FILE - The OpenAI logo is seen on a mobile phone in front of a computer screen displaying output from ChatGPT, March 21, 2023, in Boston. (AP Photo/Michael Dwyer, File)
SINGAPORE: Generative artificial intelligence (Gen AI) promised a roaring business revolution, but 18 months after ChatGPT’s big-bang debut, its impact is more like a whimper.
Enterprises had underestimated what they needed to use gen AI and overrated its impact on businesses, observers said.
Consultancy firm McKinsey, which in July 2023 projected gen AI to add a US$4.4 trillion bump to global corporate profits annually, said in April that adoption among corporations has not gone far beyond Gen AI pilots and a few use cases or situations in which the technology could be used.
Gayatri Shenai, a senior partner at the firm, said in a webinar: “The value really comes from enterprise-wide adoption and scale. And here, most organisations are struggling.”
The hubris following ChatGPT’s debut in November 2022 might explain a few let-downs.
Some of the test projects that firms picked in 2023 were not scalable – like launching Gen AI for a function performed by eight people – and would not have delivered value, said Matthew Candy, IBM Consulting’s global managing partner for Gen AI.
“There was pressure, questions being asked from above, and lots of people were feeling like they’ve got to do something, show something,” he said.
Results of a poll of 300 executives by the MIT Technology Review Insights for Telstra International in late 2023 found that first adopters of Gen AI reported lower confidence in their firm’s technology than others.
Commenting on the findings, MIT Sloan School research fellow Michael Schrage said: “It is disappointing how tactical that experimentation has been. Too many experts look at Gen AI as a way of automating or augmenting existing workflows and processes, rather than rethinking use case fundamentals or the desired outputs and outcomes they really want.”
Hallucination – the inclination of Gen AI to fabricate stuff based on probability – is a major obstacle.
OCBC Bank, which launched a ChatGPT bot in October 2023, has 250 AI models and a 150-strong data science and engineering team that oversees them.
It uses a multitude of tools to test the accuracy, limitations and biases of its large language models.
Donald MacDonald, the bank’s head of group data office, conceded: “Gen AI got many companies excited with the possibilities, but they are now finding it challenging to deploy in production.”
Organisations are also falling short of AI enablers – such as an AI strategy, talent, data, infrastructure, operating models and culture.
Firms need to have a fit-for-purpose data architecture, sufficient computing power and robust connectivity to handle the immense data volumes, said Geraldine Kor, Telstra International’s head of global enterprise.
Data that is good enough to generate real-time insights is another top constraint.
In a survey conducted by Salesforce and YouGov around March, about eight in every 10 Singapore workers said the data ingested by their AI models must be accurate, comprehensive and secure, before they would use them with trust.
“For real-world applications, we simply cannot rely on luck. We need people to trust the technology,” Professor Lam Kwok Yan, a cyber-security expert from the school of computing at Nanyang Technological University, said at a conference on May 7.
Then there is the shortage of AI talent, with no short-term relief in sight.
In a global poll of 2,000 executives in nine developed economies by recruiter Adecco and Oxford Economics in late 2023, Singapore topped the list for employers who planned to hire AI talent from outside their firms.
Given that in 2023, 34% of resident employees lasted only one to five years in a job, Betul Genc, Adecco’s head of Asean, was not surprised. Managers may find it more financially viable to hire over reskilling, she said, even if that is not sustainable.
She said: “This approach risks creating a skills shortage that may drive salaries upwards, intensifying competition in the job market.”
Still, employers are too tied up with the AI rat race to focus on training their employees, according to another study by workforce solutions provider Persolkelly.
In its February survey, only 27% of local employers said they have a comprehensive plan to help employees replaced by Gen AI find new jobs, compared with the Asia-Pacific regional average of 36%.
“Despite being acknowledged as the company’s most significant asset, employees are nevertheless frequently disregarded,” it wrote in the report.
The question is how many businesses will be ready to deploy Gen AI effectively.
Workers will remain at the core.
Brad Anderson, president of products, user experience and engineering at experience management firm Qualtrics, said: “Qualtrics research shows that Singaporeans are some of the most open to using and engaging with AI globally. But when we look at how they want to use AI, it’s clear that customers and employees want AI in the passenger seat assisting them, not the driver’s seat replacing them.”
Clients are using its AI tools to push up customer responses to human agents, and help managers act on both customer and staff feedback, he added.
Businesses will find out that a patchwork of models and technologies, and not one Gen AI model, is what they need, IBM’s Candy said.
He said: “People are starting to realise the complexity of the landscape that’s going to evolve, and therefore the need to think about how I govern it, how I manage it, how I build a fabric and a layer that can connect all of these different technologies and skills.”
Companies without data good enough to use Gen AI might start in areas where data reliability is less critical, suggested Sujith Abraham, general manager of Salesforce in Asean.
“For instance, account summaries or case summaries are not as reliant on data quality. Businesses can start there and continue to improve and refine its data by experimentation,” he said.
Ang Yuit, president of the Association of Small and Medium Enterprises, said many small and medium enterprises are using gen AI tools for personal productivity tasks rather than job or process transformation.
These firms could integrate gen AI into business processes, such as enhancing customer service through sentiment analysis or for swift and accurate first-level customer inquiries.
“A 20% increase in productivity for the owner or team can significantly impact operations,” Ang said. — The Straits Times/ANN