6 cornerstones of financial services reinvention with data and generative AI
Before I joined Microsoft in 2021, most of my professional life was spent on a trading floor. I was an analyst in Sydney in the 1990s and later in London and New York, and through all of it, there was one constant: computer screens. At one point, I think I got up to 12 simultaneously.
Today, this feels like a fading memory. Technology has advanced the financial services landscape, fundamentally and forever, and the trend is only accelerating. With Microsoft Cloud for Financial Services, the future is being redefined in real time, as Microsoft, our global partners, and customers across the industry are pushing the boundaries of what advanced analytics and AI can achieve.
Since the introduction of ChatGPT one year ago, the interest in advanced data and generative AI has exploded in a way we haven’t seen in a generation. Financial services firms are moving quickly to evaluate and deploy solutions, and not because they have suddenly stopped being risk averse (they have not), but rather because they recognize the risk of doing nothing and quickly finding themselves at a competitive disadvantage.
In my work with leaders across financial services companies, I’ve found it valuable to set the stage in understanding the AI opportunity with a set of core principles, which I’d like to share in this blog post.
Microsoft Cloud for Financial Services
The future of financial services in the era of AI
6 principles for successful adoption of advanced data and AI
The following six principles can help illuminate the journey to successfully evaluating and implementing data and AI solutions in financial services.
1. Redefine value
In the early days of my career, I relied on a specialist software platform that delivered the data and insights I needed to do my job. I synthesized that information with any other data I could imagine and find (hence the overload of screens). The metric of value in that world was my daily active usage (DAU) of that platform—for example, my eyes glued to the screen. That began to change when the platform could alert me about something important based on predetermined criteria, and later when a bot could ping me even when I was off the platform.
AI transformation in action—how organizations are innovating today
Read the blogWith the arrival of generative AI, firms need to think differently about the value of their technology investments. It’s no longer about measuring the amount of time users spend on a specialist platform; it’s about understanding how well technology empowers people in the right moments and the right ways. It’s about data working constantly in the background, and the platform engaging users in the most high-value, low-friction ways possible.
2. Meet users where they are
Once you embrace the power of automation, alerts, and bots, the next question is how to extend specialist platforms with high-value notifications, without incurring new costs and headaches. Over the past 18 months, many firms have had a major ah-ha moment, realizing that the answer is already on most of their employees’ desktops.
Microsoft Teams is widely deployed in capital markets firms across the industry, and it is designed explicitly to enable things like distributing alerts and enabling bots—even serving as a unified front end. Microsoft Excel is also nearly ubiquitous, and it is likewise built for integration, especially when coupled with Azure and Microsoft Power Platform. So now we’re seeing a wholesale embrace of firms taking advantage of these solutions, which not only add value to users’ daily workflows, but also “springboard” them back to specialty platforms when need be. Critically, this also provides the benefits of security, compliance, and productivity that are essential to deploying mission-critical, AI-based solutions.
3. Deliver timely, actionable insights
Investment bankers, asset managers, brokers, and other financial services professionals appreciate a good insight and dislike unnecessary distractions. A good insight is one that provides the right information at exactly the right moment to help the employee decide or close out an “approve, deny, investigate” workflow. This is a high bar that requires the ability to compress and structure data and deliver it in either a graphical or tabular format, which previously would have required the use of specialist software, but now can be delivered directly into Teams and automated easily and precisely through the use of a tool like Data Activator (now in public preview for users of Microsoft Fabric).
A bad insight is one that only adds noise to the workday. This is a matter of personal preference, which is why it’s critical that insights be hyper-personalized. People must be empowered to control how they receive notifications—in what tools and on what terms. Firms will of course have requirements for the delivery of certain types of information. But people should be empowered to design and curate their own notifications, modify them as they go, and learn what works best for them over time with the help of AI.
4. Atomize workflows
It’s not difficult to imagine the impact of timely, actionable insights on both efficiency and better business results. But it can be daunting to consider how to architect them and bring them to life.
The answer is to “atomize” workflows, by which I mean examining the full range of tasks involved in key organizational and business processes, breaking them down into their component parts, then asking what resources are best suited to each. This sets the stage for creating new or reinvented workflows with AI-powered capabilities in optimal roles, so that you can best automate, optimize, and evolve processes over time. A good tool to help with this process is Microsoft Power Automate, a cloud-based service that lets you create and run flows that connect various services and apps, including sending notifications across the full set of Microsoft apps.
The goal is not just greater efficiency; AI presents the likelihood that there will be entirely new tasks, workflows, and workers. The explosion of unstructured data and the democratization of AI (that is, making it broadly useful even to non-technical users) are giving rise to things that are impossible to even imagine today within the constraints of human capital costs.
5. Embrace hybrid intelligence
In 1997, Garry Kasparov, the reigning world chess champion, was beaten by a supercomputer. However, it wasn’t long before the combination of humans and chess-playing programs performed better than any computer.1 It’s not that human intelligence is greater than AI. It’s that hybrid intelligence—a mix of the two—is greater than either alone. The implication for building better workflows is that you don’t want to replace human effort with AI, but rather deploy it in ways that let people do their most important work better, while handing ancillary tasks off to AI.
This is in addition to what I call augmented intelligence—a human doing all the work, only better with technology. With hybrid intelligence, there are some tasks where a human does not need to be involved. End-of-day reporting to regulatory authorities, for example, is a human-capital intensive task in which augmented intelligence is being employed now to improve efficiency. But where a firm might have a huge volume of data, hybrid intelligence might be the better option. The key consideration is knowing when and where a human needs to be in the loop. This obviously requires careful planning, evaluation, and oversight.
6. Build on Azure OpenAI Service
As employees in financial services firms put generative AI to work with the recently released new Microsoft 365 Copilot for commercial customers, the groundswell of interest and demand for customized solutions is growing dramatically. In my work with customers, I stress that the fastest and most productive route to such innovation is to build on Microsoft Azure OpenAI Service. It provides a powerful platform to build new solutions for use cases such as customer service, risk management, financial crime detection, and portfolio optimization. These can take the form of Microsoft 365 Copilot plugins when you want to empower users with your data, or Azure-based copilots for user experiences within your own applications.
What makes Azure OpenAI Service so attractive is that it gives you the power of ChatGPT and GPT-4 while being deployed on your firm’s Azure tenant, so that all data and content stay within the bounds of the organization. That makes it easy to build intelligent solutions that take advantage of advanced machine learning and natural language processing capabilities, while maintaining security and compliance. Augmenting it even further is Microsoft Fabric, a new data analytics platform designed to simplify and streamline data and analytics workflows and lay the foundation for the era of AI.
Where to go next
Most financial services firms are well into their generative AI journeys, although it is still the early days for all of us. In our work with customers, we review these six principles and challenge the business and IT leadership to understand the unique needs of their business, where AI will deliver the most value, and how to make smart early moves in technology investments and early use cases.
The critical first step of the journey is to get your infrastructure in order and complete your migration to a hyperscale cloud platform, taking care to also include a comprehensive data strategy. Next, you can identify the specific scenarios with the greatest business impact for your firm over the long run, work with your partners and with Microsoft to identify the data sets that are required to light up those scenarios, and invest in the relevant solutions and start exploring.
You can learn more about Microsoft Cloud for Financial Services by visiting our website or contacting your Microsoft representative or partner.
In my next blog, I will update you on our partnership with London Stock Exchange Group (LSEG) and examine key aspects and lessons that have relevance for every financial services organization eager to embrace advanced data and generative AI.
