “You need to track the sources of the information GenAI uses, as it can create misleading content.”
Staying objective in financial decisions is vital. “Whenever a decision affects someone, accuracy is key. Systems offer consistency, but that doesn’t guarantee objectivity,” she notes.
At Bloomberg, 350 AI engineers contribute, but many others from diverse backgrounds play critical roles. Stent highlights the importance of subject matter experts for data input and AI-informed product managers for shaping development.
The skills needed at Bloomberg range from solid mathematical expertise to an understanding of client challenges. “We can teach effective AI usage without requiring in-depth math skills,” Stent assures.
Bloomberg’s AI engineers often hold PhDs in fields like physics or computer science, while subject matter experts have advanced degrees in finance. Sales and product managers often come from Bloomberg’s client side, adept at translating needs into actionable tasks.
Stent believes the rise of AI will transform job landscapes, as history shows no revolution lacks job transformation. “AI will augment many roles,” she asserts, noting that while some tasks may fade, new roles like prompt engineering will emerge.