FinTech LIVE London: Insights from Starling Bank

November 26, 2024

undefined mins

Jason Maude, Chief Technology Advocate at Starling Bank, explored the vital principles organisations need to implement AI

At FinTech LIVE London Global Summit 2024, Jason Maude delivered an engaging keynote presentation that explored the foundational concepts organisations should consider to utilise AI.

Understanding the purpose behind AI use

Jason Maude begins his presentation by addressing a pivotal question: why use AI?

He notes that organisations often feel pressured by hype to incorporate AI into projects without carefully considering its necessity.

“It can be very easy to get sucked into the hype around AI,” Jason explains, "the question you should be asking yourself is - is this the right tool for the job?”

This principle emphasises that businesses must critically evaluate their goals and determine whether AI is the best solution.

Focusing on clear objectives, Jason warns against investing in AI merely for appearances, as it could lead to abandoned projects akin to failed blockchain initiatives of the past.

Jason Maude, Chief Technology Advocate at Starling Bank, at FinTech LIVE London Global Summit 2024

Laying the groundwork: the importance of data quality

Jason stresses that successful AI implementation hinges on robust data quality and engineering.

“Garbage in, garbage out,” he remarks, underlining that even the best AI models will fail without clean, well-managed data.

He advises organisations to prioritise hiring skilled data engineers alongside data scientists to ensure seamless data pipelines and proper system management.

Beyond foundational data practices, such as those mandated by GDPR, Jason highlights the complexities of training AI models.

For instance, models must balance realistic and diverse datasets when detecting fraud to avoid bias while maintaining relevance.

Synthetic data, he explains, can sometimes address data limitations, but it must be generated carefully to prevent skewing model outcomes.

“You have to spend 90% of your time developing these machine learning models, getting the data correct and making sure that it looks good and works well.”

Jason Maude explore the key principles for implementing AI at FinTech LIVE London 2024

Continuous monitoring: the safeguard against obsolescence

The dynamic nature of data requires constant vigilance to ensure AI models remain effective.

“You have to have excellent monitoring of your models” Jason urges, explaining that usage patterns, consumer behaviour and external conditions are constantly evolving.

AI models can quickly become irrelevant or even harmful if they aren’t continuously observed and updated.

Monitoring must be proactive and ongoing, not limited to quarterly reviews.

This vigilance enables organisations to detect when models stray from their intended boundaries and take swift corrective action.

Jason also recommends maintaining non-AI backups for critical systems to ensure operational continuity in the event of AI failure.

The case for internal AI use

During the Q&A session, Jason addresses a question on whether AI is more suitable for internal or external applications.

He argues that internal use cases often carry fewer risks, particularly when it comes to regulatory compliance.

External-facing AI systems, he warns, can adopt problematic behaviours if influenced directly by customer inputs.

Citing an infamous case of an AI bot that adopted offensive ideologies on Twitter, Jason explains why internal systems tend to provide safer environments for experimentation.

Jason Maude, Chief Technology Advocate at Starling Bank