Banks bet on AI agents as fintechs raise the stakes.
Banks deploy AI agents at scale.
Financial institutions are moving beyond basic chatbots to deploy more powerful "agentic" AI – systems that don’t just answer queries but can execute complex tasks and make decisions. In recent days, multiple major banks have unveiled platforms to embed AI agents more deeply into their operations ([1]).
Citigroup, for instance, announced a new in-house AI platform called Arc that will let its developers create and scale AI agents across the entire bank ([2]). This builds on a strong base of employee adoption – more than 80% of Citi’s 180,000 staff with access to AI tools already use them regularly after extensive prompt-training programs ([3]). In the UK, Lloyds Banking Group similarly launched “Envoy,” a Google Cloud-powered internal platform enabling teams to train, use, and share AI agents securely, with templates and an internal marketplace for reusing agents across the organization ([4]) ([5]). These platforms include built-in compliance and safety checks and keep humans in the loop for oversight, reflecting banks’ focus on reaping efficiency gains without sacrificing control ([6]).
Banks are also extending AI agents to customer-facing functions. Spain’s CaixaBank is rolling out a generative AI virtual assistant as the first point of contact for customers across its digital channels ([7]). Within weeks, the AI agent will be available in all of CaixaBank’s web and mobile app chats, handling ~6,000 customer conversations per month and supporting queries and applications for around 40 different products ([8]). The AI can answer questions, provide tailored information, and guide customers through processes up to the final step, at which point a human adviser takes over to complete the sale ([9]). By automating routine inquiries and product guidance at scale, such agent technology promises faster service and free capacity for staff to focus on higher-value interactions.
Fintechs and big investors pile into AI.
The race to innovate with AI in finance isn’t limited to incumbent banks. Venture capital and big strategic investors are pouring money into AI ventures poised to disrupt financial services. On May 4, AI startup Anthropic announced a partnership with investment giants – including Goldman Sachs, Blackstone and private equity firm Hellman & Friedman – to launch a $1.5 billion joint venture aimed at accelerating AI adoption in businesses ([1]). The new entity will deploy Anthropic’s advanced AI model, Claude, directly into mid-sized companies (starting with those in the investors’ portfolios) by embedding engineers to redesign those firms’ workflows around AI agents ([2]). By addressing the shortage of talent to implement AI, this alliance seeks to fast-track the digital transformation of portfolio companies and beyond.
Meanwhile, fintech innovators are attracting sizable funding to develop finance-specific AI. This week, New York-based AI platform Rogo announced a $160 million Series D funding round led by Kleiner Perkins (raising its total funding to over $300 million) ([3]). Rogo’s flagship AI agent, called Felix, can autonomously perform complex investment banking tasks – from deal sourcing and due diligence to drafting pitch decks and engaging potential buyers – all with no human intervention ([4]). The platform is already used by over 35,000 professionals at 250 financial institutions, including leading global banks and advisory firms, to speed up M&A processes ([5]).
For industry incumbents, these big bets highlight both opportunity and urgency. Major banks are increasingly collaborating with and investing in fintech AI providers (J.P. Morgan’s venture arm joined Rogo’s latest round ([6])) even as they build in-house capabilities – a dual strategy to stay at the cutting edge. The huge capital flows into AI for financial services suggest that the gap between early adopters and cautious followers could widen quickly. Senior executives should consider where to partner or invest to avoid being left behind in this fast-moving wave of AI-driven innovation.
AI turbocharges lending and risk management.
Artificial intelligence is also enabling significant leaps in efficiency and adaptability in core banking activities like credit decisioning and fraud prevention. In the UK, SME lender Allica Bank has begun live trials of an end-to-end AI lending system that can read unstructured loan applications (even simple emails) and issue a credit decision in about 12 minutes – with no human in the loop ([1]). That represents a striking improvement over traditional small business lending, where the same process can take days or weeks of manual review ([2]). The AI-driven loan approvals have so far been applied to about 10% of incoming applications via two broker partners, with roughly half of those cases fully automated from submission to decision ([3]). Allica says the technology could eventually help crack the challenge of efficiently serving “complex” mid-sized business borrowers that larger banks often avoid, although the bank cautions there is still "a huge way to go" in refining the AI for more nuanced lending scenarios ([4]).
In Australia, Commonwealth Bank (CBA) is using AI to bolster fraud detection and response. CBA has deployed an agentic AI system that monitors over 80 million customer interactions and transactions each day to flag emerging fraud and scam patterns ([5]). The AI doesn’t just identify suspicious activity faster than humans; it also generates new fraud detection rules to counter novel threats on the fly, and has already helped to develop or update 75% of CBA’s card-fraud rules ([6]). To maintain control, any AI-suggested rule changes are vetted by the bank’s fraud analytics team before implementation (human-in-the-loop oversight) ([7]). This blend of AI speed and human governance is becoming a model for managing risk with advanced algorithms.
Early adopters in lending and risk are showcasing what’s possible when AI is applied to decision-making and security. Competitors will be watching closely: if AI can slash credit approval times or catch fraud that humans miss, banks may need to accelerate their own automation plans to remain efficient and secure. At the same time, these examples highlight the importance of robust model risk management. Fully automated credit decisions and self-adjusting fraud controls bring new questions about accountability, fairness, and transparency – areas where banks will need strong frameworks to avoid unintended biases or compliance breaches even as they reap AI’s benefits.
Regulators pivot from principle to oversight.
The rapid rise of AI in financial services is prompting regulators worldwide to reassess their approach. A recent global study found that while more than 80% of financial institutions are using AI, 48% of financial regulators are still not engaged in AI adoption ([1]). This gap in capabilities is becoming a concern as supervisory bodies grapple with how to monitor AI-driven activities. In fact, the chair of the Basel Committee on Banking Supervision warned that, if left unchecked, AI models could even amplify future banking crises ([2]).
In the US, regulators are beginning to act. The Securities and Exchange Commission (SEC) made oversight of AI a priority in its 2026 examination plans, including reviewing whether firms have proper controls for AI tools and can explain AI-driven decisions affecting customers and markets ([3]) ([4]). And at a recent Financial Stability meeting, Federal Reserve Governor Michelle Bowman highlighted how an advanced AI system built by Anthropic, nicknamed “Mythos,” was able to identify cybersecurity vulnerabilities faster than any human – a development she said demonstrates the "dynamic" new risks of AI ([5]). Bowman indicated that U.S. banking regulators are now drafting new guidance to promote the safe adoption of AI in financial services ([6]) ([7]).
Meanwhile, the UK’s Financial Conduct Authority (FCA) has so far insisted that its existing principles (such as requiring fair customer outcomes under the Consumer Duty) can govern AI without the need for new rules ([8]). The FCA has encouraged innovation through initiatives like sandboxes and tech sprints, but it has also hinted at tweaks on the horizon. In a recent policy review for the payments sector, the FCA said it will consider updates to rules to accommodate AI-driven payment agents – acknowledging unresolved issues of consent, liability, and consumer protection when algorithms initiate transactions on customers’ behalf ([9]). Globally, regulators face a delicate balance: they do not want to stifle beneficial financial innovation, but they are making clear that stronger oversight, transparency requirements, and accountability for AI outcomes are likely if the industry cannot demonstrate safe and ethical AI use.
Workforce and operating model shifts.
Financial industry leaders are also weighing how AI will reshape their workforces and organizational structures. Some cost-conscious banks are considering sizable job cuts alongside automation. HSBC grabbed headlines with discussions about reducing up to 20,000 roles – more than 10% of its global headcount – over the next five years as it rolls out AI to streamline processes ([1]). The bank reportedly plans to achieve much of this through attrition and by divesting units, targeting mostly back-office and support roles that AI can replace or augment. If implemented, it would mark one of the largest workforce restructurings in modern banking driven by AI.
Other institutions are focusing on re-skilling rather than layoffs. JPMorgan Chase, for example, has kept its headcount stable while using AI to automate tasks, reallocating employees to new roles as needed. CEO Jamie Dimon recently noted that “we have displaced people with AI – and we offer them other jobs,” emphasizing the bank’s “huge redeployment” strategy to avoid layoffs ([2]). Similarly, many banks are launching internal AI training programs to boost digital skills – Citi implemented broad "prompt engineering" courses for staff, and Lloyds has appointed a Chief Data and AI Officer to guide its AI transformation and workforce upskilling efforts ([3]) ([4]).
As AI becomes embedded in everyday work, the nature of many jobs in finance will change. Mundane tasks – from data entry to routine customer inquiries – are poised to be increasingly handled by AI agents, while human employees focus on higher-value activities like complex problem-solving, relationship management, and oversight of AI-driven processes. Demand for finance professionals with AI and data literacy is rising. For senior executives, this means that a proactive talent strategy is essential: investing in retraining programs, updating job roles, and fostering a culture of human-AI collaboration will be key to unlocking AI’s productivity gains while retaining human talent.