AI Talk: The modernisation moment - how AI transforms architectures, operating models and outcomes  

Views on news 

AI is already unlocking significant financial and productivity gains for companies in the technology and telecommunications sectors with productivity improvements of 15% to 25%. AI is now critical for progress but is also creating a much higher bar for CIOs and IT organizations. It requires serious investment in modern platforms, data foundations and governance. 

 

To keep pace with the business, IT teams must build and deploy with the customer at the centre, translating user needs into solutions, while CIOs need to modernize their governance and compliance, streamlining decisions while balancing risk with speed. IT must also tackle the inaccuracies of AI. At this inflection point of AI adoption, IT must reinvent itself to stay relevant.

 

The companies floundering with AI adoption tend to dump AI strategy on the IT team, which often tries to execute it with a classic IT mindset, while those who succeed have a product-centred approach.  

 

The modernisation of applications and platforms 

AI serves different purposes for every company. Some of them just want to add some AI-enabled features to their platform, while others reimagine entire products from the perspective of AI capabilities. Sometimes that requires the complete modernisation of the tech stack too. The cost of modernisation has significantly come down in the past few years, which can even enable businesses to leapfrog directly from mainframes to a microservices architecture. But, due to constraints, the reality on the ground is often a balancing act between legacy systems and state-of-the-art technology.  

 

Mastercard has an AI programme that extends to helping its employees understand AI more and encouraging them to come up with application ideas. To get bigger AI impact, businesses should follow the 80/20 rule, focusing their AI resources intensely on the 20% of their business processes, products or customers that drive 80% of their impact or revenue.  In highly regulated areas such as payments, innovation must be aligned with risks and due diligence. IT’s new role should involve monitoring, managing and mitigating a hugely complex set of risks too, including emerging ones linked to AI. However, IT’s task is not to stall but enable modernisation via managing the risks that only IT experts can fully understand. That said, some experts debate whether it should be IT alone that owns the risks of AI deployments, advocating for a cross-functional AI leadership committee. However, in order to lead the charge, IT must evolve and scale.  

 

So far, customer facing applications and digital platforms bring the best RoI in banking in the areas of reconciliation, document and KYC processing and customer service. One of the greatest RoI can be achieved when an application or product gets upgraded from a legacy to a state-of-the-art SaaS platform. Beside cost saving, there are some other measures of deployment success too: the frequency of deployments, failure rates and recovery times. The speed of software development is accelerating fast. However, the paradox is that the pace of software testing hasn’t increased significantly, which creates bottlenecks. Another challenge is how AI agents can be supplied with more context, where creating test cases based on requirements and feeding them to agents before they start coding can lead to better outputs. While SDLC will continue to remain the backbone of processes, the integration of agentic AI requires some more feedback loops to be added to it in order to get better, faster and more accurate results.  

 

The panel’s advice 

• Technological deployments must go hand in hand with changes in operation models to create value.  
• DevSecOps when integrated with AI require explicit ownership across the value chain.  
• Democratising technology without any expert control from IT can lead to serious technical debt, which may cause user experience and compliance issues too.  
• For a start, map out what percentage of your core workflows have been truly re-designed rather than just augmented  by AI.