By Steven Rosenbush
Few companies today have widely deployed AI agents and gotten value back, one of many reasons that concerns are growing over the astronomical spending on artificial intelligence. But don't overlook those early adopters who say the payoff is looking good.
The pack could appear prudent in the not-so-distant future for waiting to see how agents mature, or the first movers might gain a long-term, sustainable advantage. I have tried to be open to either outcome, or some combination of the two. It could be that the vanguard crew, the fast followers and the laggards all have good reasons for taking different approaches, and that as long as they execute well on their own time frames, things will work out.
But now, as more AI agents come out of the pilot stage, there is evidence to support the early adopters. I had the opportunity to speak with two of them last month at the Gartner IT Symposium/Xpo in Orlando, Fla., where I participated in a panel pointedly called "Agentic AI: Is it Real?"
"It's very, very hard to build an agentic framework," said Leigh-Ann Russell, chief information officer and global head of engineering at financial services provider BNY, during our conversation on stage. "But we have 117 solutions [including agentic] touching everything that happens at the bank and we're seeing really, really tangible outcomes that impact our bottom line in terms of growing capacity."
In perhaps the most dramatic example, Russell said the company has about 100 "digital employees" that possess their own distinct login credentials, communicate via email or Microsoft Teams, and report to a human manager, a system designed to provide a framework for managing, auditing and scaling the agent "workforce."
One "digital engineer" at BNY scans the code base for vulnerabilities, and can write and implement fixes for low-complexity problems.
The agents are built on top of leading models from OpenAI, Google and Anthropic, using additional capabilities within BNY's internal AI platform Eliza to improve security, robustness and accuracy.
Walmart uses AI agents to help source products, informed by trend signals such as what teenagers are buying at the moment, according to Vinod Bidarkoppa, executive vice president and chief technology officer at Walmart International, and another panelist.
In the fashion industry, it takes around six months to design and produce a collection of clothing, from ideation and concept building to the day items end up in stores and online. Walmart's Trend-to-Product agent shortens the traditional production timeline for Walmart fashion by as much as 18 weeks. Trend-to-Product has helped create specifications and patterns for products such as Walmart's No Boundaries Off-Shoulder Mini Dress.
AI is a "force multiplier" that will lift productivity across all lines of business, particularly in engineering, where significant increases are already visible, but not a means to replace human decision-making, Bidarkoppa said. "This is real value, if all of these agents are working together," he said.
Some of that work is also behind the scenes, which might help explain the general sense that we are still waiting on agents to deliver more value. "Walmart is a good example of a customer now working with us on the commerce side, but also using a lot of our technology internally around things like how to merchandise, how to handle risk and so on, on their site," OpenAI Chief Financial Officer Sarah Friar said Nov. 5 at the WSJ Tech Live conference.
Early studies reporting low adoption rates and nonexistent returns might have created more drag on agents' reputation. These studies weren't necessarily wrong, but I suspect they were snapshots in time that might have lured some into concluding the entire effort was destined to misfire.
BNY's Russell referred to a Massachusetts Institute of Technology study that generated headlines last August because it reportedly found most generative-AI pilot projects had failed to generate meaningful returns. "For us, that's a hard false," Russell said. MIT didn't respond to a request for comment.
There are other signs of agentic life. Software-as-a-service giant Salesforce told me it has seen "5x growth" since the launch of its Agentforce platform last year.
This is just another moment in time, of course, far from the end of the story. Adoption remains in the early stages. Plenty of trial, error and spending lie ahead. But when it comes to the usefulness of AI agents, an important demonstration of efficacy has been established by some first movers in the market.
Write to Steven Rosenbush at steven.rosenbush@wsj.com
(END) Dow Jones Newswires
November 12, 2025 07:00 ET (12:00 GMT)
Copyright (c) 2025 Dow Jones & Company, Inc.
Comments