The race is on. Like many industries, UK retail banks are energetically pursuing many different AI opportunities, each aiming to be first to exploit the promise of better customer experiences, reduced operational costs and new revenue streams.
But as they surge forward, there is a greater need than ever for a sense of control and visibility over the cloud resource needed and how it is going to be paid for. This is not a gold rush: banks know that they need to keep things in proportion. The real aim is to create a coherent and sustainable, organisation-wide AI capability. Whether this involves enriching the user experience with GenAI or delivering AI-enhanced business functions that are time critical, the key is to turn the art of the possible into enduring business advantage.
And this is why FinOps is going to prove invaluable in managing the best AI outcomes for cost, sustainability and business value.
The fragmentation of the cloud
As different parts of the organisation race off at different speeds with different agendas, there is a risk of inefficiency as separate pockets of cloud resource are used, driven by innovation in AI.
There are also different types of workload. Some lighter generative AI tasks, for example, might use standard cloud infrastructure that is relatively easy to manage and optimise. But there are more compute-intensive requirements too, such as fraud management, which is not just a complex task, it also needs to be done quickly since customers are waiting. This might demand GPU technology – graphical processing units, which enable many tasks to be processed in parallel. GPUs offered by public cloud providers raise the overall processing capacity enormously and deliver the high processing requirements than many AI workloads require. However, this power costs more than conventional cloud infrastructure and AI workloads are not always constant, so it is even more important that the usage and cost of these resources are effectively managed and optimised with a FinOps approach.
Google Cloud also uniquely offers TPUs, processors aimed at workloads such as AI that operate in a ‘neural network’, working together to achieve AI processing collaboratively. This can deliver AI workloads very quickly and efficiently.
With different workloads, a range of charging models and a generally distributed approach to AI experimentation and development, retail banks are working in an increasingly fragmented environment.
Why banks need full-stack FinOps
FinOps gives retail banks the two things they need right now if they want to instil some order and efficiency into the headlong rush for AI-inspired value: it provides visibility, which then drives cost-efficiency.
Visibility means being able to see into the detail with accurate cost, usage and (if required) sustainability data, and thus to make more considered decisions on how compute resource is used. FinOps is about overall efficiency as well as continuous improvement and not just the headline cost.
Another key area for AI is data because AI consumes and is dependent on data to be effective. Managing the cost of data is equally important when delivering AI and drilling into unit costs such as data cost per query using Unit Economics, for example, can highlight how cost-efficiently data is accessed. In a recent survey conducted by the FinOps Foundation, the key priority was “reducing waste or unused resources” and this is one area where AI costs can be managed and reduced.
Another essential consideration is to apply FinOps principles across the whole cloud portfolio. The full-stack FinOps approach used by NTT DATA has helped us optimise costs for our clients across both public and private cloud helping reduce, reuse and reallocate, thus releasing funds for hard-pressed IT departments who are under pressure to innovate while also dealing with finite budgets. Whatever cloud solution you use, we can build that capability with full stack FinOps anywhere, show a joined-up cost perspective and understand how they are made up.
Leading cloud providers such as Google Cloud are working hard to develop pricing models and features that reflect the way cloud use is changing, but in the meantime there is a pressing need to understand how it is being used and how to optimise. Winning the AI race is not just about speed. The retail banks that make best use of FinOps are the ones most likely to succeed with a co-ordinated use of AI across the organisation that actually translates into a cost-effective business advantage.
Want to know more about how AI is being used in the UK banking sector – and how to accelerate its implementation? FinOps is just one of a number of key issues discussed in these recent e-books from NTT DATA:
The Top Five Retail Banking Use Cases for traditional and generative AI
Five ways to accelerate value in AI in retail banking