LONDON — A U.K. commission on Thursday unveiled its recommendations for how the country should regulate artificial intelligence in medicine, as health authorities globally try to determine how to continuously review ever-changing products after they have been authorized instead of simply clearing them once for the market. 

In its report, the commission sought to strike a balance between ensuring that the U.K. takes advantage of AI’s potential in medicine — the ability to review the millions of scans that are generated each year to track the eye health of patients with diabetes, for example — while prioritizing safety, tracking device performance over time, and treating patients equitably.

The 44 recommendations include some that would allow for the staged authorization of new AI models and others that would build a system for providers to report how tools are performing in their clinics — including when they malfunction or potentially inhibit patient care — as the models learn and adapt and perform differently depending on the setting.  

Continue to STAT+ to read the full story…

ARPA-H, the government agency that funds cutting-edge health research, plans to commit $62.7 million to develop artificial intelligence bots that direct treatment of heart failure. Among the goals of the program, called ADVOCATE, is to produce partially autonomous AI devices authorized by the Food and Drug Administration to help treat patients, including assessing symptom severity, prescribing drugs, and ordering lab tests. 

ARPA-H on Wednesday announced the first batch of awards to health tech companies Atman Health, UpDoc, Tempus AI, and teams from Stanford University, Duke University, and the Kaiser Permanente health system. ARPA-H may still fund additional teams. The amount committed for the first year is $33.7 million, and the remainder may be renegotiated up or down. 

Many of the 6.7 million Americans with heart failure don’t get optimal treatment because of difficulty accessing specialists, and the hope is that AI agents developed with ARPA-H funding can help address the gap, especially in rural and other underserved settings.

Continue to STAT+ to read the full story…

You’re reading the web edition of STAT’s AI Prognosis newsletter, our subscriber-exclusive guide to artificial intelligence in health care and medicine. Sign up to get it delivered in your inbox every Wednesday. 

It’s hard to resist the world’s obsession with productivity and efficiency and to force yourself to be present in a quiet space. I’ve been enjoying the wandering_cassettes Instagram account, which posts videos of Fisher Price cassette players playing songs in different settings. This one is particularly calming, but they’re all timeline cleansers.

The newsletter will be off next week as I work on a big new story, but I’ll be back in your inboxes Sept. 23.

AI scribes only fix a tiny slice of the problem

Today’s AI Prognosis — and my new story — are brought to you by “The Pitt.” Not in any financial way — I just mean that I finally started watching “The Pitt.” (No spoilers! I’m only halfway through season two.)

Everyone is right — it is extremely compelling television. But what struck me the most was how “The Pitt” feels exactly like the ambulance ride-alongs and emergency department shadowing visits I’ve done: Dropping in on the worst day of someone’s life, but doing it over and over every 40 minutes, for eight to 12 hours, without enough resources.

Continue to STAT+ to read the full story…

Felipe Huici explains how Unikraft achieves millisecond cold boots, stateful scale-to-zero, and extreme density for sandboxing AI workloads. He discusses isolation primitives, Linux kernel optimizations, and snapshotting tricks, demonstrating how to maintain sub-10ms performance at scale while integrating seamlessly into Kubernetes environments with hardware-level security.

Ezekiel J. Emanuel and Abe Baker-Butler have been debating the proper place for AI in medicine with American Medical Association CEO John Whyte. Now, they are taking their discussion to STAT’s First Opinion. Read Emanuel and Baker-Butler’s essay below and read Whyte’s essay here.

In 1867, Joseph Lister published his research on carbolic acid and antiseptic surgical technique.  In September 1871, he was summoned to Queen Victoria, who had a rapidly growing abscess in her left armpit. Using his antiseptic surgical technique, Joseph Lister successfully drained the pus. Queen Victoria recovered without fever or other complications. The antiseptic technique quickly gained approval in the U.K. and Europe, but not among American physicians.  

Read the rest…

Ezekiel J. Emanuel and Abe Baker-Butler have been debating the proper place for AI in medicine with American Medical Association CEO John Whyte. Now, they are taking their discussion to STAT’s First Opinion. Read Whyte’s essay below and read Emanuel and Baker-Butler‘s essay here.

Would you want artificial intelligence to tell you that you have cancer?

Read the rest…

BOSTON — The building that houses the Brigham and Women’s Hospital emergency department opened in 2022. But by 2025, it was already too small. One early September evening, the emergency room had 59 patients — 152, if you counted everyone in the waiting room. 

“In terms of rooms for acute care, we have 61,” said Christopher Baugh, an emergency physician at the hospital. “You can’t see 152 patients in 61 rooms.” Patient beds lined every possible area, with a full circle of beds surrounding the staff workstations in the center of the bay. Figurines of Star Wars characters R2D2 and C3PO peered over beds 80H and 81H — where “H” stands for “hallway.”

To some, this may look like a scene from another country, Baugh said. But to him, it’s just Thursday.

Continue to STAT+ to read the full story…

OpenAI’s latest mathematical milestone has quickly become mired in controversy. Today, the company announced that its agents have solved one of the Millennium Prize Problems, some of the most important open problems in mathematics. Under normal circumstances, that solution would be a huge feather in OpenAI’s cap.

But the announcement has been overshadowed by accusations that OpenAI used NYU mathematician Tristan Buckmaster’s and Anthropic employee Levent Alpöge’s AI-assisted work on the problem as a jumping-off point and failed to credit them. OpenAI has denied the accusations.

It remains uncertain if OpenAI’s models made use of the work completed by Buckmaster and Alpöge, though Sébastien Bubeck, a member of the technical staff at OpenAI, said in a press briefing that the team was inspired to pursue the problem after hearing a rumor about Buckmaster and Alpöge’s efforts. But whether or not OpenAI’s models took advantage of Buckmaster and Alpöge’s research, this episode may mark a turning point in the history of mathematics.

AI models now seem essential for making progress on the most important mathematical problems of our time, and solving them may demand resources only available at a couple of frontier AI companies, which often defy the norms of academic collaboration that undergird most mathematical progress. If that’s the future we are headed for, it is unclear how human mathematicians will fit into it. 

The problem that OpenAI claims to have solved is known as the Navier–Stokes existence and smoothness problem. It is one of seven Millennium Prize Problems selected by the Clay Mathematics Institute in 2000. Solutions come with a one million dollar prize; before today, only one other Millennium Prize Problem had been solved. 

The Navier–Stokes problem concerns a set of equations that describes how fluids, such as water and air, flow over time. The equations are widely used in the field of fluid dynamics, and they have proven powerful, but physicists and mathematicians didn’t understand them completely. In particular, it was unknown until today whether the equations might, under some conditions, break down and predict an impossible state of affairs—such as a fluid having infinite velocity.

On Monday, NYU’s Buckmaster posted a proof on the social media site Mastodon showing that a simplified version of the Navier–Stokes equations can indeed break down—a major step forward on the Millennium Problem. He and Alpöge had worked on the problem for almost a year, using publicly available models from both OpenAI and Anthropic.

Then today, OpenAI presented a proof showing that the full Navier–Stokes equations can break down as well. The proof was obtained using an internal model that dramatically outperforms the already-impressive Astra model, which was only released last week. The company says it does not plan to claim the million-dollar prize for solving the problem.

These mathematical achievements are indisputably impressive, but they have attracted far less attention than the controversy about their origins. Along with the proof, Buckmaster posted a document detailing his interactions with OpenAI employees after he heard rumors about their work and reached out to one of them. According to him, OpenAI employees presented two possibilities to him: Either he and Alpöge could post their work and OpenAI would post their Navier-Stokes solution the following day, or he could work with OpenAI on a Navier-Stokes paper that excluded Alpöge from authorship, due to his affiliation with Anthropic, OpenAI’s biggest rival.

Buckmaster also wrote that he asked the employees whether the agents had obtained access to transcripts of the work that he and Alpöge had done with OpenAI models, which they denied; and whether OpenAI models had been trained on those transcripts, to which they offered no response. MIT Technology Review reached out to Buckmaster for comment, but didn’t hear back before publication.

The clear implication of the document is that OpenAI’s models somehow made use of Buckmaster and Alpöge’s work. That scenario is plausible on its face. The Buckmaster/Alpöge and OpenAI proofs both make use of an approach to the Navier-Stokes problem pioneered by the mathematicians Diego Córdoba and Luis Martínez-Zoroa.

According to Javier Gómez-Serrano, a mathematics professor at Brown University, this approach was one of several that was thought to hold promise for solving the Navier-Stokes problem. So, while it’s by no means impossible that both teams could have arrived at this approach independently, it’s also conceivable that Buckmaster and Alpöge’s work could have influenced OpenAI’s.

In the press briefing, Mark Chen, OpenAI’s chief research officer, again denied that any agents or OpenAI employees accessed Buckmaster and Alpöge’s transcripts—but given what has been revealed about the Hugging Face hack, it’s clear that OpenAI is not always entirely aware of what its agents are doing. 

If OpenAI’s models did train on Buckmaster and Alpöge’s work, or if its agents somehow gained access to it, then the company’s failure to track down the truth and assign those researchers appropriate credit reflects poorly on it. But there might be a thin silver lining to that version of the story for mathematicians, because it would suggest that the hard work of two humans, one of whom is a prominent expert on Navier-Stokes, was essential to the agents’ ability to solve the Millennium Problem.

Experts have long identified “research taste,” or the ability to choose promising research questions and directions, as a major obstacle for AI in science and mathematics. If the OpenAI agents did indeed choose to follow the Córdoba–Martínez-Zoroa approach because Buckmaster and Alpöge had done the same, then human research taste played an essential role in OpenAI’s success.

Even so, the bigger picture here is sobering. The progress that Buckmaster and Alpöge made over almost a year of collaboration with publicly available models speaks to the promise of human–AI collaboration. But they were not able to achieve a full solution. Meanwhile, OpenAI brute-forced a solution in a few days using an internal model, and their successful solution came at an astronomical cost: In the press briefing, Bubeck and Chen said the team was only able to solve the problem by running about 10,000 agents concurrently, at a cost of millions of dollars.

Over the past few months, I’ve heard from several researchers that mathematicians are becoming depressed, and it’s not difficult to see why. Mathematics is quickly becoming the province of frontier AI companies with impressive internal-only models, money to burn, and a lack of collaborative spirit. “Whether AI companies will decide to spend their money on doing one thing or another, I truly don’t know,” says Gómez-Serrano. “What is clear is that very few mathematicians will have resources of that scale.”

If OpenAI and Anthropic keep striving for more and more impressive mathematical accolades, there might not be any open problems left for human mathematicians outside of those companies to wrestle with. That would dramatically change the field of mathematics.

Last week, UCLA mathematician Terence Tao wrote a Mastodon thread describing how important mistakes, wrong directions, and incomplete solutions are for the field. “In most cases in pure mathematics, the problems are posed not because we desperately want the solution to these problems in and of themselves, but because we have seen from past experience that human-directed efforts to solve these problems tend to spur further development of the field,” Tao wrote.

“Prematurely solving the problem by purely AI-powered methods—particularly without full transparency into the solution process—can contaminate this process to the point where it actually becomes a net negative for the progress of mathematics as a whole.”

Humans might take longer than agents to solve mathematical problems, but in the process, they uncover new mathematical approaches and ideas that might inspire their peers and even birth their own subfields.

But when AI agents solve those problems instead—and when private companies keep the agents’ wrong turns from public view—those benefits disappear. It remains to be seen what else will vanish in the process. 

The panelists explain how platform teams adapt to support AI-assisted engineering, highlighting which capabilities belong in the platform. They discuss trade-offs between standardization and developer autonomy, while sharing strategies to manage AI tooling, security guardrails, and shifting workflows.

GitLab warns that isolating an AI coding agent in a sandbox does not necessarily make the agent safe. In a new security analysis, the company describes an internal evaluation in which an AI agent escaped its sandbox by exploiting a vulnerable package proxy that had been explicitly placed on the sandbox's allowlist.

Danijar Hafner’s office in San Francisco’s SoMa district sits mostly empty. His brand-new startup is still in stealth mode and doesn’t even have its name on the door. On the day I visit, there’s only one other person there, and little in the way of furniture. But what it lacks in decor, it makes up for in robots. Humanoids of various shapes and sizes hang like marionettes from racks that run down the center of the wide-open space.

While Hafner, 31, won’t say too much about his new venture just yet, he describes it as a continuation of his longtime work to enable AI to navigate environments it has not encountered in training. The humanoids, which he imports from China, are the next evolution of this work—and its physical embodiment. Their ability to react in previously untested scenarios will be key to getting robots into human spaces. Because if you want to send a robot into a person’s home, for example, it needs to be able to handle a floor plan and furniture it’s never seen before. 

To achieve this, Hafner relies on something called model-based reinforcement learning. He develops world models—AI models designed to emulate physical reality—and trains agents within them. The agent essentially treats the model as a real-world simulation and learns how to act there. It then uses those experiences to make predictions (to dream or imagine, Hafner might say) about future outcomes. That allows agents—or the robots they’re embedded in—to navigate unfamiliar situations IRL.

“I get to interact with a lot of really smart people in research at Google, and he easily sits in the top half of 1%.”

Timothy Lillicrap, Google DeepMind

Unlike other efforts, Hafner’s technique enables agents and the robots they control to execute massively complicated tasks without the real-world trial-and-­error training that’s traditionally been used in robotics. 

Hafner grew up in a rural town in northeastern Germany, where his parents were both classical musicians. He learned programming from a neighbor, and in high school he began taking online courses about AI, which quickly developed into a passion. “I was always fascinated with how thinking works,” he says. AI offered him a way to emulate it on a computer.

In 2015, as a second-year under­graduate studying engineering at Hasso Plattner Institute in Potsdam, he won a role as a student researcher at Google Brain. From there, he went on to a dozen internships and other positions at the company, including stints with Google Brain and Google DeepMind (the two have since merged under DeepMind) in the UK, Canada, and the US. He worked with industry legends including Geoffrey Hinton, who is often referred to as one of the godfathers of AI, and Ashish Vaswani, coauthor of the groundbreaking research paper “Attention Is All You Need,” which described the transformer technology used by today’s large language models.

One of Hafner’s former managers and coauthors at Google, Timothy Lillicrap, describes him as a standout among standouts. “I get to interact with a lot of really smart people in research at Google, and he easily sits in the top half of 1%,” Lillicrap says. “In many cases he would build, single-­handedly, things it would take entire teams of engineers to build.”

Over the years, Hafner has honed and proved his approach by pitting agents trained within his world models against popular video games. His first breakthrough was PlaNet, a model that allowed agents to execute actions by planning ahead. His Dreamer 2 was the first agent to hit human-level performance playing Atari 2600 games using a world model. Dreamer 3 was the first one to solve the Minecraft Diamond challenge—successfully mining in-game gems on its own. And Dreamer 4 went a step beyond that by learning to mine diamonds from an offline data set of recorded game-play videos, without ever interacting with the game directly. 

More recently, he’s begun to migrate his agents out of the virtual world and into physical reality. His DayDreamer project used the Dreamer algorithm to let robots operate themselves in novel environments and react to new experiences (such as being pushed over) without any specific training. 

Today, Hafner is working on his new startup, which he left Google DeepMind to form in the fall of 2025. Though he’s coy about his next steps, it’s clear he’s dreaming big: “I was interested in solving a problem,” he hints, “that would change the world.” 

Coca-Cola is using AI to recommend which products Malaysian retailers should order and in what quantities through its Coke Buddy platform.

The Perfect Basket feature uses Coca-Cola’s Central Recommendation Engine to analyse previous orders, ordering frequency, seasonality, weather, and purchasing patterns among similar businesses.

Coca-Cola said Coke Buddy currently supports about 39,000 retail outlets across Malaysia. The company describes Coke Buddy as a self-ordering platform that allows retailers to buy products through its app, website, or WhatsApp, with personalised order suggestions and order tracking also available.

Perfect Basket builds on those existing ordering functions by recommending both products and quantities before a retailer completes an order. Retailers can review the recommendations and retain control over what they purchase.

How Perfect Basket guides retailer orders

Coke Buddy already uses previous purchase history to suggest products a retailer is likely to order again. Perfect Basket adds other signals, including seasonality, weather, ordering frequency, and purchasing trends among comparable businesses.

Perfect Basket recommends products and quantities before retailers submit their orders through Coke Buddy. Fulfilment is handled separately by Coca-Cola Refreshments Malaysia or its suppliers under existing sales and distribution arrangements.

Coca-Cola said its sales teams remain involved with retailers alongside the digital ordering system, while retailers retain control over the final purchasing decision.

Coca-Cola recently disclosed usage figures for Perfect Basket following its Perfect Basket, Perfect Ride campaign, which ran from January to April 2026. The campaign encouraged retailers to use the recommendation feature when placing orders and received more than 4,500 entries from over 4,000 retailers in Malaysia.

During the campaign, 83% of participating outlets adopted Perfect Basket recommendations, according to Coca-Cola. The figure applies only to retailers taking part in the campaign, not the full network of about 39,000 outlets supported by Coke Buddy.

Coca-Cola also said participating outlets that followed the recommendations recorded higher sales revenue growth than comparable retail outlets. The company did not disclose the size of the difference or provide detailed performance data showing how individual recommendations affected sales or inventory levels.

The available Malaysian campaign data does not provide figures for forecast accuracy, stock availability, inventory levels, or logistics costs.

Suggested orders extend beyond Malaysia

Coca-Cola has deployed similar suggested-order capabilities elsewhere in its bottling network. In its first-quarter 2024 results, the company said it and its bottling partners had connected nearly eight million customers to B2B platforms globally, while AI-enabled suggested-order capabilities had reached more than three million outlets in Latin America.

Coca-Cola has said these systems combine customer data, external information, and AI to generate predictive order recommendations. Then-chief executive James Quincey said in 2024 that digital ordering also allows retailers to adjust deliveries without waiting for a salesperson to visit.

Coca-Cola has also linked suggested orders to changes in its sales process. Quincey said AI-generated orders allow pre-sales staff to spend less time taking routine orders and more time on account development, while retailers continue to make the final purchasing decision.

Coca-Cola has reported results from earlier pilots using similar recommendation systems. In its second-quarter 2024 earnings call, the company said retailers receiving AI-generated product recommendations based on previous orders and market data were more than 30% more likely to purchase the recommended SKUs in initial pilots. These results did not relate specifically to Perfect Basket in Malaysia.

In a separate demand-prediction project, Coca-Cola combined historical sales data with weather and geolocation information to generate replenishment recommendations. CIO Neeraj Tolmare told Fortune in 2025 that a three-country pilot recorded sales 7% to 8% higher than outlets that were not using the AI algorithm.

Perfect Basket remains available after the campaign. Coca-Cola said it plans to continue developing Coke Buddy and the recommendation feature using retailer feedback and data, while retailers will continue to have access to the company’s sales representatives alongside the digital ordering system.

(Photo by Mahbod Akhzami)

See also: MG Ship adds AI route optimisation as logistics returns accelerate

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.

The post Coca-Cola uses AI to improve retailer ordering in Malaysia appeared first on AI News.

Google’s newest AI weather forecasting model predicts wind speed at 100 metres above the ground, roughly the height of a modern wind turbine. It also forecasts cloud cover and how much sunlight reaches the surface, and it updates every hour. Energy traders, grid operators and wind and solar developers already pay other companies for that data. The introduction of WeatherNext 3 now puts Google in their market.

Google DeepMind and Google Research released the model on September 3. It produces a global forecast every hour at up to five-kilometre resolution for surface variables such as temperature and moisture. The previous version, WeatherNext 2, worked on a 25-kilometre grid and refreshed every six hours. Google says the new energy variables are meant to help grid operators and developers predict how much power their wind and solar assets will generate, then match that against demand.

The consumer side of the launch has had most of the attention. WeatherNext 3 now powers weather results in Google Search, the Gemini app, Google Maps and the Google Maps Platform Weather API. Behind it sits an enterprise layer that matters more commercially. The same forecast data can be queried in BigQuery and Earth Engine or downloaded in bulk from Google Cloud Storage, with no model setup required by the customer.

Why the energy sector is buying AI weather forecasting

Grid operators are running a system that has become harder to predict at both ends. On the generation side, renewables now account for most new capacity. S&P Global Market Intelligence’s US Grid Outlook 2026 projects solar and energy storage as the primary sources of new capacity this year, at 51.2GW and 25.7GW respectively out of more than 90GW of planned additions. 

Solar and wind generate according to the weather rather than demand, so each gigawatt added makes a short-term forecast more accurate.

On the consumption side, the new load is coming from AI. S&P Global identifies the spread of data centres across North America as a primary driver of the recent surge in electricity demand, forcing utilities to revise their load forecasts upward. Deloitte’s 2026 Power and Utilities Industry Outlook projects peak demand growing by roughly 26% by 2035, with data centre demand alone potentially reaching 176GW, five times its 2024 level.

The cost of getting a forecast wrong is straightforward. If an operator underestimates how much wind power will arrive, it has to buy replacement electricity at short notice, usually from gas plants kept on expensive standby. If it overestimates, wind and solar farms end up being paid to switch off because the grid cannot absorb what they are producing. Both outcomes are expensive, and both are forecasting failures.

The market Google is entering

Selling weather forecasts to the energy sector is an established business. Vaisala, Solcast, DNV’s WindGEMINI and IBM’s HyperWatch all compete in it. So does Jua, a Swiss firm that claims its EPT-2 model beats Microsoft Aurora and DeepMind’s earlier GraphCast on accuracy while updating 24 times a day, against what it describes as a typical four updates a day among competitors.

Google’s advantage is reach. The same forecast appears as a table in BigQuery, a layer in Earth Engine, an API in Google Maps Platform and the default answer in Google Search. No specialist vendor has that spread, and the hourly refresh closes the update-frequency gap those vendors have used to differentiate themselves.

The incumbents have one technical argument left. Jua’s published position is that physics-based models such as ECMWF’s HRES still outperform purely data-driven AI models during record-breaking extreme weather, because physics models encode rules about how energy and mass move through the atmosphere, while AI models learn patterns from past data. 

Jua sells a physics-constrained product, so the claim serves its own interests. It also describes the conditions grid operators worry about most, when a storm falls outside anything the model has seen in training.

What is new, and what is being oversold

The architectural claim behind WeatherNext 3 is that it learns from real observations instead of from simulations. Most AI weather models, WeatherNext 2 included, are trained on output from numerical weather prediction models, which are supercomputer-driven physics simulations that carry a six-hour data lag. That lag can introduce bias in fast-changing variables such as rain and surface temperature. WeatherNext 3 ingests live geostationary satellite imagery and trains directly on readings from individual weather stations.

The shift is real, though narrower than much of the coverage has suggested. Google’s own system diagram shows the model taking in one-hour satellite mosaics alongside traditional historical analysis. DeepMind senior research scientist Ilan Price told Bloomberg the gain comes from not waiting for the next analysis and using the most recent information available instead. 

Reporting puts the remaining data lag at three to four hours, down from about seven. Dependence on numerical weather prediction has been reduced, not removed.

The accuracy figures need similar care. Google reports improvements of up to 60% against NASA’s IMERG satellite product, 30% against MRMS radar and 10% against rain gauge readings at early lead times, measured using a standard scoring method for probability forecasts. Those are three separate baselines, and the percentages do not add together. The widely repeated claim of 50% better precipitation forecasting applies specifically to forecasts a day or more ahead. Every figure carries an “up to” qualifier, which makes each one a best case rather than a typical result.

Google published no independent third-party validation alongside the launch. It points instead to live evaluations by Brightband, whose leaderboard it cites in claiming WeatherNext 3 is the most accurate global weather model to date. A utility considering a switch away from a paid specialist will care more about performance in its own service territory, on its own assets, than about a global leaderboard position.

Google’s own stake in the problem

Google is selling forecasting tools into a grid problem its own industry helped create. The data centre build-out driving the load growth utilities are struggling to serve is led by the hyperscalers, Google among them, and Google has signed multi-gigawatt renewable procurement agreements to supply its own facilities.

Accurate prediction of wind and solar output is directly useful to a company matching large volumes of clean energy against a load that is both growing and variable. That is commercial logic, and it goes some way to explaining why the energy variables shipped in this release.

Google has not published pricing for enterprise access to WeatherNext 3, or said whether the BigQuery and Earth Engine data carries standard Cloud query charges or a separate licence. Utilities weighing a move away from a paid specialist will want that figure before they weigh any accuracy claim.

2025, more than 65,000 employees in its Corporate and Investment Bank were actively using the platform, while more than 90% of its engineers were using AI coding assistants.

The bank also said AI-based transaction screening allowed it to review more than twice the previous transaction volume while reducing manual operator checks by half.

Bank of America is using a generative AI-enabled system called EricaAssist with more than 18,000 customer service employees. The tool summarises why a customer is calling, retrieves relevant information, and recommends possible next steps while keeping the employee responsible for the interaction.

Bank of America said in July 2026 that EricaAssist can deliver contextual guidance in under three seconds and has reduced average call times by nearly one minute. The bank plans to extend the system to additional servicing scenarios and business lines later in 2026.

(Photo by Google)

See also: MIT AI forecasts extreme weather without historical data

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.

The post AI weather forecasting enters the energy market as Google targets grid operators with WeatherNext 3 appeared first on AI News.

M&T Bank has deployed AI copilots to more than 15,000 employees as the US regional bank applies AI to internal operations, customer service, software development, and risk management.

The bank uses AI to analyse call-centre conversations, draft reports, generate code, identify customer needs, and flag portfolio risks, according to Fast Company. M&T is also examining agentic AI applications in cybersecurity and fraud detection.

American Banker reported in September 2025 that 16,000 of M&T’s roughly 22,000 employees were already using Microsoft Copilot for tasks including drafting emails and reports and summarising call-centre conversations.

Before the wider rollout, M&T initially restricted employee access to public large language models. Chief data officer Andrew Foster told American Banker that the bank blocked the tools because employees could potentially enter sensitive company information into public-facing services.

M&T later evaluated enterprise providers and selected Microsoft Copilot, starting with a pilot involving about 800 employees before expanding access across the organisation.

Foster said using generative AI to summarise call-centre conversations saves about six minutes per call. Software developers at the bank also use GitLab tools to generate code, while employees remain responsible for reviewing AI-generated work.

M&T’s human-review requirement is also reflected in its 2026 Code of Business Conduct and Ethics. The policy requires employees to use approved AI tools and prohibits confidential, proprietary, customer, employee, or regulated information from being entered into unapproved systems. Employees remain responsible for the accuracy and appropriateness of AI-assisted work.

Building the technology and data foundation

M&T’s AI deployment follows a technology overhaul that began in 2018. The bank said more than half of its technology specialists were external workers at the time, compared with an 80% in-house technology workforce today.

M&T now has about 2,000 technologists working across more than 300 agile teams and has hired more than 1,000 technology specialists during the programme.

The bank has also replaced dozens of older platforms. M&T said technology outages have fallen by more than 80% since 2018, while the number of system upgrades completed annually has increased by 300%.

Technology spending exceeded $1.2 billion in 2025, nearly three times its 2017 level. Wisler told Forbes in August 2026 that annual technology releases increased from about 15,000 in 2018 to 65,000 in 2025.

Wisler joined M&T as chief information officer in 2018 before becoming senior executive vice-president for technology and operations in 2025. His current remit covers both technology and operational functions across the bank.

M&T’s data programme developed alongside the broader technology overhaul. Foster, who joined the bank in 2023, began building a data-lineage programme to track where information originates, how it is used, and how it moves between systems.

Foster told American Banker that the data-lineage work was not created in response to generative AI. He described it as a core capability for understanding M&T’s data estate.

The bank also established a Data Academy focused on data governance and data skills, with around 2,000 employees participating in the programme.

M&T has created an internal repository called Edison containing authoritative documents and information on bank policies. The bank also uses data-lineage software from Solidatus and Monte Carlo to trace information as it passes through databases, applications, and business-intelligence systems.

The lineage work gives M&T visibility into the source, meaning, quality, and governance of individual data elements, according to Foster. He said one application for that governed data is the bank’s use of Copilot.

M&T also uses retrieval-augmented generation with internal, governed data, according to American Banker.

Scaling AI into daily banking operations

Wisler told Forbes that M&T is pursuing generative AI through three routes: general employee use, AI capabilities embedded in existing applications, and proprietary systems built around the bank’s own data and processes.

M&T operates more than 1,800 applications, many supplied by third-party vendors. Wisler said one of the bank’s AI pathways is identifying useful AI capabilities already embedded within those applications.

M&T’s third pathway involves proprietary AI development around the bank’s own data and processes. Forbes reported that early applications include repetitive operational work, software development, fraud prevention, and cyber defence.

Fast Company’s September report also said M&T continues to assess both internally developed AI systems and external tools, including general enterprise software and technology designed specifically for banks.

Earlier workforce use cases centred on drafting, summarisation, call-centre work, and software development. Fast Company reported that newer applications include identifying customer needs and flagging portfolio risks.

Other large US banks have also expanded generative AI across employee workflows.

JPMorganChase launched its internal LLM Suite platform to more than 200,000 employees in 2024. By 2025, more than 65,000 employees in its Corporate and Investment Bank were actively using the platform, while more than 90% of its engineers were using AI coding assistants.

The bank also said AI-based transaction screening allowed it to review more than twice the previous transaction volume while reducing manual operator checks by half.

Bank of America is using a generative AI-enabled system called EricaAssist with more than 18,000 customer service employees. The tool summarises why a customer is calling, retrieves relevant information, and recommends possible next steps while keeping the employee responsible for the interaction.

Bank of America said in July 2026 that EricaAssist can deliver contextual guidance in under three seconds and has reduced average call times by nearly one minute. The bank plans to extend the system to additional servicing scenarios and business lines later in 2026.

(Photo by Alain Pierre-Lys)

See also: Bank of England reviews AI rules for agentic AI in finance

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.

The post M&T Bank expands enterprise AI after years of technology overhaul appeared first on AI News.

OneRail has launched an AI-powered delivery platform that uses Nvidia technology to help retailers, wholesalers, and distributors decide how individual orders should be delivered.

Called OmniSTAR, the system evaluates options including owned fleets, couriers, parcel carriers, and other delivery modes, then selects the lowest-cost option that meets the required service level, according to OneRail.

The platform combines Nvidia’s cuOpt decision optimisation engine and cuDF data processing software with OneRail’s delivery pricing and performance data. Nvidia accelerated computing infrastructure is used to process the routing and delivery-mode calculations.

OneRail said the system can reduce computation times by as much as 10 times. A calculation that previously took 20 minutes can be completed in under two minutes, while a calculation taking a week can be reduced to about two days, according to the company.

OneRail said the shorter processing time allows the optimisation to run within live delivery operations, where multiple fulfilment options can be evaluated before an order is assigned.

“If you don’t have the ability to make lightning-fast decisions, you’re giving up margin,” Catania said in an interview with CNBC. “Last-mile fulfilment is expensive.”

From prediction to delivery decisions

OneRail’s broader AI systems use prediction and optimisation for different parts of the delivery process. The company said its machine-learning models estimate factors including service time, lateness risk, the probability of first-attempt delivery success, and expected price ranges.

OneRail said those predictions feed into optimisation systems that determine how an order should be executed. Separately, the company said OmniSTAR compares different fulfilment modes before selecting an option based on cost and service requirements.

Research on dynamic vehicle routing makes a similar distinction between predicting changing conditions and recalculating operational decisions as new information becomes available. A 2024 review in the European Journal of Operational Research identified travel-time prediction and real-time re-optimisation as separate areas of time-dependent routing research.

Nvidia cuOpt handles route optimisation

Nvidia describes cuOpt as an open-source, GPU-accelerated optimisation library for vehicle routing and other mathematical optimisation problems.

Nvidia’s documentation shows that cuOpt can account for vehicle costs, capacities, travel times, operating windows, starting locations, and other restrictions when calculating routes. Its cost models can also use distance, time, monetary cost, or a weighted combination of those measures.

OmniSTAR applies cuOpt to both routing and delivery-mode selection. OneRail said this allows the system to compare available fulfilment options for an order and identify the lowest-cost option that still meets its service requirements.

OneRail said many retailers still rely on static rules or manual planning when making these decisions, and that OmniSTAR is designed to evaluate more delivery combinations within shorter operational timeframes.

Nvidia said cuOpt does not exhaustively test every possible route. Instead, the solver generates candidate solutions and iteratively improves them using GPU-accelerated heuristics to produce high-quality results within a set computation time.

The platform also uses Nvidia cuDF, a GPU-accelerated library for tabular data processing, including filtering, joining, and aggregating datasets.

OneRail combines those capabilities with its own delivery data and operational models. Its dataset is based on millions of deliveries across a network that the company said includes more than 12 million drivers and over 1,000 logistics partners.

The data covers pricing and delivery performance across different transportation modes. OneRail said OmniSTAR can use the information to identify delivery rules that increase costs and assess how delivery choices affect item-level profitability.

The architecture disclosed for OmniSTAR centres on GPU-accelerated data processing and mathematical optimisation. Nvidia describes cuOpt as the optimisation component used for problems including vehicle routing.

Because cuOpt is stateless, changes in operating conditions require the optimisation problem to be modelled and submitted again. Nvidia cites vehicle breakdowns, driver absences, road blockages, traffic, and new high-priority orders as examples of changes that can prompt this type of dynamic reoptimisation.

OneRail said OmniSTAR can rerun delivery scenarios as variables including fuel costs, weather, and shipping conditions change. The company has separately said its use of cuOpt allows it to evaluate more routing scenarios and recalculate routes faster than its previous approach.

OmniSTAR moves into live operations

OmniSTAR is already deployed with selected enterprise customers.

At US Foods, OneRail said the system identified delivery configurations that were reducing margins, including low-margin products being transported long distances using higher-cost equipment. US Foods subsequently used the findings to adjust pricing and restructure some delivery patterns, according to OneRail.

OneRail also told CNBC that an unnamed large tire distributor using the platform achieved $40 million in run-rate savings over three years. The customer was not identified, and the savings figure was provided by OneRail. The company also told CNBC that it expects OmniSTAR to exceed $6 billion in gross merchandise volume during the fourth quarter of 2026.

CNBC reported that OneRail and Nvidia had worked on the project for three years before its launch. OneRail said the collaboration included direct engagement with Nvidia’s cuOpt engineering team on last-mile delivery and large-scale logistics optimisation, alongside its participation in the Nvidia Inception programme.

In March this year, FedEx launched FedEx SameDay Local in collaboration with OneRail, connecting customers to a national network of more than 1,000 delivery providers.

(Photo by Brecht Corbeel)

See also: A quarter of Nvidia’s business next year comes from labs it is financing

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.

The post OneRail uses Nvidia AI for real-time last-mile delivery optimisation appeared first on AI News.

Amid rapidly growing adoption of enterprise-level AI agents, there’s a disconnect emerging between ambition and execution. 

Although 85% of organizations say they want to be agentic within the next three years, 76% say their current operations and infrastructure can’t support that change. They cite a lack of readiness across people, processes, and workflows. 

The sticky tape problem

The challenge is that many organisations are often layering AI agents onto existing operations, rather than reimagine the operating model and how work will need to be rewired, explains Prasun Shah, global CTO for workforce consulting and chief AI officer at PwC UK Consulting. “They’re embedding AI employees into what is a human operating model,” layering on AI agents to existing workplace structures when “this is like adding sticky tapes to parts of an operating model that is breaking.”

Doing so may be preventing organizations from unlocking the full value agentic AI offers, creating circumstances where disillusionment can quickly creep in. That full value lies in agents’ capacity to execute entire workflows with limited human input. They can coordinate complex tasks, make independent decisions, adjust to changing conditions, and iterate performance. 

In early proving grounds that span customer service, HR, and sales, it’s already estimated that AI agents could accelerate business processes by as much as 30% to 50% and low-value work time by 25% to 40% when deployed at scale. But with this capability comes greater complexity and the need for an enterprise-wide change.

Growing the AI vocabulary 

Enterprise agentic AI platform Ema describes this change as agentic business transformation (ABT), a term it coined last year in partnership with HFS Research, in an attempt to plug what it sees as a gap in the existing lexicon about AI agents, and to provide enterprises with a new framework with which to think about their own adoption of the technology. 

“None of the existing vocabulary captures the full scope of the change,” explains Ema CEO and founder Surojit Chatterjee. “Digital transformation was about moving from paper to software. AI transformation was about adding artificial intelligence to existing processes. Co-pilot is about AI assisting in various human tasks. But ABT is something categorically different: It’s the integration of AI agents into the fabric of the organization.” 

For Shah, the dedicated term (ABT) “helps drive the need to redesign an organization in its entirety: its operating model, its workflows, decision rights, and performance management systems.” He emphasizes that “everything that’s needed to ensure those agents are actually active participants in value creation, rather than just point tools or productivity aids.”

According to Ema, ABT encompasses three core pillars: an organization’s technology stack, its workforce, and the metrics used for success. 

AI agents as connective tissue

The first pillar of ABT is the technology stack. “Your existing tech stack was designed for human-operated, application-centric workflows,” says Chatterjee. “It needs to be reconsidered when the actor is an AI agent operating at machine speed across multiple systems simultaneously.”

 As AI agents are integrated into an organization, enterprises will need to pivot from a set of linear processes and steps, to rewiring work in a very different way, explains Shah. That’s because the value in AI agents isn’t as another layer in an existing technology stack but as a connective tissue, he explains, moving between or across layers to coordinate a high-level task or retrieve and interpret data from multiple discrete applications. AI agents can create “a true competitive differentiation for an enterprise” by making decisions based on this capacity to contextualize, he says. “That is where the next battleground will be.”

To build this connective tissue, leaders need to adapt their technology stack to surface higher quality decisions from AI agents, prioritizing access to multiple datasets and applications simultaneously to develop tacit knowledge. “Organizations that make this architectural shift become genuinely more adaptive,” says Chatterjee. “When a new business requirement emerges, you don’t wait six months for a software vendor to build a feature. You configure an AI employee using natural language and connect it to the systems it needs. The time from business to production workflow drops from months to days.”

The workforce, redesigned

As AI agents are deployed for more use cases, enterprise leaders must consider what this means for dynamics across their workforce, the second pillar of ABT.

Workforce structures today deviate little from the hierarchical model of the early days of industrialization. To maximize efficiency and scale, processes are standardized, tasks are clearly delineated between strategic business units (SBUs), and employees progress up through an organization based on their capacity to optimize output from teams below them. But with AI agents that can execute, coordinate, and optimize tasks—often without managerial coordination—the lines of that established hierarchy become blurred.

In a workforce that blends AI agents and human employees, managers will be freed up from many execution-based tasks but take on new responsibilities associated with managing hybrid teams. Managers “will need to be able to manage issues around trust, explainability, psychological safety, and even status dynamics” to navigate new tensions that could arise in a hybrid workforce, says Shah.

The impact of agentic AI on existing workforce structures goes far beyond the management layer, too. McKinsey predicts that by 2030, three-quarters of current jobs will require redesign, upskilling, or redeployment, and organizations will need to act swiftly to amend recruitment, retention, and remuneration. 

From output to outcome

Success metrics are the third and final pillar of ABT. 

As AI agents assume greater ownership of core enterprise processes, taking on collaborative roles alongside human employees, traditional workforce metrics that focus on activity or output—such as calls handled or reports filed—no longer make sense. 

“When you add AI employees into the workforce, activity metrics become meaningless or actively misleading,” says Chatterjee. “An AI employee can handle a thousand customer interactions in the time it takes a human to handle ten. If you measure success by interactions handled, you’ll conclude the AI is working brilliantly while missing whether any of those interactions actually drove customer satisfaction, retention, or revenue.” To correct this, enterprises must develop a new set of metrics that focus on outcome rather than output. That is, metrics on the broader benefits or changes achieved, rather than individual deliverables. 

For example, when one of Ema’s large enterprise customers overhauled its own metrics, switching from tool metrics like cost per query and AI accuracy, to outcomes like the percentage of contracts reviewed without human escalation, the measured ROI from agentic AI tripled within two quarters. The changes meant “this customer stopped building point solutions in high-volume, low-complexity workflows and started deploying AI employees where the outcome value was highest,” says Chatterjee.

Integrating new metrics may also require a complete reconfiguration of reward and talent management processes, as well as accountability and ownership within organizations, points out Shah. In human-AI teams, for example, although ethical and fiduciary responsibilities will likely remain with human employees, operational accountability will become significantly more diffused to reflect the systemic role of AI agents.

This change will raise new questions that senior leadership teams will need to wrestle with, Shah adds. They’ll need to consider: Who is accountable when an AI employee makes a mistake? What happens when AI and humans disagree? What guardrails should be erected to safeguard customers? 

Laying the groundwork for systems-level change

Systems-level change is gradual. These are complex lines of inquiry that experts continue to grapple with. But in kickstarting internal dialogue about the core pillars of ABT—the workforce, the technology stack, and the metrics by which success can be gauged—leaders can lay the groundwork for an enterprise better poised to embrace AI agents at a systems level and start to close the gap between their ambition and execution. 

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff. It was researched, designed, and written by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.

Autonomous AI systems are beginning to move beyond software environments and into warehouses, delivery networks, and public spaces. The development is drawing attention to whether current AI rules cover systems that operate in physical environments.

Most existing AI governance frameworks have focused on online harms and model outputs, including bias, misinformation, and harmful content. Embodied AI systems carry risks in physical environments, where failures can affect infrastructure, property, or human safety.

Singapore’s Infocomm Media Development Authority published version 1.5 of its Model AI Governance Framework for Agentic AI on May 20. The framework sets out guidance for organisations deploying AI agents that can plan, make decisions, and take actions across multiple steps to complete user-defined goals.

The framework says agents can interact with tools, external systems, and other agents, including systems that update databases, write files, control devices, or perform transactions. It lists access controls, monitoring, and human approval among governance measures for deployment.

AI moves into physical systems

At an AI summit in Singapore last week, discussions around robotics and embodied AI focused on operational safety issues more commonly associated with aviation, industrial systems, and critical infrastructure oversight than conventional software regulation.

Speakers also discussed whether autonomous systems can operate safely and reliably in unpredictable real-world environments over extended periods.

Dr. Ya-Qin Zhang, founding dean of the Institute for AI Industry Research at Tsinghua University, said embodied AI systems amplify risks already associated with autonomous software. He said failures can directly affect transport systems, drones, logistics networks, and critical infrastructure.

“Any risk in the digital domain will be amplified in the physical domain, and the physical domain will have a physical consequence,” Zhang told MLex on the sidelines of the summit.

He added that vehicles, drones, smart grids, and other infrastructure could become exposed as AI systems are embedded more deeply into physical operations.

Speakers discussed reliability, operational monitoring, and post-deployment assurance as governance concerns. Summit discussions pointed to deployment-based governance models built around simulation, telemetry, and iterative testing, rather than one-time certification alone.

IMDA’s framework also recommends gradual rollouts, continuous monitoring, and further testing after deployment. It says agents interact dynamically with their environment and not all risks can be anticipated before release.

Monitoring becomes a deployment issue

Grab, which is piloting autonomous vehicles and delivery robots in Singapore’s Punggol district, said deployment governance depends heavily on simulation, testing, and continuous monitoring.

“We do a lot of simulation, we do a lot of testing in closed courses and open courses in order to make sure our robots are reliable,” Suthen Thomas Paradatheth, Grab’s chief technology officer, said during one of the summit panels.

“Before we scale to hundreds of robots, we make sure we crack it first in simulation and with a few robots,” he added.

Grab also pointed to monitoring systems designed to track robot performance and detect unexpected failures after deployment.

“There’s a long tail of issues that could emerge,” Paradatheth said.

The IMDA framework says organisations should assess agentic AI use cases based on data access, external system access, autonomy, and task complexity. It also points to the scope and reversibility of agent actions, third-party involvement, and overall system complexity.

It also recommends limiting agent access to tools and systems, applying least-privilege permissions, and defining standard operating procedures for agent workflows. Organisations should also set mechanisms to take agents offline when they malfunction.

Accountability spreads across more actors

MLex reported that embodied AI systems can involve several parties across development, manufacturing, and deployment. These include AI developers, robotics manufacturers, semiconductor suppliers, and infrastructure operators.

MLex also noted that responsibility can be harder to assign when systems continue adapting after deployment through software updates, telemetry, and operational data.

IMDA says organisations and humans remain accountable for agent actions, even when agents operate autonomously. The framework calls for clear responsibility across the agentic AI value chain, from model and platform providers to deployers, tooling providers, and end users.

Applied Materials said large-scale robotics deployment is also tied to semiconductor economics and systems integration. Om Nalamasu, the company’s chief technology officer, said robotics systems will depend on better sensors, energy efficiency, advanced packaging, and computing architectures.

Nalamasu said robotics systems would require purpose-built designs adapted to specific industrial ecosystems rather than a single solution for all environments.

Zhao Yuli, chief strategy officer of Chinese robotics startup Galbot, said Beijing is prioritising deployment scale and industrial commercialisation through government-backed testbeds, industrial partnerships, and long-term funding initiatives.

Galbot has deployed humanoid robotics systems in retail, warehouse, and pharmaceutical operations in China. These include autonomous stores that operate around the clock. Zhao said semi-structured industrial environments are likely to become an early commercialisation path because they offer more controllable operating conditions.

Japan is placing more focus on standards-setting, robotics datasets, and safety governance. Professor Yutaka Matsuo of the University of Tokyo’s Graduate School of Engineering pointed to an “AI Association” project aimed at collecting 100,000 hours of robotics data to support robotic foundation models.

Matsuo also referred to Japan’s AI Safety Institute and the Hiroshima AI Process as part of broader efforts to develop governance standards for embodied AI systems with Singapore and other Asian countries.

Singapore sets out agent controls

Singapore’s framework sets out four governance areas for agentic AI. These cover upfront risk assessment, human accountability, technical controls, and end-user responsibility. The framework describes them as an iterative process rather than a one-time assessment.

The framework says human oversight has to be adapted for agentic systems because continuous review of all workflows becomes impractical at scale. It recommends human approval at significant checkpoints, including high-stakes actions, irreversible actions, and outlier behaviour.

IMDA also identifies automation bias and alert fatigue as risks when humans supervise capable agents. It recommends auditing oversight through indicators such as human override rates and response times, and using automated real-time monitoring to flag unexpected behaviour.

The framework says users should be told what actions an agent can take, what data it can access, and what responsibilities remain with the user. It also recommends employee training on human-agent interaction, oversight, and the professional skills needed to assess agent outputs.

Companies test AI in regulated workflows

JPMorgan is implementing AI tools across its global investment banking business, Paul Uren, the bank’s Asia Pacific head of investment banking, told Reuters. The bank said the tools help bankers access more information and synthesise it with internal systems. They are also being used to prepare content and support client engagement.

JPMorgan CEO Jamie Dimon told Bloomberg News that the bank would hire more AI specialists and fewer traditional bankers. Reuters reported that global banks are increasing AI investment, reshaping workforces, and changing job roles.

The bank is also among selected organisations permitted by Anthropic to use its Mythos cybersecurity model under a controlled initiative known as Project Glasswing. According to Anthropic, Mythos can detect old vulnerabilities in browsers, infrastructure, and software.

Reuters reported that Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley also have access to, or are testing, Mythos, citing sources and company executives.

IMDA’s framework includes a case study from OCBC Bank of Singapore on source-of-wealth analysis. The system parses income-related documents and drafts a source-of-wealth memo. It does not make credit, onboarding, or risk decisions autonomously.

In that case, the workflow is limited to task-level autonomy and operates only when triggered by predefined workflows. Human review is required at critical decision points, and final validation remains with designated reviewers.

Robots move into industrial use

In Japan, one-third of companies are already using or considering AI-powered robots, according to a Reuters survey conducted by Nikkei Research from May 1 to May 15. The survey contacted 492 companies, with 220 responding on the condition of anonymity.

About 4% of respondents said they already use AI robots, 5% plan to deploy them, and 25% are considering doing so. The remaining 66% said they had no such plans.

Transportation equipment manufacturers were the most active group in the survey, with 80% already using AI robots or considering deployment. By comparison, 94% of wholesale sector respondents said they had no plans to deploy AI robots.

Among companies using, planning to use, or considering AI robots, 71% selected manufacturing as a use case. Another 19% selected dangerous tasks, while 11% selected customer-facing services.

The Japanese government expects AI robots to help address the country’s chronic labour shortage and support its position in industrial robotics. Japan is home to robotics companies including Fanuc, Yaskawa Electric, and Kawasaki Heavy Industries, but faces competition from China and the United States in AI-enabled robotics.

Retail agents expand beyond search

Walmart has outlined plans to use agentic AI across shopping, employee, supplier, and developer workflows.

In July 2025, the retailer announced plans for four AI-powered “super agents.” They are designed for shoppers, store employees, suppliers and sellers, and software developers. Walmart said these agents would become the main entry point for AI interactions across those groups.

One of the tools, Sparky, is already available in Walmart’s app as a generative AI-powered shopping assistant. Hari Vasudev, Walmart’s US chief technology officer, said its expanded version would be able to reorder items and plan events. It would also use computer vision to suggest recipes based on the contents of a shopper’s fridge.

Walmart is also developing an Associate super agent for store workers and corporate staff. A separate Marty agent is being built for sellers, suppliers, and advertisers. The retailer is also working on a Developer super agent for testing, building, and launching future AI tools.

The company declined to say whether the agents would replace jobs. Dave Glick, senior vice president of enterprise business systems, said the tools would create new jobs, without giving further details.

(Photo by Growtika)

See also: OpenAI opens Singapore AI lab as IMDA updates AI framework

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.

The post Autonomous AI systems test governance in physical environments appeared first on AI News.

Haven’t you heard? White-collar jobs are going away, decimated by AI. Waves of layoffs in the tech sector (most recently at Coinbase and Meta and Cisco) are said to presage what will soon come for all of us knowledge workers. But before you quit your job as a software developer or financial analyst—or tech journalist—and look to join the plumbers’ union, it’s worth considering today’s economic research on whether artificial intelligence has actually begun to devour white-collar work.

Analysis of the data gathered for the US Bureau of Labor Statistics (BLS) shows that the unemployment rate for the jobs potentially most affected by AI is actually lower than that for occupations less exposed to the technology. And, critically in the mind of economists, there are no signs that large numbers of people are shifting from jobs threatened by AI to supposedly safer ones, such as those involving mostly manual labor.

While the current labor statistics don’t preclude a sudden job upheaval in the coming years, they do throw doubt on the inevitability of the doomsday scenarios and the pace at which they’d unfold. Everyone in the AI community, it seems, is predicting that the technology will soon wipe out jobs, and everyone, it also seems, knows some young wannabe workers who can’t find one. Perhaps we haven’t seen any major disruption in the labor market statistics yet, people often say, but just wait. 

But maybe we should pay attention to what the data is showing us. And right now, the numbers paint a picture of a relatively stable labor market in which AI disruptions remain largely speculative.

“It could be disruptive, but the data is telling us right now that disruption is not yet here, and we have time to plan.”

“All of the available evidence to date suggests that AI’s impact on current labor market conditions is likely small right now,” says Erika McEntarfer, a labor economist who headed the BLS until President Trump fired her last fall after a jobs report that displeased the administration. (Not surprisingly, BLS reports of sluggish job growth have continued since her dismissal.)

McEntarfer, who is now a fellow at the Stanford Institute for Economic Policy Research, says the relatively small impact that AI is having so far on today’s labor market “surprises many people, but it shouldn’t. What we know from history is that it takes time for innovations to work their way through changes in industries and changes in occupations. AI is unlikely to transform labor markets until it first transforms businesses.”

McEntarfer points to US Census data showing that only one in five companies are using AI in any business function. “The data are a great reality check on the fear that AI will be enormously disruptive,” she says. “It could be. It likely will be disruptive, but the data is telling us right now that disruption is not yet here, and that we have time to plan.”

Things ain’t great—but the question is why

The US job market, to be sure, sucks for many, especially younger would-be workers. Unemployment rates for recent college graduates stand at around 5.6%, well above the level for all workers. It’s a rate not seen since the pandemic and the years immediately after the 2008 recession. Even more troubling is that hiring rates have been particularly dismal during the post-covid economy, a trend that hits hard at young people trying to enter the workforce. If you’re a recent college graduate and looking for a tech job, no one, it can seem, is hiring.

There are signs that AI is contributing to the pain for the 22-to-25-year-olds seeking jobs in software development and other occupations that are feeling a big impact from AI. But these professions represent just a sliver of the overall labor market. What’s more, it’s uncertain how much blame AI should get for the job woes. Similarly unknown is whether the loss of entry-level jobs in AI-exposed occupations is a harbinger of what’s coming for others or simply an isolated symptom of what economists refer to as a “low-fire, low-hire” labor market caused by a variety of macroeconomic forces.

Insights into these uncertainties will tell us much about our working fates in the transition to an AI economy. There are no shortage of confident assertions and predictions about what is about to happen; while some people forecast the end of work, others say economic history teaches us that technology advances always lead to more and better jobs eventually. 

The honest answer is that no one knows for sure what AI will bring and whether this time will be different. To help figure it out, we need better and far more comprehensive data.

The statistics gleaned from the federal government’s monthly survey of 60,000 households for the BLS provide a broad overview of the changes to the labor market, while academics and even some AI companies have begun trying to gain a more granular view of specific jobs that are being affected. But the existing data-gathering tools don’t adequately explain how AI is affecting the huge and diverse US labor market.

There’s a long list of questions that we don’t have the data to fully answer. How is AI being used in the workplace? Does the increased use of AI mean the technology will replace workers, or will it make them more productive and valuable? Which occupations and skills are most affected? Who is in most peril from the changes? As David Deming, a professor of economics at Harvard University, puts it: “We’re sort of flying blind.”

To gather more insight into some of these questions, Deming and his colleagues have been surveying several thousand people every three months since 2024, asking them basic questions: Do you use generative AI, and how often? Does it save you time at work? Tracking the answers over time gives the economists important clues (it’s used by a little over 40% of workers but adoption varies by sectors) and allows them to estimate productivity gains (they’ve found some, but nothing economy-shaking). It has also helps document how quickly AI has been adopted in the workplace and how it compares with earlier technologies such as the PC and the internet (the pace has been faster but roughly in the same ballpark).

It’s far from a complete picture of how AI is changing work. But it provides some intriguing results; for example, a fair number of workers in manufacturing and other industrial sectors have tried AI. Deming’s results show that while businesses in general might be relatively slow to formally adopt the technology, lots of their employees are using it.

The young are most vulnerable

But by themselves the exposure results are not a true predictor of which jobs will be lost to AI. That depends on the kinds of tasks done by the technology, the extent to which the AI is adopted, various business calculations about the value of workers, and even the costs of deploying AI. But the exposure findings are a valuable starting point. 

In a working paper called “Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence,” researchers at the Stanford Digital Economy Lab looked at 950 jobs, placing the occupations into five categories from least exposed to most. Then they used a vast data set from ADP, the world’s largest payroll provider, to look at employment growth in each of the categories. Their exclusive access to the ADP data set, which is far larger than the one available through the BLS, allows the researchers to better spot impacts by demographic. When they examined what was happening to different age groups, says Erik Brynjolfsson, the director of the lab who led the effort, “it was extremely striking.”

They spotted the drop in head count for 22-to-25-year-olds in the most exposed occupations, such as software development and customer service, beginning in late 2022, when ChatGPT was first publicly released. Other researchers reported evidence that the decline in these jobs began well before ChatGPT and questioned whether the labor market could react so quickly to the introduction of AI technology. 

But while the Stanford researchers acknowledge that other factors in addition to AI probably contributed to the early declines, they say that after controlling for those factors, they saw convincing evidence of a significant effect from AI after 2024 and growing in 2025 to a 16% decline in entry-level jobs in AI-exposed occupations. In contrast, head count grew for older workers in the same occupations, as did the number of jobs in the less exposed occupations.

Digging deeper into the data, the researchers found another important clue, though one that wasn’t totally unexpected. The impact on head counts depended on how AI was being used. It was specifically the jobs where tasks could be automated (that is, AI could do them “with minimal human involvement”) that accounted for the decrease in employment—jobs for people like software developers. In jobs where AI was mainly used but to augment human work, head counts grew faster than the average for entry-level workers.

That’s consistent with one explanation for the woes of many young would-be workers. It could be, according to the Stanford paper, that entry-level jobs depend more on the types of knowledge that people acquire through education but that can readily be mimicked by AI; the authors call this codified knowledge. It might be particularly easy to automate such tasks as entry-level coding. In contrast, older workers have more so-called tacit knowledge, the type based on their experience. That type of wisdom is harder for AI to replace.

Despite the findings about AI’s impact on young workers, Bharat Chandar, an economist at Stanford and one of the authors (along with Brynjolfsson and Ruyu Chen), stresses that it’s still early when it comes to understanding how the technology will affect jobs in the future. It could be that the job loss will spread to older workers and to less AI-exposed occupations, he says. But Chandar says it is also possible that firms and workers will adjust to shifting labor demands, and the effects will level off or even disappear.

A recent paper by economists at the Federal Reserve Board found, not surprisingly, that annual employment growth for coders has slowed significantly—by about 3%—since the introduction of ChatGPT. But here’s a critical detail: Overall employment for coders continues to grow. Employment in coding jobs is still rising, they noted, just more slowly than before 2022. 

In short, coding jobs are not going away, at least not anytime soon. But it’s an occupation that is clearly being transformed by AI.

One of the somewhat surprising wrinkles uncovered by recent research is that wages in sectors highly exposed to AI have risen relatively fast since the introduction of ChatGPT. One explanation is that employers are still willing to pay for the kinds of knowledge and experience that are, at least for now, hard to replace with AI. If true, this suggests not the end of work in AI-exposed jobs but, more specifically, the demise of the typical career model in which young graduates are hired to do software tasks that can be automated and are slowly trained to gain that valuable tacit experience. The earn-while-you-learn model might finally be broken—at least for some occupations.

The simple truth could be that coding skills are no longer a guarantee of a job. That may help to explain the drop-off of computer science majors at schools around the country. Future canaries in the cubicles are sniffing out the dangers of looking for a job when their skills can be matched by AI.

But a closer look at the data shows that students are not necessarily turning away from AI-related careers. Rather, they appear to be tailoring their skills to the changes they see underway as AI becomes increasingly important for various disciplines. Interest is rising in AI-adjacent fields like data science and cybersecurity. One fast-growing major: artificial intelligence itself (a recent addition to many college offerings).

Is this time different?

Anxiety over the potential of AI to replace workers is nothing new. I wrote “How Technology Is Destroying Jobs” in 2013, describing how a slew of new digital technologies, including AI, were beginning to threaten white-collar work. I wasn’t alone. It was a popular theme at a time when the labor market was sluggish and jobs were scarce. 

In one of his last days in office in late 2016, President Obama issued a report written by his top economic and science advisors warning that AI was threatening workers. Among the findings was that automated vehicles—especially driverless trucks—could eliminate 2.2 million to 3.1 million existing US jobs.  Around the same time, one of the pioneers of AI, Geoffrey Hinton, said that “people should stop training radiologists” because it was “completely obvious” the occupation was soon to be replaced by AI.

Perhaps this time is different, and we can put aside the lessons of economic history. Certainly, AI has gained unimaginable powers to do humanlike tasks. Perhaps it will devour jobs in ways that we’ve never seen before. And perhaps that will happen abruptly, without a warning buried in the labor statistics. But the previous bouts of AI job anxiety still hold a prescient lesson: Our real focus needs to be less on the dystopian fears and more on the very real transitions in the workplace that will likely affect millions of people.

“Even if there is not mass or even increased unemployment, the transition could still be very difficult,” says Jed Kolko, senior fellow at the Peterson Institute for International Economics and former undersecretary of commerce in the Biden administration. “And what does a difficult transition period mean? It means people losing jobs, or people’s jobs being redefined in ways that make those jobs pay worse or be less meaningful. And some people whose jobs are threatened may not be able to adapt.”

The more we understand this transition, the better prepared we’ll be to deal with it.  And for that we’ll need better and more complete data.

For McEntarfer, the former commissioner of the BLS, the real question is the speed of any disruption. “If it happens at the normal pace of technological change, labor markets will have time to adapt. If there is a sudden and severe disruption, then that will be a big challenge for policymakers,” she says. “That’s really the most important question facing us right now: how rapid this transformation is going to be.” And, she adds, “we’ll know by watching the data.”

Two decades ago, the country was caught flat-footed by the so-called China shock as free-trade policies led to an influx of imports and the devastation of manufacturing jobs in many parts of the country. It took years for researchers to understand the data showing how the trade policies, generally welcomed by economists, were destroying communities. Today the threat of an economic transformation brought on by AI is far larger and points to potentially far more damage for huge groups of workers.

To head off another devastating labor transition, we will need well-timed government and business policies, especially programs to train and reskill workers. If McEntarfer and other labor economists are correct, we probably have time to design deliberate and effective strategies to manage the transition. But first we need to better understand what is going on—and how fast.

It’s hard to find an economist who is more enthusiastic about AI’s future than Stanford’s Brynjolfsson, who believes that we’re likely on the brink of a huge boost that will transform the economy. “Perhaps the best productivity growth of my lifetime is coming up,” he says.

But Brynjolfsson also warns that a lack of data is severely limiting our visibility into the economic and societal impacts that are coming. At a time when hundreds of billions are being spent on rolling out the technology, he says, “we’re not investing even 1% of that on understanding the transition.”

Artificial intelligence has not so far produced a clean story of mass unemployment. Aggregate employment in developed countries remains broadly stable, and recent assessments have found limited evidence that AI has shifted the headline numbers. But a troubling change may be hiding beneath the surface: the quiet weakening of the first rung of the career ladder.

The most worrisome evidence is showing up exactly where we should expect it first: in early-career hiring. A working paper released in November 2025 by the Stanford Digital Economy Lab found that workers aged 22 to 25 in the most AI-exposed occupations experienced a 16% relative decline in employment after the spread of generative AI, even after controlling for other factors that might affect firms’ employment decisions. An Anthropic report from March 2026 provides suggestive evidence that led to a similar conclusion.

More experienced workers in those same occupations did not suffer the same decline. Employment is not also declining in the entry-level jobs with low AI exposure. The concern is specific to early-career jobs that are exposed to AI.

That is not a minor signal. It suggests that firms may be using AI to substitute for the junior tasks through which people traditionally gain their first foothold—at least for those in jobs where generative AI is used extensively, like software developers, customer service representatives, computer programmers, and information systems managers.

The time is now to make changes in the way we train, prepare, and support young people who are about to enter the workforce. Educational institutions need to reorient for the era of an AI-augmented workforce. Governments must incentivize businesses to hire and train early-career workers. Businesses, in turn, need to recognize the importance of developing a long-term workforce experienced in AI—a process that begins with entry-level workers. And students themselves should take on the responsibility of not only becoming AI fluent but learning how to apply that knowledge in various fields.

In short, we must change the way we have traditionally thought of entry-level work.

This is especially true because the broader labor market for recent graduates is also softening. The Federal Reserve Bank of New York reported that in the fourth quarter of 2025, the unemployment rate for recent college graduates rose to 5.6%, while the underemployment rate (the share of graduates working in jobs that typically do not require a college degree) reached 42.5%, its highest level since the covid pandemic. No single statistic can prove that AI is the sole cause of that deterioration. Hiring in general is way down post-pandemic, and young people are particularly vulnerable to the slowdown. But it would be a mistake to ignore the possibility that AI is accelerating an already difficult transition from school to work.

Behind these statistics is a great deal of personal distress. Recent graduates today often submit hundreds of applications before they receive a single offer, and surveys consistently find elevated rates of anxiety, financial precarity, and burnout among young workers in extended job searches. If AI quietly closes the door on typical early jobs, people will pay the price in delayed independence, postponed family formation, and the sense that their first serious professional efforts have been refused.

It also matters because entry-level jobs are part of the economy’s training system. Junior analysts learn which numbers can be trusted. Young software developers learn how production systems fail. New marketers learn how customers behave outside the neat language of dashboards. Early-career legal and financial staff learn how rules, judgment, deadlines, and human relationships actually interact. If AI absorbs more of the drafting, triage, coding, summarizing, and administrative preparation that once helped train entry-level workers, firms may become more efficient in the short run while society becomes less capable in the longer run.

The right way to improve the skills of young workers is not to tell them, “Learn to code.” That advice, which shaped more than a decade of federal initiatives and university expansion, rested on the premise that coding was a stable, scalable skill almost anyone could learn and parlay into a middle-class job. The premise no longer holds. The layer of work AI handles well—translating a specification into routine code, reproducing standard patterns, debugging predictable errors—is precisely the layer that “learn to code” programs were built around.

Supervising AI systems in their work is now a much more relevant skill. So understanding the outputs AI systems produce will become very important.

To help people develop such skills, we should require universities, community colleges, and professional programs to embed AI literacy, data literacy, prompt-based workflow skills, verification skills, and domain judgment into ordinary degrees. Every graduate should know how to use AI tools, check their output, understand their limits, and combine them with human expertise. This matters even for graduates entering occupations that look relatively safe from AI, such as those in health care. Almost every job contains tasks—drafting, summarizing, scheduling, research, basic data work, routine communication—for which AI is already a substantial productivity tool.

The competition most young workers will experience is not human versus machine but colleague versus AI-augmented colleague. For most young workers, the realistic path to making themselves valuable is not to avoid AI but to become fluent in the technology and combine that with domain judgment, contextual reasoning, and human relationship skills. To this end, schools should emphasize paid co-ops, apprenticeships, and employer-linked projects so students build judgment in real workplaces before they graduate.

Governments should also create targeted tax credits, wage subsidies, and training grants for employers that hire early-career workers into structured, AI-augmented roles. The architecture for this kind of conditional, behavior-linked subsidy already exists in US tax policy. What is missing is a version of these instruments built specifically around early-career AI-augmented work.

Firms, for their part, should stop making hiring decisions based only on short-run cost savings from AI. Young workers are not valuable only for the tasks they perform this quarter. Their value lies in learning, skill formation, institutional memory, and future productivity. Entry-level hiring is not just an expense. It is an investment in the future stock of judgment inside the firm. The most effective AI-augmented senior workforce of the late 2030s will be drawn overwhelmingly from the junior cohort of today. Firms that automate away the learning stage may improve their immediate margins but find themselves, a decade from now, without anyone who understands how their own AI-driven workflows actually behave.

Students graduating this spring and next face a tough labor market in transition. AI fluency is becoming a commodity. Domain expertise without AI fluency is being outpaced. The combination is what is genuinely scarce. The mechanical engineer with knowledge of manufacturing and AI proficiency; the software programmer with knowledge of financial services who is also a whiz at AI—these are the types of people who will be in demand.

Georgios Petropoulos is an assistant professor at the USC Marshall School of Business. His research focuses on the implications of information technologies for innovation, competition policy, and labor markets.

Google's SynthID, designed to embed imperceptible signals into AI-generated content, is adding a new Content Detection API on Google Cloud's Gemini Enterprise Agent Platform, after gaining adoption by several industry players including Nvidia and OpenAI.