Temenos says banks must resolve data bias to succeed with AI post-Covid

By Sunniva Kolostyak

Banks and financial institutions must address issues around data bias and ‘black-box’ risk governance, as the Covid-19 crisis is intensifying the use of artificial intelligence, according to a Temenos report.

The seventh annual report Overseeing AI: Governing artificial intelligence in banking, conducted by the Economist Intelligence Unit (EIU) for Temenos, has looked at AI governance in the banking sector. It found that 77 per cent of banking execs believe AI will separate winning from losing banks while ensuring ethical, fair and well-documented AI-based decisions will be vital.

The survey is the second in a series focusing on how new technologies will revolutionise banking (released in May) and AI governance, with reports on customer experience and Open Banking to be released later this year.

Speaking to IBS Intelligence, Prema Varadhan, Chief Architect at Temenos, said AI was already a rising trend in the banking sector pre-COVID-19, but the onslaught of the global pandemic has brought it further to the fore.

She said: “Banks are looking at leveraging AI to meet changing customer expectations and find ways to better engage with customers in new ways. This is not possible through traditional methods due to constant change in transactional and customer behaviour, which are time-consuming, hence the need to apply data-driven self-learning methods as in AI.”

The report has highlighted four key governance challenges: ethics and fairness, explainability and traceability, data quality and skills.

“First and foremost, banks need to ensure that the AI models they adopt are fully transparent and explainable. Financial institutions using AI must be able to demonstrate that any decisions made have been done so in a fair and just way. Black box models can be incredibly complex, but decisions cannot be reviewed. “White box” models, like Temenos’ Explainable AI (XAI), can explain in simple human language how decisions are made. This enables banks to identify and resolve bias issues when they arise and win the trust of regulators and customers alike,” Varadhan said.

For AI to work to its best ability, with a minimum chance of bias, it needs to be powered by huge amounts of high-quality data that is both clean as well as purposeful to the desired outcome.

“For banks, this consequently means having the right data architecture in place to stream data real-time and store in the right format so that it is readily available. In reality, this is incredibly hard for incumbent banks to achieve, unlike neo banks who tend to be more data-savvy. Underlying this is the need for data at volume to ensure that algorithms and models are as good as they can possibly be.

The EIU surveyed over 300 people working across all segments of banking, with almost 50 per cent of respondents being either C-suite or top banking executives from Europe, North America, LATAM, APAC, the Middle East and Africa.