The AI reckoning in banking: Why moving fast (and failing faster) is the only way forward

AI, especially large language models (LLMs), is evolving at a breakneck pace, leaving banks both excited and overwhelmed. The technology is advancing so quickly that by the time a new system is implemented, it can already feel outdated. But one thing is certain: standing still isn’t an option. We have spent a lot of time grappling with not only how we utilise the evolving technologies within our own space, but also how best to advise our clients on where and how they can implement AI in a way that doesn’t result in wasted resources. Throughout this journey we’ve learnt some valuable lessons, specifically within the financial services space.

Three pillars of AI adoption in banking

Banks are already leveraging AI in three key areas:

Staff efficiency tools – AI-driven tools are helping knowledge workers automate repetitive tasks and streamline workflows, freeing up time for higher-value work. Almost no role remains untouched by generative AI tools.
Operational efficiency automations – AI is reengineering operational processes, improving efficiency without requiring a full-scale system overhaul. This includes the likes of customer service and process automation such as OCR (optical character recognition) and document processing. However, these automations are wholly volume dependent and don’t automatically make sense for each organisation, or every use case within that organisation.
Customer-facing AI – The holy grail: digital assistants that act as ‘digital twins’ of human bankers. We’re not there yet, but the industry is moving closer every day. Right now, AI-powered chatbots and virtual assistants are handling routine customer queries, fraud detection, and even some basic financial advice. The real challenge? Building AI that understands context, intent and the nuances of human conversation. Until then, hybrid models - where AI supports, rather than replaces, human bankers - are proving to be the best path forward, ensuring smoother, more personalised customer experiencer while laying the groundwork for the future.

Lesson 1: The proof-of-concept imperative

The biggest hurdle with AI isn’t just technical, it’s a mindset shift. Unlike traditional enterprise tech, AI doesn’t come with guaranteed, neatly packaged outcomes. That’s a tough pill to swallow in an industry built on risk management and regulatory compliance. But the key to unlocking AI’s potential isn’t endless analysis - it’s action.

The best approach? Rapid prototyping. Instead of waiting for the perfect strategy, banks should focus on small, high-impact AI experiments. Find a clear use case, implement a solution, measure the impact, and adjust quickly. If something isn’t working, pivot, because in AI, indecision is the real risk.

Lesson 2: The hard reality - AI integration is no easy feat

AI also isn’t plug-and-play. Successful implementation requires expertise across multiple disciplines - data science, engineering, cloud computing, and machine learning. Without the right engineering disciplines, infrastructure and integration experience, scaling AI beyond the POC into a productionised capability that can enhance customer experience is an uphill battle.

Another challenge? Justifying AI investment. Banking leaders must move beyond traditional ROI-based thinking and embrace a new success model, one that prioritises incremental progress, adaptability and long-term value creation.

Our advice: Move fast, fail fast

Based on the various projects we’ve successfully implemented, and the lessons learnt along the way, here’s what we are currently advising our banking clients:

Be agile and lightweight in adopting and in replacing
Don’t over invest – AI technologies are experiencing rapid advances and corresponding cost declines. Use this trend to your advantage
Architect solutions for agility, prioritising modular, pluggable and loosely coupled systems
Finally, don’t adopt early if there isn’t a clear business case

What’s next? A bumpy (but exhilarating) ride

The future of banking won’t be shaped by simply layering AI onto old processes - it will come from rethinking banking itself. Just as placing computers on desks in the ‘80s didn’t automatically boost productivity, plugging AI into legacy workflows won’t drive real transformation. In many cases banks need to re-engineer their processes from the ground up, assessing the way AI can reshape how financial services are delivered.

This won’t be smooth sailing. AI-powered banking will require constant recalibration, bold decision-making, and a willingness to embrace the unknown. But for those who take an iterative approach failing fast, learning faster and moving forward with intent - the rewards will be transformative.

The AI revolution in banking isn’t coming - it’s already here. The only question is: will you lead the charge or be left playing catch-up?

A leading international provider of bespoke software solutions, BBD’s four decades of technical and domain expertise spans the education, financial services, insurance, gaming, telecommunications and public sectors. BBD employs over 1200 highly skilled, motivated and experienced IT professionals, curating flexible teams from our hubs across South Africa, India, Netherlands, Portugal and the UK.