Technology strategy and choices

- To become the AI powerhouse you can be, you'll need to transition from your technology platforms of the past to the technology foundations of the future. Every organization will need to evolve their technology systems, some more than others. In designing the next phase of your technology platform, there are two fundamental principles to take into account, flexibility and scalability. Predicting advances in technology or your business requirements is not just difficult, it's constantly becoming harder. You need to transition to technology systems that will let you readily reconfigure workflows, adopt new approaches, and switch vendors. You need systems that will allow you to change as fast as the business environment. Every change in your technology platforms should create more options. For AI projects to have a real impact, they will need to be able to scale and to grow with your business. This requires both scalable platforms and internal capabilities. These are likely to include, for example, efficient processes for data handling, access to compute power, and modular software architecture. These two characteristics, flexibility and scalability, are greatly enabled by cloud solutions. There are many choices, including whether you adopt public, private, hybrid, or multi-cloud. Cloud providers will give you access to their own and sometimes third party AI systems. This means you will need to consider which AI platforms best suit your needs. Where possible select providers that make it easy to change vendors down the track. The reality is most organizations have what are called legacy technology systems built in the past and not necessarily the most effective for today and tomorrow. For some organizations, the transition to more modern systems that support the latest AI technologies could take time. The core systems that underpin the functions of some banks, for example, are well over a decade old. Data may need to be converted into different formats or reconciled across data silos. You'll need to set a phased integration strategy, including establishing interfaces between your existing ERP and operational systems and the newer platforms you are using for AI projects. Over time, data and applications can be migrated to your new systems. You will certainly need to tap external as well as internal capabilities, including cloud, AI, and software providers. As you make strategic choices on what you build internally and access outside, consider factors including urgency, complexity, availability of resources, and competitive advantage. Logistics and shipping giant, Maersk, has chosen to develop its own AI for business critical applications, as this is at the heart of their competitive differentiation. However, they use off the shelf software, which incorporates AI for more commonplace functions such as HR. You are very likely to draw on professional service providers to deliver and integrate AI projects. However, you also need to develop your own capabilities. When working with service providers, ensure that knowledge transfer and capability development are central to your agreed relationship. Specify the delivery of training programs, upskilling, and implementing in-house processes. There are a raft of different AI frameworks and models that you can use, both in machine learning and generative AI. You should select a set of frameworks or models appropriate to your highest value applications. This allows you to develop your expertise and capabilities around a narrow range of AI frameworks, rather than dealing with a diverse and fragmented set of tools. Particularly for generative AI, you have fundamental choices, including whether to build your own foundation models, adapt open source models, use enterprise models from major AI firms, or fine tune models for your particular objective or use case. Morgan Stanley Wealth Management wanted to provide generative AI capabilities to help its advisors better serve their clients. It opted to use fine tuning, using its own data on an enterprise vendor's large language model. Consistently moving towards more flexible, more scalable technology foundations is a vital underpinning of your AI transformation.