Addressing issues will help navigate businesses through the economic landscape for the next decade.
Addressing AI bias and reskilling issues will help CIO’s future proof the business for the next decade and beyond as organisations across APAC continue to increase their reliance on AI systems and platforms.
In an interview with CIO Tech Asia, Richard Morgan, vice president Australia at professional services company, Genpact said every country in APAC is at a different stage of maturity when it comes to AI developments and implementations, AI bias and reskilling are issues that we see across the board.
“Having the right teams in place, and the right skills means that CIOs will be able to build AI engines that are not only fit for purpose but also help the organisation as a whole address growing concerns from consumers and workers,” Morgan said. “It will also help develop opportunities to build stronger customer relationships.”
At a time when the C-Suite will be looking at technology closely to help the business post-COVID-19, CIOs have a chance to actively contribute to building brands and organisations that are fit for purpose and technologies that everyone can rely on without concern in the future.
“What will change from one country to another – and from one organisation to another independent of geographies – is the level of maturity that individual organisations or geographies have reached,” he said.
“Based on that level of maturity and based on each country’s economic and social context around AI priorities, strategies will differ.”
According to Morgan CIOs need to build tailored approaches. For example, CIOs in Australia are under greater pressure to address AI bias. The findings from Genpact research were quite consistent across all regions, including APAC, Europe and the U.S, and overall show that AI bias and reskilling issues are a main concern.
“As 70 per cent of Australians – the highest amongst the countries we surveyed as part of our AI360 report – worry about AI discriminating against them and 67 per cent (highest too) fear that AI will make decisions that affect them without their knowledge,” he said. “CIOs in Japan might feel less external pressure as Japanese consumers tend to be less worried about those issues.
However, a staggering 80 per cent of Japanese workers say their employers often do not offer the necessary AI training, and 84 per cent of senior executives say that companies in their industry generally do not provide equal opportunities to men and women for AI reskilling.”
As for worries and issues around AI, Genpact found that AI benefits overall tend to be the same for companies across the board, said Morgan.
What changes from country to country is the prevalence of each one of those benefits and concerns or issues.
For example:
- For Australian organisations, the main benefit of AI is to improve customer experience, while in Japan freeing up employees’ time emerges as the top AI business benefit
- While a majority of Australians (63 per cent) are more likely to recommend a company that can demonstrate its AI algorithms are bias-free, and more likely to purchase products or services from such businesses (59 per cent), this seems like a lesser decision-making factor for Japanese consumers with only 40 per cent of consumers feeling the same.
The majority of Australian executives believe that integrating AI into various talent processes will help reduce three main issues:
- Gender bias in recruiting (56 per cent)
- Hiring (54 per cent)
- Promotion (62 per cent)
“Australia is leading the AI deployment race compared to other countries, with over a third (36 per cent) of senior executives saying their organisations are implementing AI extensively to fundamentally reimagine their business,” said Morgan. “This number is significantly low in Japan, with just over a quarter (26 per cent) or Japanese executives responding in the same direction meaning that AI implementations should be a bigger priority in this country if Japan doesn’t want to lag behind.”
Morgan said bias occurs when an AI algorithm reflects the siloed thinking and approach of the individual developer and/or when the data set used to train the system is limited. AI bias can be reduced by:
- Building teams that reflect broader experiences and views to develop algorithms
- Creating reskilling programs that target a complete diversity of profiles (gender, background, experience, and more), as these will be the teams that will develop tomorrow’s AI
- Peer reviews of algorithms to build in broader thinking
- Training algorithms with more comprehensive datasets
However, while the same overarching techniques and processes can be used to reduce AI bias and address reskilling issues, each approach is unique and based on factors such as culture which may change from region to region and country to country.
“Taking a successful model from one geography and implementing it ‘as is’ in another is not recommended as local markets are different and what might be a completely appropriate non-biased model in one environment may be entirely non-suited or biased in another,” he said.
“It is important for organisations in the APAC region to be mindful of each country’s unique contextual, cultural, economic and social differences and adapt their AI bias and reskilling models accordingly.”
