
一项新的Finextra调查显示,英国和爱尔兰绝大多数高级银行高管感到被迫在内部准备就绪之前加速人工智能项目。尽管降低成本、欺诈检测和个性化服务的承诺不可否认,但受访者指出了三个关键摩擦点:监管合规、人才短缺和遗留系统集成。
这项针对42家主要银行的运营总监、首席技术官和创新负责人的调查显示,78%的人感受到来自监管机构、行业同行和市场预期的外部压力,要求在接下来的12个月内交付AI驱动的解决方案。然而,只有34%的人认为当前的治理框架能够充分评估英国金融行为监管局(FCA)和欧洲银行管理局(EBA)所要求的模型风险、数据来源和可解释性要求。
人才短缺加剧了这一困境。高管们报告称,62%的AI相关职位仍然空缺,而且与科技巨头争夺数据科学家和机器学习工程师导致薪资超出了传统银行的预算。这种短缺迫使许多机构依赖第三方供应商,从而引发了关于供应商风险管理和数据主权的问题。
遗留的核心银行平台增加了另一层复杂性。虽然许多银行已经探索了“即插即用”的AI模块,但要在不干扰交易处理的情况下进行集成仍然是一项昂贵且耗时的努力。Hyperlayer最近筹集的1950万英镑凸显了市场对将AI能力从核心系统中抽象出来的中间件的需求,但此类解决方案仍处于早期阶段,需要严格的测试。
从更广泛的生态系统角度来看,加速采用AI的压力可能会催生一波标准化浪潮。监管机构可能会发布关于模型验证和审计跟踪的更明确指南,促使供应商在其工具包中嵌入合规检查点。相反,如果模型产生偏见结果或在压力下失败,过早部署可能会带来声誉损害,从而可能引发更严格的监管审查。
对于首席财务官和金融科技构建者来说,要点很明确:速度必须与严格的风险管理相平衡。投资稳健的模型治理平台、提升内部人才技能以及建立明确的供应商风险协议,将区分那些实现可持续AI价值的银行和在匆忙中出错的银行。
所表达的观点基于公开可用的调查数据,不构成财务建议。读者在采取任何与AI相关的投资或运营决策之前,应进行独立的尽职调查。
图片:Luke Chesser / Unsplash (https://unsplash.com/@lukechesser)
New BIS research shows fintech platforms like Wise now offer significantly lower costs for cross-border transfers than traditional banks and stablecoins, which still face friction issues.

Visa’s live pilot with Billtrust demonstrates how trusted AI agents can automate invoice settlement, promising speed gains but also raising compliance and risk considerations.

EPC’s new framework aims to standardise EU digital identity wallets for PSPs, promising smoother onboarding and AI‑enhanced AML checks.

A new Finextra and ACI Worldwide report, 'Payments 2030: The Payoff Year,' offers crucial insights into the evolving payment landscape. This analysis explores how AI agents and advanced technologies will be central to navigating future challenges and unlocking efficiencies for financial institutions.

评论 (2)
I'm curious, what specific regulatory compliance challenges are bank executives in the UK and Ireland facing with AI deployments, given the FCA and EBA mandates?
The biggest friction point is the FCA and Central Bank of Ireland's strict operational resilience mandates for third-party model risk management. When algorithms act as black boxes making credit decisions, proving auditability to regulators without violating proprietary IP is proving to be a massive hurdle for CFOs.
The 78% acceleration pressure versus the 34% governance readiness gap is a classic "move fast and break things" scenario, but in banking, the "things" that break are usually customer trust and regulatory standing. I’m skeptical that banks can outsource the talent gap without sacrificing the internal domain knowledge required to actually operationalize these models for leaner processes. If the legacy integration costs aren’t factored into the ROI, these projects will stall in the pilot phase.
You hit the exact friction point, because relying on third-party vendors without building internal model risk management creates a regulatory blind spot that auditors won't forgive. If CFOs don't reallocate budget from maintenance to domain upskilling right now, those stalled pilots are going to turn into expensive write-downs by Q3.