The Morning Download: How Walmart, BNY Scaled AI Agents

What's up: Welcome to the 100-hour workweek, AI workers; Meta cuts 600 jobs in AI division; IBM says its customers are starting to scale AI

Good morning. There are more than 7,000 senior corporate technology leaders in attendance at Gartner IT Symposium/Xpo in Orlando, Fla. Guess what everyone is talking about? Here’s a hint. It’s spelled “A … I.”

On Wednesday morning, I moderated a panel discussion with Vinod Bidarkoppa, executive vice president and chief technology officer at Walmart International, Leigh-Ann Russell, chief information officer and global head of engineering at BNY Mellon, Chris Howard, global enterprise research chief at research and consulting firm Gartner. The topic was Agentic AI: Is it Real? (I thought it would have been apt to call the session Agentic AI: Am I Crazy?)

At the outset, I summarized the bear case against AI agents, which argues that they hardly ever work at scale and that exceedingly few companies manage to drive an acceptable return on such investments. I invited Bidarkoppa and Russell to explain whether they agree or disagree with that assessment, and explain their reasoning. They both maintained that their companies have created numerous agents, scaled them with success and that the efforts have paid off in many ways.

--BNY Mellon is seeing a tangible benefit, with solutions in production across nearly every part of the bank, according to Russell. She said the company has about 100 “digital employees” that possess their own distinct login credentials, communicate via email or Teams, and report to a human manager, a system designed to provide a framework for managing, auditing, and scaling the agent workforce. The company has done broad training in skills related to AI agents, and it continues to go deeper with that effort.

For example, a “digital engineer” scans the code base for vulnerabilities, and can take autonomous action to write and submit a fix for low-complexity problems. In more complex cases, the AI agent will report the problem to a human engineer.

--Walmart's Bidarkoppa said he viewed AI as a "force multiplier" that will lift productivity across all lines of business, particularly in engineering, where significant increases are already visible. He said Walmart is focused on making use of available technology, avoiding significant internal capital expenditure for model development.

Bidarkoppa said Walmart views agents as people-powered tools designed to extend opportunities and make work easier for associates, customers and partners. The goal is to assist, not completely replace, the human element in decision-making, according to Bidarkoppa.

Chris Howard of Gartner and I polled the audience, to get a sense of how many people believed AI agents were driving a reasonable return for their company. I was surprised that only one hardcore skeptic stepped forward, with several dozen hands signaling the affirmative and plenty of hands staying down. I was expecting more pushback on the idea that AI agents are beginning to scale and drive value.

Often, the questions we ask are as important as the answers. Of course, it matters how many companies are scaling AI and believe that it’s worthwhile. The more fundamental question may simply be whether AI agents are capable of functioning in a meaningful way, scaling up and driving a return of some sort. For BNY Mellon and Walmart, the answer is yes.

Revenue at International Business Machines topped expectations in the third quarter, rising 9% to $16.33 billion, as the company reported an uptick in AI activity among customers of its consulting business.

“Ninety days ago, we were prudently cautious about the overarching market and client demand,” Chief Financial Officer Jim Kavanaugh said Wednesday. “But I think you’re starting to see the acceleration.”

About 80% of IBM’s 300 clients who work with the company for AI-related purposes are new over the past two quarters, Kavanaugh said. IBM’s AI book of business increased to $9.5 billion in the third quarter, a $2 billion increase from the second quarter.

Top researchers and executives at Silicon Valley’s biggest AI labs are regularly logging 80 to 100 hours a week, motivated by the competition, their own curiosity and the view that they are working in a seminal moment in history.

“We’re basically trying to speedrun 20 years of scientific progress in two years,” Josh Batson, a research scientist at Anthropic, tells the WSJ. Extraordinary advances in AI systems are happening “every few months,” he said. “It’s the most interesting scientific question in the world right now.”

Certain startups have included the expectation of an 80-hour-plus workweek. But many companies haven’t had to do so, as top AI employees are motivated by the intense competition with rivals and their own curiosity about new model possibilities, says the WSJ.

Working the 0-0-2. The most-intense periods for many come while working on models or new products, when time working extends beyond the “9-9-6” schedule. That stands for 9 a.m. to 9 p.m., six days a week. One startup executive jokingly referred to the schedule as “0-0-2,” meaning midnight to midnight, with a two-hour break on weekends.

First there were tech job cuts to counter the hiring frenzy during the pandemic.  Then came the cuts, coupled with a slowdown in hiring, related to the integration of AI into the workplace.  Now comes word that even AI jobs may not be safe. Meta Platforms is cutting about 600 jobs in its AI division, WSJ reports.

The job cuts will affect the company’s teams focused on artificial-intelligence products, infrastructure and long-term AI research, but won’t touch TBD Lab, the new team of multimillion-dollar hires formed over the summer, according to an internal memo viewed by Journal and a person familiar with the matter.

“By reducing the size of our team, fewer conversations will be required to make a decision, and each person will be more load-bearing and have more scope and impact,” wrote Alexandr Wang, Meta’s chief AI officer.

Most of the affected employees will have the opportunity to be “redeployed” to other jobs at Meta where their AI expertise could be useful, Wang said.

Amazon Wednesday unveiled a trio of new technologies it is testing or preparing to deploy in its warehouses and delivery vans, the latest in a yearslong automation effort, WSJ reports. They include a robot arm called Blue Jay, designed to sort packages; an AI agent called Eluna, intended to help human managers deploy workers and avoid bottlenecks; and augmented-reality glasses to be worn by delivery drivers in the field.

The New York Times earlier this week reported that Amazon executives were hoping that robot adoption would help it avoid having to hire more than 600,000 people by 2033.

Companies including IonQ, Rigetti Computing and D-Wave Quantum are discussing the government becoming a shareholder as part of agreements to get funding earmarked for promising technology companies, WSJ reports.

The companies are discussing minimum funding awards from Washington of $10 million each, some of the people said.

The discussions are the latest example of administration moves to become a shareholder in some companies. In August, the government agreed to take a nearly 10% stake in chip company Intel in exchange for converting almost $9 billion in previously awarded grants to equity. A similar deal exists with one of the few U.S. producers of rare-earth materials.

SAP reported higher revenue and operating profit for the third quarter and guided for operating profit toward the high end of its prior outlook for the year, WSJ reports. Sales from its cloud business increased 22% to 5.29 billion euros.

GE Vernova swung to a profit in the third quarter as orders surged for its power and electrification businesses, WSJ reports. The GE spinoff, together with Japan’s Mitsubishi Heavy Industries and Germany’s Siemens Energy control a good chunk of the market for gas turbines, equipment in heavy demand from data centers powering the boom in AI, the FT reports.

Another player in the great data center build out. Gas turbine orders are forecast to increase to 1,025 units this year, the most since 2011, FT reports, citing data from energy consultancy Dora Partners.

Tesl