Nvidia CEO Jensen Huang pledged to invest £2 billion ($2.6 billion) to supercharge the country’s AI startup ecosystem. Wayve, the U.K.-based self-driving tech startup, has signed a letter of intent with Nvidia to evaluate a $500 million strategic investment in the U.K. startup’s next funding round. Wayve has gained attention and investors for its automated driving system that uses a self-learning versus rules-based approach to its self-driving software. Wayve’s end-to-end neural network doesn’t require high-definition maps and only uses data to teach the vehicle how to drive. That data-driven learning approach is used for “eyes on” assisted driving and an “eyes off” fully automated driving system. The company plans to sell its “Embodied AI” to automakers and other tech companies. Wayve’s self-learning approach, which is similar to the strategy that Tesla uses, is seen as particularly appealing to automakers because it’s not reliant on specific sensors or maps. This means Wayve’s system can work with existing sensors like cameras and radar. The automated driving software captures data from those sensors, which directly informs the driving decisions of the system. Wayve’s generation 2 self-driving platform, which is integrated into its Ford Mach E test vehicles, uses Nvidia GPUs. This week, the startup unveiled gen 3, a platform that uses the in-vehicle compute autonomous vehicle development kit called Nvidia Drive AGX Thor. The gen 3 will allow Wayve to offer eyes-off advanced driving-assistance systems and Level 4 driverless features that will work on city streets and highways.
Citi’s new retail head of retail banking says simplified strategy distinguishes Citi from larger banks; credits fintechs with pushing traditional banks to reduce complexity
Kate Luft, Citi’s new head of U.S. retail banking, draws inspiration for customer engagement from the airline industry. “I think of it like an airline,” she said. “The more you do with us, the more we recognize you.” Luft led the overhaul of Citi’s U.S. retail strategy last year, simplifying products, consolidating checking accounts, and introducing “relationship tiers” that reward customers with perks like waived fees when they meet balance thresholds. “Really what we did was redefine our products and value [propositions],” she said. “Our mandate was, how do we make it super-simple for our clients?”
By integrating Discover’s direct savings bank with Capital One’s “Digital First” model, the combined entity can now rival the “Big Four” banks, unlocking a critical advantage- Discover’s exemption from the Durbin Amendment
Capital One’s $35.3B acquisition of Discover Financial creates a top-2 U.S. credit card player with 19% loan share and vertical integration in issuing/processing. By combining Discover’s proprietary payment networks (PULSE and Diners Club) with Capital One’s data-driven underwriting, the merged entity now holds 19% of U.S. credit card loans and 22% of the customer base. That’s second only to JPMorgan Chase. But the real magic lies in vertical integration: the combined firm now owns both the issuing and processing sides of the credit card business, a rare feat in an industry dominated by intermediaries like Visa and Mastercard. This integration unlocks a critical advantage: Discover’s exemption from the Durbin Amendment. That means the merged entity can bypass interchange fee caps on debit transactions, a $1.2 trillion market. For investors, this isn’t just a line item—it’s a new revenue stream that can be reinvested into customer rewards or used to undercut competitors on pricing. The result? A playbook that challenges the status quo. The merger redefines the fintech landscape. By integrating Discover’s direct savings bank with Capital One’s “Digital First” model, the combined entity can now rival the “Big Four” banks. Discover’s 70 million global merchant acceptance points and Capital One’s cloud-based tech stack create a platform for embedded finance—think payments in SaaS platforms or fintech partnerships. American Express and JPMorgan Chase remain formidable, but the merged entity’s agility in targeting mid-tier spenders and underbanked segments is a game-changer. For example, the Discover Cashback Debit card—a rare offering in the industry—now serves 100 million customers, enabling Capital One to monetize this segment in ways competitors can’t replicate. For long-term investors, this deal is a high-stakes poker game. The short-term pain of integration costs is offset by long-term gains in market share, innovation, and financial inclusion. The key is patience.
Only 33% of developers trust AI accuracy in 2025, down from 43% in 2024 while 66% cite “AI solutions that are almost right, but not quite” that demand careful analysis as their top frustration
New data from Stack Overflow’s 2025 Developer Survey exposes a critical blind spot: the mounting technical debt created by AI tools that generate “almost right” solutions, potentially undermining the productivity gains they promise to deliver. AI usage continues climbing—84% of developers now use or plan to use AI tools, up from 76% in 2024. Yet trust in these tools has cratered. Only 33% of developers trust AI accuracy in 2025, down from 43% in 2024 and 42% in 2023. AI favorability dropped from 77% in 2023 to 72% in 2024 to just 60% this year. Developers cite “AI solutions that are almost right, but not quite” as their top frustration—66% report this problem. Meanwhile, 45% say debugging AI-generated code takes more time than expected. AI tools promise productivity gains but may actually create new categories of technical debt. AI tools don’t just produce obviously broken code. They generate plausible solutions that require significant developer intervention to become production-ready. This creates a particularly insidious productivity problem. Most developers say AI tools do not address complexity, only 29% believed AI tools could handle complex problems this year, down from 35% last year. Unlike obviously broken code that developers quickly identify and discard, “almost right” solutions demand careful analysis. Developers must understand what’s wrong and how to fix it. Many report it would be faster to write the code from scratch than to debug and correct AI-generated solutions. The workflow disruption extends beyond individual coding tasks. The survey found 54% of developers use six or more tools to complete their jobs. This adds context-switching overhead to an already complex development process.
BoA’s CashPro Forecasting is clearest examples of AI-driven efficiency; it learns from a client’s historical cash flows, automatically selecting the most accurate balance for each account and using it to forecast future cash positions
Bank of America’s CashPro platform is being transformed by artificial intelligence, real-time data, and advanced digital tools, supporting over 40,000 corporate and commercial clients worldwide. The platform now includes biometric mobile logins and machine learning–driven forecasting. “Clients … are looking for more productivity and efficiency out of tools that help them manage cash, payments and receivables,” said Tom Durkin, global product head of CashPro. He emphasized the platform’s ability to deliver “predictive, more personalized recommendations to clients to help drive the right treasury decisions,” highlighting tools like CashPro Forecasting, CashPro Chat, and QR Sign-In. Internally, AI is embedded across the bank’s technology stack. “We’re leveraging AI … to help [develop] software,” said Andrew McKibben, head of International Technology. “It improves the productivity of a software engineer — helps them write code, helps them write test cases [and] improve time to market.” Bank of America has issued over 7,800 patents, including 1,400 in AI and machine learning. “We showcase it and we celebrate it,” McKibben said. In generative AI, the bank uses tools for content classification, summarization, and generation. “You can prompt and ask questions of all of our research reports in a library and generate content that might be useful [for someone internally, or in discussions with] a client,” McKibben noted. CashPro Forecasting learns from historical cash flows to predict future balances, retraining its models daily. “There’s nothing more important to the treasurer than preserving and understanding where their cash is,” said Durkin. Forecasts that once took a week are now generated in minutes, with capabilities extending to scenario-based modeling across global subsidiaries. “If I create certain models and events, how will the model scenario work? How will that work within this unit — if I have a subsidiary operating in the EU versus a subsidiary that’s operating out of Brazil?” he explained. CashPro’s self-service tools, such as account verification letters, are seeing rapid adoption — up 21% in Q1 2025. “Clients no longer have to call the service center,” Durkin noted. CashPro Chat, powered by the same AI as consumer assistant Erica, now handles 40% of client queries. The CashPro App has evolved from a transactional tool to a personalized experience. “The app itself really started as … a transactional tool,” Durkin said, but it now supports over $1 trillion in annual payment approvals. Daylon Bailey of Highgate Hotels called it “our saving grace,” noting its intuitive design and strategic utility: “Primary administrators like myself, a lot of times, we’re generally going into CashPro App to make decisions, so it’s nice to have front and center that information that we need right then and there — whether I need to approve a user change, I need to approve a wire or look at positive pay. It’s very intuitive. It’s like having the web-based platform in the palm of your hand.” Security enhancements include QR sign-in and push notifications, with QR sign-in adoption up 60% year over year. McKibben said, “We’re deep into evaluating it,” referring to agentic AI.
AWS new neurosymbolic AI feature to mix symbolic or structured thinking with the neural network nature of generative AI to validate truth or correctness in an AI system against a set of policy or ground truth data, lending greater confidence in deploying AI agents
AWS is banking on the fact that by bringing its Automated Reasoning Checks feature on Bedrock to general availability, it will give more enterprises and regulated industries the confidence to use and deploy more AI applications and agents. It is also hoping that introducing methods like automated reasoning, which utilizes math-based validation to determine ground truth, will ease enterprises into the world of neurosymbolic AI, a step the company believes will be the next major advancement — and its biggest differentiation. Byron Cook, vice president at AWS’s Automated Reasoning Group, told the preview rollout proved systems like this work in an enterprise setting, and it helps organizations understand the value of AI that can mix symbolic or structured thinking with the neural network nature of generative AI. Automated Reasoning Checks validates truth or correctness in an AI system by proving that a model did not hallucinate a solution or response. This means it could offer regulators and regulated enterprises worried that the non-deterministic nature of generative AI could return incorrect responses more confidence. Cook brought up the idea that Automated Reasoning Checks help prove many of the concepts of neurosymbolic AI. Automated reasoning works by applying mathematical proofs to models in response to a query. It employs a method called the satisfiability modulo theories, where symbols have predefined meanings, and it solves problems that involve both logic (if, then, and, or) and mathematics. Automated reasoning takes that method and applies it to responses by a model and checks it against a set of policy or ground truth data without the need to test the answer multiple times. Cook said that agentic use cases could benefit from automated reasoning checks, and granting more access to the feature through Bedrock can demonstrate its usefulness.
AI coding tools on auto-run mode are posing security risks in compromise and data leakage by allowing agents to run command files on a user’s machine without explicit permission, due to vulnerabilities in model repositories and with malicious models granting access to cloud environments
One problem identified by the cybersecurity community at this year’s Black Hat is that shortcuts using AI coding tools are being developed without thinking through the security consequences. Researchers from Nvidia Corp. presented findings that an auto-run mode on the AI-powered code editor Cursor allowed agents to run command files on a user’s machine without explicit permission. When Nvidia presented this potential vulnerability to Anysphere Inc.’s Cursor in May, the vibe coding company responded by offering users an ability to disable the auto-run feature, according to Becca Lynch, offensive security researcher at Nvidia. Part of this issue can be found in the sheer number of application programming interface endpoints that are being generated to run AI. Security researchers from Wiz Inc. presented recent findings of a Nvidia Container Toolkit vulnerability that posed a major threat to managed AI cloud services. Wiz found that the vulnerability allowed attackers to potentially access or manipulate customer data and proprietary models within 37% of cloud environments. Despite the popularity of LLMs, security controls for them have not kept pace. This threat of exploitation has cast a spotlight on popular repositories where models are stored and downloaded. At last year’s Black Hat gathering, researchers presented evidence they had breached three of the largest AI model repositories. If model integrity fails to be protected, this will likely have repercussions for the future of AI agents as well. Agentic AI is booming, yet the lack of security controls around the autonomous software is also beginning to generate concern. Cybersecurity company Coalfire Inc. released a report which documented its success in hacking agentic AI applications. Using adversarial prompts and working with partner standards such as those from the National Institute of Standards and Technology or NIST, the company was able to demonstrate new risks in compromise and data leakage. “There was a success rate of 100%,” Apostol Vassilev, research team supervisor at NIST, said.
Banks accelerate AI deployments as agentic tools gain traction | CIO Dive
Banks ramped up AI adoption as agentic tools began to gain traction in the sector during the first half of the year, according to an Evident Insights’ AI report. The number of new use cases launched by 50 of the world’s largest financial firms doubled compared to the last half of 2024 while the number of technologists working on agentic AI grew more than tenfold, the analysis found. More than half of the 173 use cases deployed by the banks analyzed leveraged generative AI capabilities, Evident said. Nine of the 50 firms documented AI agents in the pilot or production phase, but BNY, Capital One and JPMorgan Chase were the only firms to disclose details of the supporting architecture for agentic workflows. Banks are grappling with twin objectives, according to the report. “They are optimizing potential outcomes from generative AI deployments, while simultaneously assessing early experiments with agentic AI,” Evident said. Banks expect to reap substantial rewards from sustained investments in generative AI capabilities. The technology’s ability to harness vast stores of untapped enterprise data, digest raw documentation and deliver insights that expedite analytics and customer service processes are already transforming daily operations, as developers test the mettle of natural language assistants to ease cumbersome coding tasks. While most of the focus thus far has been directed toward internal use cases, a shift toward the customer is underway. Wells Fargo, for example, beefed up its Fargo virtual banking assistant with Google’s Gemini 2.0 Flash and several other smaller internal models, Evident said. The bank is aiming to up its agentic game through an expanded partnership with Google announced earlier this month. Commerzbank leaned on Microsoft’s Azure AI toolkit to roll out AI avatar Ava in April. The app-based assistant answers general banking questions and is designed to provide customers with personalized account information. Ten of the leading banks in the Evident AI Index, including JPMorgan Chase, Citigroup and Bank of America, have collectively placed AI tools in the hands of over 800,000 employees, representing two-thirds of their workforce. Banks have had the biggest AI wins to date in front office applications and IT and security functions, according to Evident. More than two-thirds of use cases with reported outcomes, including tools that assist sales, were concentrated in these areas.
Moody Ratings says stablecoin growth could cause a 1% decrease in both bank assets and bank lending — that is, a $325 billion reduction, as issuers favor Treasuries for reserves, raising systemic financial concerns
Banks have been concerned about stablecoin issuers coming for their deposits, but the growing popularity of the digital asset could have wider implications, including a reduction in available credit. While stablecoins are still early in their evolution, they are bound to scale massively, Rajeev Bamra, associate managing director, head of strategy, digital economy at Moody’s Ratings, told American Banker. This scale could impact traditional lending, investment products and marketwide risk as the use of Treasuries as stablecoin reserves impacts other sectors of finance. “Stablecoins’ role in the plumbing of financial markets … is making them more systemically important,” Bamra said. Stablecoins have been growing at a fast clip, with circulation doubling from January 2024 to July 2025, accounting for $30 billion of transactions daily, or less than 1% of global money flows, according to McKinsey and Company. That growth is not expected to slow anytime soon. Today, the stablecoin market is just over $250 billion, with Tether and Circle’s stablecoins making up the lion’s share at $165 billion and $67 billion, respectively. Stefan Jacewitz, assistant vice president and economist at the Federal Reserve Bank of Kansas City, believes that the stablecoin industry will eventually grow large enough to boost demand for Treasury bonds, but that growth comes at a cost as the role of Treasuries declines elsewhere in banking. Presently, the role of Treasuries in the stablecoin market is limited. Stablecoin issuers such as Circle and Tether favor U.S. Treasuries as a backing for their stablecoins in circulation because they are low risk and highly liquid. Both issuers hold about half of their assets in U.S. treasury notes: As of June 30, Circle held just less than half of its total $61 billion in assets, $27 billion, in Treasuries, and Tether held $105 billion in Treasuries to back its USDT stablecoin, according to the two company’s respective transparency reports. “If all issuers held a similar proportion of their assets as Treasuries, they would hold around $125 billion in Treasury bills — less than 2% of the $6 trillion in outstanding Treasury bills,” Jacewitz wrote in a research bulletin. “While this sum is not negligible, the stablecoin industry is not as yet considered a major part of the Treasury-bill market, and issuer behavior likely has a limited effect on overall Treasury liquidity.” The stablecoin industry would need to grow to about $900 billion to reach the size of the next smallest category of U.S. Treasury owners, which are private pension funds that hold a little more than $450 billion in Treasuries. By comparison, the largest private holders of T-bills are mutual funds, at $4.5 trillion, according to Jacewitz. But as stablecoin issuers grow their share of coins in circulation, so too will the demand for T-bills, Jacewitz said. JPMorganChase has estimated that the market will grow to $500 billion by 2028, and Standard Chartered estimated the stablecoin market would grow to $3 trillion by 2028. Analysts at Bernstein are also bullish, and predicted the market could grow to $4 trillion by 2035. “This potential flow of funds from bank deposits into stablecoins could increase Treasury demand but also could reduce the supply of loans in the economy,” Jacewitz said. “Assuming the stablecoin market grows from $250 billion to $900 billion … the $650 billion in growth could represent a shift from bank deposits to stablecoins,” Jacewitz said. “This shift would represent a 1% decrease in both bank assets and bank lending — that is, a $325 billion reduction in bank loans to the economy.”
AI agents evolve how automated customer service works- deplolying automated work assignment through ServiceNow Task Intelligence and integrated GenAI capabilities is allowing engineers to resolve problems 36% to 38% faster without forwarding calls
Scott Steele, CEO of Thrive, discusses the use of AI in contact center operations, focusing on end-user support. The company uses ServiceNow as its primary workflow engine to maintain data across various platforms, ensuring accuracy and alignment across the business. Steele emphasizes the importance of policy, governance, and management in AI strategy to ensure successful implementation. Thrive’s main uses for AI in the contact center include process management, which involves understanding bottlenecks and improving customer experience. They have deployed automated work assignment through ServiceNow Task Intelligence and integrated GenAI capabilities, allowing engineers to resolve problems 36% to 38% faster without forwarding calls. This has reduced the time spent on the phone with customers and improved the route of calls. However, some agents still want to remain call center agents and are being moved to other call centers. As the industry continues to evolve, it is expected that autonomous AI agents will take on more decision-making responsibility. Steele predicts that in five years, AI will become the exoskeleton for individuals, making them bigger, faster, and stronger. As AI becomes more digital-oriented, chat will improve, reducing the need for voice-side assistance. However, Steele acknowledges that getting away from humans 100% may be difficult. Automation has already saved hundreds of thousands of hours and improved efficiency and cost structure, making agentic AI an opportunity for businesses to drive efficiency and cost structure.