Artificial intelligence is gaining traction right across the banking and capital markets (BCM) industry, touching almost every aspect of day-to-day banking functions: fraud detection, risk assessment, compliance and regulatory reporting and algorithmic trading. Banks also can deploy AI in operations – to automate email processing, for example – and in customer engagement to provide tailored advice and guidance.

But a critical issue remains in applying AI successfully: how to close the gap between experimentation and production. Banks and other financial institutions are struggling to mainstream AI, industrialize it and make it work at scale.

Secrets of Applied AI

To reap the benefits of AI, banks need to know how to go successfully from proof of concept to production applications by addressing practical, organizational, technical and regulatory challenges. What paths should banks take to achieve these goals and improve business performance?

Here are some key considerations:

  • Set up a Center of Excellence (COE). Banks are challenged with knowing how to organize around deploying AI. Because disparate business units often use different AI approaches, deploying a Center of Excellence model can bring together expertise in model development and design, while establishing standards around architecture, governance and compliance issues such as explainability and fairness. A COE also acts as a valuable magnet in recruiting AI specialists and in career development, with staff rotating between business units and the COE to cross-pollinate techniques, tools and frameworks.
  • Have the right architecture and infrastructure in place. Industrializing AI requires having a flexible, scalable, cost-effective infrastructure and architecture that can easily adapt to future business needs. Many banks applying AI and ML solutions address infrastructure too late in the project lifecycle, running into limitations in performance and facing high energy costs driven by the huge computational power it can take to train and deploy AI models. Infrastructure needs to be considered at project inception. Cloud-based solutions, for example, can provide the scalability and flexibility needed to handle large AI workloads, but in some cases on-prem private models are needed for security, privacy and competitive reasons. Here investing in the right specialized hardware for the right AI workloads ensures optimal performance and controls costs.
  • Ensure that compliance issues are addressed early. A complex web of regulation governs the use of AI, made up of general law, banking regulations and, in many regions now, AI-specific law. Banks need to develop frameworks, principles and governance to ensure compliance with this web of regulation. For example, AI models must be explainable to both business people and regulators. Safeguards need to be in place to prevent bias and discrimination. Also, banks need to make sure the model that has been developed is the same model that’s deployed, with a clear line of sight across the deployment cycle. Failure to address compliance early on is an expensive mistake, as projects can be delayed, reworked or even scrapped if not properly considered.
  • Avoid data drift and model drift. Maintaining model performance is a key challenge for banks. Banks need to avoid model drift, where the performance of AI models declines over time. Then there is data drift, where the data the model was trained for has changed. Or perhaps the business environment has changed, so the data becomes less relevant. Banks need to monitor model performance and shifts in data on a technical level to avoid data drift and model drift.
  • Focus on obtaining the best talent. Once upon a time, banks were the employer of choice for analysts and data scientists, but nowadays start-ups and tech firms are often perceived as more dynamic and leading edge. Banks have to compete hard in this seller’s market by creating a compelling offer, where the working environment, opportunity to acquire skills and exposure to different technologies are just as important as the financial package. Equally, since the AI space comprises countless niches, banks must develop resourcing strategies to cope with inevitable skills gaps.

How DXC can help

DXC provides the full range of technology expertise and IT services that can help banks and other financial services institutions (FSIs) leverage the power of AI. This includes advisory services, model build, training and testing, integration and development of end-user applications. In addition, DXC provides MLOps services across the full lifecycle, as well as managed services and specialized infrastructure services to address the distinctive requirements of training and running models. This portfolio of AI expertise is bolstered by DXC’s vast ecosystem of global partnerships – which includes the leading AI players in the industry such as Microsoft, AWS and Google Cloud Platform (GCP) – and relationships with AI specialists such as NVIDIA.

With decades of experience, DXC professionals know how to help banks with their AI deployments in practical, meaningful ways. For example, DXC has developed AI solutions to automate email processing, and intelligent chatbots using Generative AI to enable customers to find answers to their specific questions in complex financial product documentation. In addition, DXC provides MLOps services to a European financial services customer where a particular challenge was demonstrating to regulators a clear audit trail from model development and training through to deployment and run. Moreover, through the application of technology such as DXC Platform XTM, DXC leverages AI to provide intelligent observability that predicts, detects and proactively addresses issues so that your IT operations are resilient and robust.

Banks have a good understanding of how AI can help them achieve their business goals. What’s preventing industrialization of AI on a wide scale is a lack of knowledge and expertise in the practical application and adoption of the technology. That’s where DXC can come in to provide the knowledge and support needed to apply AI to help deliver successful business results.

 

Learn more about DXC Banking & Capital Markets and DXC Data & Analytics.


About the authors

David Rimmer is industry managing partner (Banking & Capital Markets), where he assists customers in adopting innovative technologies to grow revenue, reduce cost and successfully engage in today’s intensely competitive banking market. David has a background in AI, having been COO of an AI start-up in France that developed natural language processing (NLP) solutions. Connect with David on LinkedIn.

Dave Wilson is a chief technologist in Banking & Capital Markets, where he works with clients in modernizing legacy enterprise IT, and adopting new technologies. Dave is currently focused on areas such as how best to use operational data analytics and ways in which banks can leverage the metaverse. Connect with Dave on LinkedIn.