The persistent evolution of global finance is not always obvious, but every few decades, a new technology, event or agreement fundamentally resets the monetary system’s core machinery. We sit at one of those inflection points and few realize how consequential it could become. In 1944, world leaders gathered in Bretton Woods, New Hampshire, to architect a new international financial order. Out of the ashes of war, the dollar became the central “settlement asset,” global exchange rates found their anchor, and the rules of global commerce became standardized and interoperable. That system, for all its flaws, brought a clarity and stability that propelled half a century of cross-border growth. Today, as multiple countries accelerate pilot programs for central bank digital currencies, or CBDCs, and experiment with sovereign blockchains, we’re witnessing the birth of a new international monetary framework.
Nvidia’s research shows small language models with 1.5B parameters can handle 70-80% routine enterprise AI tasks; with large models as fallbacks for complex cases
Nvidia’s latest research makes the case that small language models (SLMs) could prove more practical and more profitable in the enterprise. The argument is straightforward: SLMs are powerful enough for many real-world tasks, cost less to run and can be deployed at scale without the same infrastructure burden as large language models (LLMs). The research offers both a technical framework and a business case. Nvidia introduces a conversion algorithm that rethinks how enterprises deploy artificial intelligence. Instead of sending every request to a heavyweight LLM, the system routes repetitive tasks such as document parsing, summarization, data extraction and draft generation to SLMs. To illustrate, Nvidia introduced its Hymba line of SLMs with a hybrid design that balances precision with efficiency. The Hymba-1.5B model, with just 1.5 billion parameters, has been shown to perform competitively on instruction-following benchmarks at lower infrastructure cost than larger frontier models. For business leaders, the key takeaway is not the architecture but the economics; smaller models are now capable enough to handle professional tasks without the infrastructure burden that has limited LLM adoption. If SLMs can complete 70% to 80% of routine steps cheaply and reliably, and LLMs backstop the rest, the ROI profile for enterprises improves. The hybrid model is not about eliminating error but about routing work to reduce exposure and optimize cost. If Nvidia’s thesis holds, enterprises could evolve toward architectures where SLMs handle most routine work and LLMs act as fallbacks. That shift would redefine how organizations design AI systems and how they measure value.
Uber Eats integrates Avride autonomous robots for Shake Shack deliveries in Jersey City, enhancing last-mile automation for thousands of global retailers with average delivery times under 30 minutes
The fast-casual restaurant chain is collaborating with the Uber Eats delivery subsidiary of Uber Technologies to deploy autonomous delivery robots from a store location in Jersey City, N.J. The program is an extension of a sidewalk delivery robot initiative the two companies initially launched in Los Angeles in 2024 and then expanded to Atlanta in June 2025. In its latest automated delivery pilot, Shake Shack and Uber Eats are rolling out robots from Avride. When customers order from Shake Shack’s participating Jersey City location, or other participating restaurants on the Uber Eats app, they may be eligible for robot delivery. A notice that reads “autonomous vehicles may deliver your order” will appear on the restaurant’s Uber Eats storefront when the service is available. Uber Eats leverages Uber’s technology and logistics capabilities to partner with 825,000 retailers in more than 11,000 cities globally, with an average delivery time the company says is under 30 minutes.
Agentic workflows shift enterprise automation from rules to goal-driven orchestration, with UiPath’s Maestro platform orchestrating adaptive and scalable agentic workflows blending robots, AI agents, and human oversight
Agentic workflows are moving automation into a new era, shifting from rigid rule-based tasks to adaptive, goal-driven processes that respond to real-world complexity. The result is a demand for orchestration platforms capable of uniting humans, robots and agents into one cohesive system that can evolve alongside business needs, according to Graham Sheldon, chief product officer of UiPath Inc. The adoption of agentic workflows is not about replacing existing systems but orchestrating them in new ways. UiPath’s Maestro platform is designed as a control plane that allows enterprises to blend robots, agents and human oversight without reinventing their entire technology stack. This approach enables businesses to deploy automation quickly while adapting to unique operational needs, Sheldon explained. A critical advantage of this orchestration model lies in its ability to surface insights and optimize processes over time. By modeling and visualizing workflows, organizations gain visibility into bottlenecks and can make iterative improvements while maintaining reliability and compliance.
Databricks acquires Mooncake to eliminate ETL pipelines, enabling instant real-time synchronization between PostgreSQL and lakehouse data, accelerating agentic AI workloads by up to 100x
Databricks is acquiring Mooncake, an early-stage startup focused on bridging PostgreSQL with lakehouse formats, to eliminate the need for ETL pipelines entirely. Financial terms of the deal are not being publicly disclosed. The technology promises to make operational data instantly available for analytics and AI workloads, with performance improvements ranging from 10x to 100x faster for common data movement operations. Nikita Shamgunov, who joined Databricks as VP of Engineering after leading Neon, told Databricks co-founder and chief architect Reynold Xin that Databricks should buy Mooncake on his literal first day at the company. Mooncake has several technologies in its portfolio. There is the ‘pgmooncake’ extension that enables analytical workloads to run on PostgreSQL. Then there is the moonlink component that Shamgunov describes as an acceleration tier. It enables real-time transformation between row-oriented PostgreSQL data and columnar analytical formats without traditional ETL pipelines. “Moonlink allows you to basically create a mirror of your OLTP data in a columnar representation in Iceberg and Delta,” Shamgunov explained. “Moonlink also supports an acceleration tier as well. So in many places, you accumulate latencies when you query the data lake by metadata lookups, or s3 both on the way in and on the way out.”
Truist is the “first super-regional” to introduce a Visa Business Infinite product — expanding access to high-value financial tools for SME and offering 2% unlimited cash back on all purchases; 10% loyalty cash bonus when cash rewards are redeemed into an eligible Truist Business Deposit account
Truist Financial Corporation announced the launch of its Truist Business Premium Visa Infinite card, a premium credit card designed for small businesses with meaningful annual spend. The new offering positions the bank as the first super-regional to introduce a Visa Business Infinite product — expanding access to high-value financial tools built for simplicity, scale and growth. Headquartered in the Southeast — one of the nation’s fastest-growing economic regions — and with a strong presence in the country’s top growth markets, Truist is uniquely positioned to support entrepreneurs where business momentum is accelerating. The launch underscores the bank’s purpose-driven commitment to innovation and relationship banking, aligning with its strategy to modernize payments, advance digital tools and deepen connections with business owners in key markets. Available exclusively through Truist branches and business banking teams, the card delivers 2% unlimited cash back with an additional 10% loyalty cash bonus, an offer to earn a $1,000 welcome bonus, and unsecured credit lines from $15,000 to $100,000 for approved clients. It integrates seamlessly with Truist’s approach to small business, combining practical advice and full-scale capabilities with digital-first features, designed to help entrepreneurs optimize cash flow and make smarter financial decisions. “This card is designed for business owners who want premium benefits without the hassle,” said Chris Ward, Truist Head of Enterprise Payments. “It’s built around simplicity, speed and safety — delivering straightforward rewards, strong protections and seamless integration with our clients’ existing Truist relationships. We’re focused on helping small businesses grow with confidence and control — and this card is a powerful step in that direction.” With nearly 80% of small businesses using credit cards to manage operational expenses — and nearly half relying on cash-back rewards to reinvest in their business — the Truist Business Premium Visa Infinite card offers a strategic solution for entrepreneurs seeking to maximize value and streamline financial management. “Visa, a world leader in digital payments, is pleased to collaborate with Truist to deliver a new solution that meets the evolving needs of small businesses,” said Denise Press, Head of Small Business, NA, Visa Commercial Solutions. “This new card reflects our shared focus on enabling smarter, more efficient ways for businesses to manage spend and reinvest in growth, ultimately helping businesses to thrive.” Key benefits of the Truist Business Premium Visa Infinite card: 2% unlimited cash back on all purchases; 10% loyalty cash bonus when cash rewards are redeemed into an eligible Truist Business Deposit account; $1,000 welcome bonus in cash-back rewards after spending $15,000 in eligible purchases within the first 90 days of account opening; No foreign transaction fees; Visa Infinite benefits including 24/7 concierge service, trip delay coverage, auto rental protection and price protection, and more; Unsecured credit lines from $15,000 to $100,000; $299 annual fee, waived the first year and in future years with more than $100,000 in qualifying spend.
UWM cuts costs for refinance borrowers offering title fees as low as $1,295 along with a $500 borrower incentive
United Wholesale Mortgage (UWM) is cutting costs for refinance borrowers with updates to its title review and closing alternative, TRAC+. The lender is offering title fees as low as $1,295 along with a $500 borrower incentive. TRAC+ allows UWM to handle all title work on refinances directly for its broker partners — eliminating the need to coordinate with a title company or settlement agent for balancing fees, scheduling closings and related tasks. The first version of TRAC, launched in October 2022, had a $1,600 flat fee plus a settlement agent fee of up to $500. TRAC+, announced in May 2024, offered a $1,850 flat fee, with no settlement agent fee. Wednesday’s update reduces the title fee to as low as $1,295, depending on the state and loan amount. The $500 incentive can be applied as a pricing credit, listed on the Closing Disclosure (CD), or issued as a check to the borrower. Brokers also gain access to additional perks: UWM is offering 9,500 points — which can be redeemable for a $300 gift card to help borrowers cover post-closing expenses — as well as virtual closings with notaries available 24/6. Closing packages can be issued in as little as 15 minutes from clear-to-close, with UWM balancing the CD. In November 2024, UWM also introduced TRAC Lite, which features a $0 title fee and reduced settlement agent fees ranging from $350 to $600. The lender was among those selected to participate in Fannie Mae’s title insurance waiver program. Both TRAC+ and TRAC Lite are now available in 37 states. UWM recently also raised the one-unit conforming loan limit from $806,500 to $819,000 for conventional and Department of Veterans Affairs (VA) loans, ahead of the Federal Housing Finance Agency (FHFA) announcement in November. It also extended a 90 basis-point incentive for rate-and-term refinances until Oct. 31.
Stripe’s new Open Issuance enables any business to launch and manage their own stablecoin with just a few lines of code
Stripe announced new products to help businesses take advantage of AI and stablecoins to grow their revenue. Announcements included Open Issuance, which empowers businesses to launch and manage their own stablecoins with just a few lines of code, and new solutions for agentic commerce, helping businesses and their customers transact via AI tools and agents. Open Issuance, a new platform powered by Bridge, enables any business to launch and manage their own stablecoin with just a few lines of code. Businesses can mint and burn coins freely, and customize their reserves to manage the ratio between cash and treasuries and choose their preferred partners. Treasuries are managed by BlackRock, Fidelity Investments, and Superstate. Cash is held by Lead Bank to provide liquidity as necessary. To establish liquidity, all new coins are fully interoperable with any others issued via Open Issuance, and Bridge’s orchestration API helps with low cost conversions to virtually any other stablecoin. Crucially, businesses can generate rewards for originating stablecoins on their platform, and use earnings from these rewards to incentivize their customers. CASH, an open-loop stablecoin designed by Phantom, is the first stablecoin issued via Open Issuance. Recently announced stablecoins like mUSD for Metamask, and USDH, built by Native
Markets for Hyperliquid, will also be issued on top of the platform. Alongside Open Issuance, Stripe announced a series of product updates that make it easy for businesses to deploy stablecoins for business growth: Businesses can now accept recurring stablecoin payments, and Stripe’s Optimized Checkout Suite now accepts stablecoin payments by default. US businesses with Financial Accounts can now hold stablecoin balances, convert between fiat currencies, spend stablecoin balances with a locally issued card, and send stablecoins to crypto wallets cross-
border. Stripe and OpenAI released the Agentic Commerce Protocol (ACP)—a new standard that already powers the new
Instant Checkout in ChatGPT. ACP establishes a shared language between merchants and AI agents. With a single integration, merchants can sell through AI agents while retaining control over their brand, catalog, fulfillment, and customer relationships. ACP is an open standard: it works across AI agents, and businesses can adopt it even if they don’t process payments with Stripe. Stripe also announced that it’s also working with early partners including Microsoft Copilot, Anthropic, Perplexity, Vercel, Lovable, Replit, and Manus to test its solutions in real-world settings to help businesses get ready for agentic commerce. Beyond ACP, Stripe announced new tools to help AI companies monetize their products. Stripe Billing now makes it simple to run hybrid revenue models that combine subscriptions with usage-based pricing, while a new API will let companies connect to LLM providers and track inference cost changes in real time. Stripe Radar is expanding to block ‘friendly fraud’ types, such as the abuse of free trial periods. AI companies see this problem every day: bad actors string together multiple free trials and rack up huge compute bills without ever paying for a service. Stripe Radar can now stop up to 62% of trial fraud at the source, protecting
margins without adding friction for legitimate users. Other announcements included:
- A new app-to-web payments flow helps businesses seamlessly redirect customers from their iOS app to a mobile checkout page powered by Stripe Checkout. Businesses like Superwall, Voodoo, Replit, and Hinge are saving up to 90% on payment costs with this new flow.
- Link, Stripe’s wallet for faster checkouts, now supports Buy-Now-Pay-Later (BNPL) with Klarna.
- Businesses can now use Stripe Managed Payments, a merchant-of-record solution, for one-time payments. Stripe manages global taxes, fraud prevention, dispute management, fulfillment, and more on businesses’ behalf.
- For businesses selling physical goods, Stripe Tax will support tax collection in 102 countries—up from 40 last year.
- Stripe Reader S710, Stripe’s newest hardware product, is now globally available, with support for cellular connectivity so users can process payments even without WiFi.
- AI platforms like Vercel and Replit can now embed Stripe sandboxes directly inside their environments, allowing developers to launch and manage payments and financial services without leaving the platform.
TD Bank’s restructuring allocates $500 million to AI-based AML remediation with independent Guidepost monitoring, targeting affluent customers and automating processes to save $750 million while staying below $434 billion asset ceiling
TD Bank laid out its vision Monday for regaining its footing, and pulling ahead, in the United States after historic anti-money laundering blunders. It’s a plan that includes not just major cost-cutting initiatives, but also efforts to target affluent customers and increase artificial intelligence capabilities. The Toronto-based company has spent more than a year reviewing its strategy and making major renovations to its operations, regulatory compliance and financial structures. Executives announced a sweeping agenda of ambitious growth goals during the company’s investor day on Monday. In the U.S., TD is aiming to cut costs by $750 million over the next four years by pruning its branch network, automating more processes and reducing its governance spending. Innovation initiatives, like data management and AI development, are expected to deliver about $200 million of those cost savings. The bank will relocate or close 10% of its roughly 1,100 locations in the U.S., which it expects to take out costs of $100-$150 million. The move comes on top of the 174 branches TD has closed and the 62 it has opened or moved since 2020. Simultaneously, TD has plans to accelerate revenue growth in U.S. co-branded credit cards and services to affluent Americans. The company expects those two initiatives to lift revenues by $700 million and $300 million, respectively. All told, TD is projecting a 13% increase in return on equity for its American business by 2029. Leo Salom, the head of TD’s $386 billion-asset American operation, said that although the bank’s deposit franchise is the “flagship” of its American operations, TD lacks “relationship depth” with many clients. The company’s goal is to drive cross-selling opportunities through the bank’s technology transformation, location optimization plan and product concentration changes, he said. The bank’s chipper outlook comes almost exactly one year after it was hit by U.S. officials with an asset cap, more than $3 billion in fines and a sweeping list of fixes it must make for money-laundering-related blunders. To stay below its $434 billion cap, the bank has reduced its asset size about 10% by selling or winding down certain loan portfolios and businesses, such as point-of-sale financing, correspondent lending and supply chain financing. TD’s U.S. balance-sheet restructuring effort resulted in after-tax losses of about $1.4 million in the quarter that ended on July 31, and total losses from the strategy should eventually land around $1.5 billion, the company said during its latest earnings call. But Salom said Monday that an investment portfolio remix has contributed about $500 million in additional net interest income this year. In 2026, as TD moves past the bulk of the major investments and expenses required by the revamp, the U.S. business is targeting $2.9 billion of net income after taxes, Salom said. “We’re applying rigor and discipline as we make investments to grow the business and achieve our goals,” Chun said about TD’s enterprise-wide goals. “As we drive change, harness AI and extend digital leadership, we’re building a more disciplined, simpler and faster bank to fund our growth and achieve our efficiency ratio target.” Still, the company is also targeting growth as it navigates through a shifting economic and political environment, Chun said. “We face new risks, economic changes, disruptive technologies and new innovations,” he said. “Here in Canada, our trading relationship with the United States is center stage.” “We know we have lots of work to do,” he said.
Katy Knox and Lindsay Hans of Bank of America, Meghan Shue of Wilmington Trust-M&T, Thasunda Brown Duckett of TIAA are among American Banker Most Powerful Women in Finance
Taken together, the effects of the digital economy—notably, the ability to reduce costs, maximize speed and open up services to new customers—is propelling a greater democratization of financial services. The leaders ranked as The Most Powerful Women in Finance are at the forefront of this technological transformation, adapting to change, driving results and exploring new opportunities that sit at the intersection of business and innovation.
Mary Callahan Erdoes, JPMorganChase CEO, Asset and Wealth Management: The CEO of JPMorganChase’s mammoth asset and wealth management (AWM) business moved into the company’s gleaming new headquarters at 270 Park Avenue a day earlier and underestimated the return time to her office after escorting a client to the entrance. Over the past year she’s shepherded the bank’s high-net-worth clients through a volatile spring, flown around the world to meet with heads of state and billionaires, spoken at big forums, including the World Economic Forum in Davos, Switzerland, and even helped fetch a record price for a venerable sports franchise. She also played a high-profile role as the most senior member of Chairman and CEO Jamie Dimon’s operating committee.
Thasunda Brown Duckett, TIAA President and CEO: When TIAA launched an IRA product broadening access to its proprietary annuity products on March 10, the stock market recorded one of its most volatile days in years. Investors were fretting over the impact of tariffs and the possibility of a recession. The day’s roller coaster ride reinforced what TIAA president and CEO Thasunda Brown Duckett already believed about the value of lifetime income annuities.
Katy Knox, Bank of America President, Bank of America Private Bank: Katy Knox is one of the longest-tenured female leaders in the U.S. private banking sector. Knox, president of Bank of America’s Private Bank since 2018 and a member of BoA’s executive management team, views her visibility as both a responsibility and an opportunity.
Lindsay Hans, Bank of America President, Co-Head of Merrill Wealth Management: In just her second year as president and co-head of Bank of America’s Merrill Wealth Management business, Lindsay Hans has met those challenges head on. In partnership with Eric Schrimpf, president and co-head of the 25,000-employee Merrill Wealth Management unit, Hans boosted year-over-year revenue by 9% in 2024 and increased the AUM balance of the Investment Advisory Program by 18% year-over-year, among other accomplishments. The unit’s success has continued into 2025. During the first half of this year, Merrill Wealth Management posted nearly $10 billion in revenue, and wealth management advisors added 12,700 clients and more than 20,000 households. As of March 31, Hans and Schrimpf oversaw client balances of $3.5 trillion.