WIRED REPORTS

“Conversational analytics” is the next frontier in payments

5 minute read
 
It may point to a future where AI agents work alongside treasurers 

IDEAS BANK

Last summer, researchers from Harvard Business School conducted an experiment1 with consultants at a major professional services firm. Participants were given 18 different tasks designed to mimic work at a top consulting company. The consultants using generative artificial intelligence (AI) outperformed their colleagues by every measured metric, completing tasks, on average, 25 percent quicker while producing 40 percent higher-quality results.

The ability for generative AI to help solve domain-specific corporate problems through a natural language interface creates obvious opportunities. One that is attracting particular interest is the notion of “conversational analytics”—using plain language to query data, create reports, or run complex analysis without having to write code or use technical software. 

Conversational analytics is potentially a powerful tool for payments professionals. “The idea is to train generative AI to be part of a corporate treasury team,” says Tony Wimmer, Head of Analytics and Insights at J.P. Morgan Payments, who is building a conversational analytics assistant for corporate treasurers. “We want to engineer it to understand the job of a treasurer. So when the treasurer asks certain questions, it knows the right answer.”  

Tools like this may also let corporate treasurers with no coding knowledge instantly translate plain language questions into the equivalent database queries, gaining insights—complete with accompanying charts and graphs—that it might currently take data analytics teams many days to produce.  

Related Solution
Analytics and Insights Solutions
Make more informed business decisions and manage your bottom-line. Our data solutions help you harness all your data into actionable insights while helping you navigate rapidly-changing economic environments with confidence.

These tools will continue to evolve. Key players in the tech industry envision a new kind of AI assistant, often referred to as an “AI agent,” that can not only operate as described above, but also carry out multi-stage actions autonomously. One company founded by AI industry veterans recently launched an AI agent that can perform office tasks such as reading and replying to users’ emails; another new start-up has showcased an agent that can handle many of the tasks of a software engineer, fully managing coding projects from writing to deployment. One can imagine potential applications of AI agents for treasurers: Perhaps it could optimize cash flows, troubleshoot accounting problems, or simply look after everyday admin. 

The technology is not perfect. Generative AI is prone to “hallucinations”—errors, essentially—and these tools will need to be rigorously tested and monitored by humans as they evolve. And whether you think we will ever see general-purpose AI agents that can function as a true, fully fledged “co-worker” is largely a matter of whether you think AI will come to match human intelligence. But it doesn’t seem like a stretch to imagine that, for certain professions and tasks, we may have specialized agents who can lend support. And it seems like even less of a stretch to imagine conversational analytics going mainstream in a way that lets payments professionals work more effectively. 

“If we get this right, then that’s the productivity gain that people are currently talking about,” says Wimmer. “Then humans can focus on the more value-adding tasks of making decisions, aligning people, driving culture, solving organizational issues: The things that really only humans can do.”  

SOURCES: AS PER WIRED, MAY 2024

FUTURE CAPABILITIES OF CONVERSATIONAL ANALYTICS ARE UNDER DEVELOPMENT; FEATURES AND TIMELINES ARE SUBJECT TO CHANGE AT THE BANK’S SOLE DISCRETION.

ILLUSTRATION: GARY NEILL

Wired