The right AI strategy for the Financial Sector
AI is the mega-trend of our time. The potential of this new paradigm appears almost limitless – including in the financial sector. Massive efficiency gains, hyper-personalised services and deeply disruptive innovations are made possible by Generative AI (GenAI). At the same time, however, critical voices are growing louder. Investment plans by tech giants for data centres and AI chips are soaring to dizzying heights. And more than a few GenAI pilots that initially looked attractive are proving, in banks’ day-to-day operations, to be costly dead ends.
Can financial institutions overcome these challenges by simply stepping up their efforts – in other words, by investing even more? Or is the relentless scaling of AI currently running into fundamental limits, possibly even pointing to a bubble? And how can banks, despite this opaque situation, ensure that their offerings remain technologically relevant, deliver a future-proof user experience to customers, and hold their own against competitors?
We address these questions below by taking a more fundamental look at the limits of AI scaling. In doing so, we show how banks can nevertheless secure competitive advantages with AI: through cooperation with external providers that enable a focus on effective AI differentiation.
At present, there appear to be virtually no limits when it comes to AI investment. Whether the challenges outlined above can actually be overcome through ever more aggressive scaling is, however, highly questionable. Many of these challenges are structural in nature. The large language models (LLMs) on which the new solutions are based are purely probabilistic systems. They have no intrinsic understanding of facts or contexts, but instead complete sentences based on statistical likelihoods.
In this form, AI has little to do with genuine intelligence. Nevertheless, it can deliver sensational results – but only if the inherent limitations of LLMs are taken into account, as these cannot be eliminated by even the most extensive scaling:
Despite these barriers, it would in principle be possible to further refine a model through specific training, for example using a bank’s proprietary data. This would be extremely resource-intensive, however, and would still not rule out hallucinations. The same limitations also apply to the next stage of the AI revolution, Agentic AI. AI agents can not only generate content, but also act autonomously. Yet this still happens on the basis of an LLM, augmented by predefined action categories.
Future developments will undoubtedly bring improvements, particularly in learning capabilities as well as speed and cost efficiency. These optimisations will, however, inevitably remain incremental and will not overcome the fundamental limits of scaling.
In addition to technological constraints, there are also financial considerations associated with scaling. The current AI hype, together with exploding investments and costs, reminds many observers of the dotcom era, when the stock-market valuations of internet companies skyrocketed. While that boom, some two decades ago, involved significant excesses followed by a sharp consolidation, the internet ultimately did deliver truly revolutionary advances. What remained was a group of giant technology companies that continue to prosper and drive digitalisation.
A similar trajectory could now unfold around AI – despite the hurdles and barriers described above. Not every company can shoulder the enormous upfront costs, but the technology will nonetheless fundamentally change the world. This also applies to banking: Switzerland’s largest banks are able to build their own AI organisations, operate state-of-the-art models and develop innovative applications that herald a new era of digital banking. These institutions can afford the significant effort involved – from computing power and energy supply to financing and the necessary AI talent.
For smaller financial institutions, the situation is naturally different. To benefit from AI opportunities beyond the hype, however, such a commitment is not necessarily required.
The decisive lever for successful AI innovation in the financial sector is cooperation with external AI specialists such as Contovista. This allows banks to avoid the major investment costs associated with in-house development (see also our blog on the “make or buy” question). The prerequisite is a focus on innovative, customer-centric solutions that deliver genuine value and enable effective market differentiation. This is all the more important given that AI is likely to become a commodity in the near future – a basic, interchangeable resource that no longer offers differentiation in its own right.
Differentiation, by contrast, is precisely what defines Contovista’s solutions. Our AI strategy is tailored to the needs of banks and their customers: Contovista combines deep industry expertise with a sustainable “data moat” and advanced AI capabilities. This “moat” is built on years of experience in analysing account transaction data in Switzerland, which, combined with proprietary AI capabilities, enables highly specific insights and applications. Building such niche expertise would require considerable effort from third parties – and it is this very fact that enables lasting differentiation.
From a technological perspective, Contovista’s approach combines two layers. The first layer consists of robust, specialised systems for categorising transaction data and analysing it using traditional machine learning (ML). To this end, we have developed our Enrichment Engine for data categorisation and our Client Analytics for analysis. This approach is economical, efficient and highly precise, while remaining strongly customer-centric: the analyses deliver actionable insights into financial situations (such as salary trends, new financial obligations, churn signals, property financing, or risk profiles).
On the second layer, we use GenAI to operationalise these insights. With lean, efficient LLMs, we build productive workflows around transaction data: efficient interfaces, automations and decision support, for example for customer advisory processes or cross-selling.
Contovista’s solutions therefore combine domain-specific, ML-based analytics with the benefits of GenAI. The strategic principle is clear: differentiation instead of boundless scaling and the associated costs – or, put differently, scaling only where it makes sense. For financial institutions, this provides a powerful transformation lever whose ROI is likely, in many cases, to exceed that of the costly mega-initiatives pursued by the largest players.
Would you like to learn more about how Swiss banks can navigate the risks of AI hype with the right strategy and implement a focused transformation? Then simply get in touch with our experts.
