Q&A: Why Platform Engineering May Be the Missing Link in Banking AI Success

BIZTECH: What makes banking environments uniquely challenging for AI deployment? 

OXENHAM: The primary challenge for financial institutions deploying AI is navigating an exceptionally tight and ever-evolving regulatory landscape while managing decades of technical debt. Unlike lower-risk sectors where AI can be deployed iteratively in a fail-fast fashion, banks operate under strict mandates regarding data sovereignty, privacy and systemic risk. 

Financial institutions cannot rely on black box AI implementations; regulators and compliance officers are required to implement “explainable AI.” If an AI model flags a transaction for fraud or denies a customer a loan, the bank must be able to trace exactly how that decision was made in its audit trails. 

EXPLORE: Follow these critical insights for adopting platform engineering. 

Many financial institutions are realizing that consuming SaaS-based public AI models presents massive systemic challenges to these requirements, whether that manifests as unpredictable API costs, fear of vendor lock-in, compliance headaches, severe data leakage risks or the catastrophic kill-switch risk of relying entirely on a third-party provider, as recent global export control restrictions have highlighted. 

Compounding this regulatory pressure is the sheer complexity of the financial services infrastructure. Most institutions operate in a deeply fragmented environment, where modern, data-hungry cloud-native applications must securely interface with legacy infrastructure. Safely extracting value from these siloed data stores without risking personal data leaks or compliance violations requires an incredibly sophisticated architectural foundation. The ultimate business challenge is balancing this heavy risk posture with the intense market expectation to gain market share in an incredibly saturated market by delivering hyper-personalized, modern consumer experiences.

BIZTECH: What AI bottlenecks can platform engineering help eliminate? 

OXENHAM: The most immediate bottleneck platform engineering eliminates is the immense cognitive load placed on a highly specialized, expensive workforce. Today, many data scientists and systems engineers spend a disproportionate amount of their time wrestling with infrastructure itself, whether that’s configuring Kubernetes clusters, patching Linux systems, optimizing GPU allocations or troubleshooting networking. Platform engineering abstracts this entire layer of operational complexity away. By creating an internal developer platform, banks can provide data scientists with a self-service model, allowing them to focus strictly on building and refining models rather than managing the infrastructure themselves. 

Additionally, platform engineering tames the ever-growing sprawl of the modern AI ecosystem. The open-source AI tooling landscape is moving at an incredible pace, with new frameworks, vector databases and models emerging constantly. 

Without a structured platform approach, organizations suffer from shadow AI, where disparate teams either download unvetted tools onto local environments or leverage external, ungoverned public AI implementations, creating massive security and operational vulnerabilities. Platform engineering standardizes these tools and the underlying infrastructure that supports them into a unified implementation, optimizing software delivery, lifecycle-management, observability, security guardrails and expensive hardware utilization (i.e., GPUs).

BIZTECH: How does platform engineering support the scaling of AI initiatives? 

OXENHAM: Platform engineering transitions an organization from treating AI as a series of bespoke, isolated investments to operating a highly efficient, industrialized assembly line. It accomplishes this through the creation of blueprints, i.e., pre-validated, tightly integrated and fully automated environment templates that, in combination with the underlying infrastructure capabilities, embed security, logging and compliance guardrails right into the foundation. 

This is the exact philosophy behind the SUSE AI Factory concept: a unifying infrastructure platform that serves as the secure enterprise bedrock, handing complete control back to the organization. When a bank successfully deploys its first AI model using a blueprint within this factory framework, it becomes instantly repeatable and scalable. Subsequent teams can launch new AI initiatives in days rather than months, massively accelerating time to market. This shift tackles the single biggest hurdle for enterprises right now: proof of concept to production velocity. The goal is short-circuiting the gap between initial AI experimentation and actual, measurable ROI, and delivering turnkey, actionable solutions that significantly reduce engineering friction and technical debt. 

READ MORE: Capital One uses serverless to move faster with less ops overhead.

This approach also addresses the scaling challenges inherent to hybrid and multicloud environments. Financial institutions frequently train large models in the public cloud to leverage elastic compute on a purely operational expenditure basis, but they may need to run inference on-premises next to core banking data for compliance and data privacy reasons. A robust platform engineering strategy ensures that the underlying operational fabric behaves identically across any environment. It breaks down departmental silos and drives infrastructure consolidation and standardization of tooling, allowing the business to remain agile and adapt to shifting market dynamics without exponentially increasing its operational risk profile. 

BIZTECH: What kinds of use cases could this apply to at financial services institutions? 

OXENHAM: A secure, scalable foundation enables use cases across the entire banking value chain, which can be categorized into front, middle and back-office operations. In the front office, it powers hyperpersonalized wealth management tools and context-aware, generative AI customer service agents capable of solving nuanced user issues safely. In the middle office, where risk management is paramount, a standardized platform allows banks to run low-latency, real-time fraud detection and automated anti-money laundering transaction monitoring across vast, disparate data streams. 

Perhaps the most transformative frontier enabled by platform engineering is in back-office IT operations through agentic workflows. Instead of human operators manually managing system health, specialized AI agents can be deployed within the infrastructure to monitor for anomalies, automatically remediate software vulnerabilities, execute routine software patching and dynamically auto-scale infrastructure resources in response to volatile market trading volumes, including bursting capacity to third-party implementations where compliance restrictions allow. This offloading to agents enables not only a more streamlined operation and enhanced security posture, but more time for human-led innovation, something every organization struggles with. 

BIZTECH: How does platform engineering address security and compliance concerns when it comes to AI projects? 

OXENHAM: Platform engineering fundamentally shifts security and compliance from a reactive, late-stage process to a proactive and fully automated concept known as “shifting left.” Instead of security teams auditing an AI application right before its scheduled launch, compliance guardrails, vulnerability scanning and strict data-masking protocols are baked directly into the platform’s automated pipelines and workflows. This ensures that any model developed or deployed on the platform is compliant by design, significantly reducing human error and the risk of accidental data exposure. 

GET THE DETAILS: What is data sovereignty in financial services?

To truly address these concerns, the financial services industry, like many other sectors, must pivot toward a framework of private enterprise AI. By private, we mean giving institutions absolute, uncompromised control over their intelligence, their data and their intellectual property, securing their technological independence, operational resiliency and business continuity. A well-engineered internal platform provides the clear audit trails that financial regulators demand. The platform automatically logs the entire lifecycle of an AI initiative: where the training data originated, who approved the model deployment, which security protocols were executed and how the model is performing over time. 

By transforming compliance from a manual, exhausting exercise into a continuous, automated byproduct of daily operations, banks and other financial institutions can innovate rapidly without compromising their strict risk standards. In a heavily saturated market, it’s not just differentiation that these organizations need to be concerned with; it’s the reputational risk that comes from a data leak, regulatory investigation or a significant outage of service. Platform engineering considerably enhances this risk posture while enabling the organization to deliver truly differentiated user experiences to continue to attract customers and open up new routes to market.

Click the banner below to subscribe to our newsletter for the latest financial services IT insights.

BIZTECH: How can managed services support banks and FSIs with establishing platform engineering? 

OXENHAM: While it is true that modern financial institutions must operate with the agility of traditional software companies, their true competitive advantage lies in their financial expertise and proprietary data, not in building complex infrastructure from scratch. Designing, securing and maintaining an enterprise-grade IDP requires deep, highly specialized cloud-native expertise. Managed services and enterprise-backed open-source providers allow banks to offload this undifferentiated heavy lifting, essentially buying the foundation to build the differentiation. 

By leveraging external managed services, banks can drastically accelerate their platform engineering journey while de-risking the open-source supply chain. The managed service provider acts as a critical buffer, vetting, stabilizing and securing rapidly evolving upstream open-source projects to meet stringent enterprise standards. This offloads the ongoing cognitive burden of platform upgrades, security patching and infrastructure maintenance, freeing the bank’s internal IT talent to focus exclusively on delivering business-critical AI capabilities and the differentiated user experiences their customers demand. 

DIVE DEEPER: Financial services firms race to operationalize AI. 

This boils down to enterprise-grade operations. Bleeding-edge AI frameworks and open-source blueprints are fantastic, but if you cannot deploy, secure and scale them with trusted operations, the stack will ultimately fail under regulatory scrutiny. Success requires partnering with entities that bring decades of experience running the world’s most mission-critical infrastructure to the AI era — handling the deployment, the lifecycle management, the air-gapped security guardrails and the full-stack validation required by an FSI. 

BIZTECH: Is there anything else financial institutions need to keep in mind about platform engineering related to AI?

OXENHAM: As financial institutions build out their platform strategies, they must fiercely protect themselves against vendor lock-in. The AI landscape is shifting too fast to tie an organization’s future to a single provider’s proprietary stack. It almost goes without saying, but organizations must carefully consider their business continuity plans. 

These companies have spent years accumulating technical debt on top of legacy infrastructure. Platform engineering and the onset of AI enables consolidation and standardization, but it also presents an opportunity to deeply build in technical independence and, ultimately, pivotability. True technical independence means achieving operational resilience, having the ability to run intelligence entirely within your own control and never making your core banking operations subject to an external kill switch. It requires providing a consistent user experience on top of a highly flexible foundation, letting the bank deploy these incredible capabilities wherever it suits the business — whether on-premises, in the cloud or at the edge — all while keeping the full power of supporting software ecosystems intact. This architectural flexibility guarantees that if regulatory requirements shift, or a superior, more cost-effective AI model or hardware accelerator emerges, the bank can pivot seamlessly without rebuilding its entire ecosystem. 

Finally, organizations cannot overlook the critical importance of AI FinOps. The computational costs associated with training and running large language models and AI applications can spin out of control rapidly if left unmonitored. From day one, a bank’s platform must feature deep observability, strict multitenancy with its associated workload isolation and granular cost-attribution capabilities. Knowing exactly how much budget, compute and GPU utilization each department or autonomous AI agent is consuming is non-negotiable for long-term fiscal sustainability.