Lightning-Fast Loans, New Markets, and New Risks: AI Shakes Up Financial Services

This week, a significant milestone for AI in banking was reached as Canada’s TD Bank announced an “agentic AI” system now handling its mortgage pre-underwriting process. Built by TD’s Layer 6 AI research unit, the tool autonomously gathers and analyzes applicant documents, calculates income, performs consent checks, and generates a concise credit memo for human underwriters. Early results are striking: the AI slashed a task that typically took loan officers up to 15 hours down to under 3 minutes, while improving accuracy. This dramatic reduction in processing time means customers can get mortgage decisions almost instantly, a leap forward in convenience and efficiency.

TD’s new AI underwriter is more than just another workflow tool—it represents one of the first large-scale deployments of autonomous AI agents in core banking operations. The bank sees it as a foundational step toward an AI-enabled enterprise, one that could eventually deliver $1 billion in annual value through faster lending and a greater volume of deals. In practice, that could mean significantly higher mortgage throughput and an edge in a fiercely competitive market. If a major lender can reliably approve home loans in minutes, other banks and mortgage lenders may face pressure to match this speed or risk losing tech-savvy customers. We’re likely at the start of an AI-driven race in retail lending.

For senior executives, TD’s initiative underlines both the potential and the responsibility of AI adoption. These “agentic” AI systems operate with more autonomy than traditional software, tackling complex, multi-step tasks without constant human guidance. That can free up valuable staff time and reduce costs, but it also demands robust oversight. Models making credit decisions must be rigorously validated and explainable to satisfy regulators and maintain customer trust. As banks follow TD’s lead, leadership will need to invest not only in AI capabilities but also in governance frameworks and training so that human experts and AI agents can work hand-in-hand. The opportunity to accelerate credit decisioning is enormous, but it comes with an expectation: AI must augment human judgment, not supplant prudence.

A new development in anti-money laundering (AML) technology this week highlights how AI can unlock markets that traditional banks have written off as too risky. Lagos-based startup SmartComply has expanded to the UK, offering its AI-powered compliance platform to help British financial firms re-establish payment corridors to Africa. This move comes as a response to a decade-long retrenchment: UK banks have cut back more than a quarter of their correspondent banking relationships with Sub-Saharan Africa, largely due to regulatory pressure and high costs of compliance and fraud risk. Yet the demand for cross-border transfers remains strong—UK remittances to Sub-Saharan Africa exceed £4 billion annually, even though fees average a steep 8.5%, far above the UN’s 3% target. There is a clear commercial opportunity for institutions that can manage the risk.

SmartComply’s “Adhere” platform aims to turn this dilemma into an opportunity. Its AI-driven system performs real-time transaction monitoring, automated customer due diligence (KYC/KYB), sanctions screening, and anomaly detection, all tailored to the nuances of African financial networks. By ingesting local data such as Nigeria’s Bank Verification Numbers and patterns from mobile money systems in East Africa, the AI can distinguish legitimate transactions from suspicious ones more accurately than generic compliance tools. According to the company, clients in Africa have already seen a 70% reduction in manual compliance workloads and a 40% drop in false-positive fraud alerts thanks to these capabilities. By greatly reducing the labor and friction involved in compliance, this kind of technology can make serving high-growth emerging markets both safer and more economically viable.

For established banks and payment providers, such advances in RegTech offer a strategic lesson. Markets that were once abandoned due to regulatory risk might now be worth a second look with the right technological support. AI-driven compliance tools can enable re-entry into regions where the balance of risk and reward has shifted, potentially unlocking new revenue streams while keeping regulators onside. Forward-looking financial institutions should evaluate partnerships with specialized fintech firms and invest in next-generation compliance infrastructure, lest they cede ground to more agile competitors. The ability to marry robust AI oversight with local market expertise could become a differentiator, allowing incumbents to win back business in corridors that were previously shut for fear of financial crime.

Even as financial institutions seize on AI’s benefits, regulators are intensifying scrutiny of the risks these technologies may introduce. In an interview with the Financial Times this week, ECB Supervisory Board Vice-Chair Frank Elderson warned that Anthropic’s new “Mythos” AI model has exposed gaping vulnerabilities in banks’ IT defenses. Large US banks with early access to Mythos have reportedly uncovered hundreds, even thousands, of previously hidden software bugs in their systems. Worryingly, some of these “low-risk” weaknesses can be combined by the AI into potent cyber threats that no human team would easily spot. Elderson called the situation “game-changing” and has summoned Eurozone bank executives to address the issue, cautioning that “the clock is ticking” for firms to shore up their cyber resilience.

The immediacy of the threat is a wake-up call. Experts note that advanced AI can reverse-engineer security patches in a matter of minutes, radically compressing the window banks have to respond to newly discovered vulnerabilities. This fundamentally changes the calculus of cyber defense: what used to be a race against days or weeks is now a race against hours or minutes. Regulators like the ECB are urging banks to revamp their incident response and software update processes accordingly, or face severe consequences. We can expect increased supervision of how banks manage AI-related risks, from cybersecurity and fraud to model risk.

Regulatory bodies globally are likewise rethinking their approach to oversight in the age of AI. In the US, the SEC has signaled concern that financial institutions might become dangerously dependent on a handful of opaque AI models or data providers, raising new systemic risk questions. In the UK, the Financial Conduct Authority has reopened its “AI SandBox” and commissioned studies on AI’s impact on retail finance. Broadly, watchdogs are emphasizing that existing rules around operational resilience, data protection, and model governance still apply – and may need strengthening. For C-suite leaders, the message is clear: aggressive AI adoption must go hand-in-hand with investment in security, transparency, and robust oversight. The cost of a misstep, whether a massive breach or a biased AI credit decision, could swiftly attract regulatory action and reputational damage.

The march of AI is also prompting difficult questions about the future of the financial services workforce. Just last week, Standard Chartered’s chief executive announced plans to replace roughly 7,000 to 8,000 jobs – about 15% of the bank’s support and back-office roles – with AI and automation by 2030. This blunt declaration, framing those roles as “lower-value human capital,” drew public backlash and even inquiries from regulators in Asia. While the phrasing was controversial, the strategy reflects an industry-wide drive for efficiency: analysts estimate over 200,000 jobs across European banks could disappear by the end of the decade as AI and digitalization streamline processes. From routine credit processing to compliance checks, many functions once handled by junior and mid-level staff are squarely in the path of AI-driven automation.

Leaders contemplating such transformations must balance cost efficiencies with long-term resilience. The allure of a 30% jump in productivity and significant cost savings is real, but so are the risks of cutting too deep or too fast. Loss of institutional knowledge and experience is one concern—even Standard Chartered has emphasized retraining and redeployment for affected employees, to retain valuable know-how. Other bank executives have cautioned that if junior staff never learn fundamental skills due to over-reliance on AI, it could undermine the industry’s future human talent pipeline.

Rather than viewing AI purely as a tool for headcount reduction, forward-looking organizations are reframing it as a way to elevate the workforce. Roles are evolving: mundane data processing can be automated, while employees focus on higher-value advisory, relationship management, and creative problem-solving tasks. New positions are also emerging in areas like data science, AI governance, and model risk oversight. The strategic takeaway for C-level executives is to integrate AI into operating models in a way that enhances human productivity and decision quality. That means investing in training, redefining job scopes, and fostering a culture where human expertise is amplified by AI – not sidelined by it.