*AI: Big gains – and big challenges too*
Strongly Agree
Agree
Neutral
Disagree
Strongly Disagree
By Martin Veitch, Industry Commentator
CIOs see high potential value from their AI and ML investments, but nobody says getting there is easy.
The ultimate accolade bestowed on any emerging technology is that every vendor stretches its meaning to include themselves in the party. Now the media lens is so heavily focused on AI and ML, almost every company I meet is either ‘an AI company’, ‘AI-powered’ or has some sort of angle, no matter how obscure, unconvincing or, well, artificial. Despite this, while many challenges remain, the ceiling on AI is stratospherically high and there are very genuine business opportunities being targeted in many ways by CIOs and their cohorts.
It may be helpful to think of AI and ML as umbrella terms for a series of technologies, the roots of which date all the way back to the mid-twentieth century. The gestation period may be uniquely high but at the current manic rate of development and investment, we are looking at a very broad portfolio.
One way to separate out the various strands here is to consider the reasons for very different rates of adoption of some of these technologies. Hard-core AI remains a tough sell for most enterprises because of uncertain near-to-medium-term return on investment, fears over weak data management foundations and exorbitant costs of research and development. This is hugely exacerbated by the scarcity of software engineers with requisite skill sets, most of whom get hoovered up by the tech startups and superpowers that can entice them with fascinating work, fat salaries or share options.
All the chat
Much current excitement centres on generative AI and the ability to create various types of content using technologies such as ChatGPT, Google Bard and Baidu Ernie. Microsoft’s investment in the first of these provides instant access to scale and effectively forced the hands of others to move fast or be bypassed. But with so much hype, it can be hard to see the nature of the opportunity here and its scale.
While these tools undoubtedly gave some of us exciting eureka moments, such is the ease of use that they offer, I do wonder whether we are moving out of what Gartner would call ‘the peak of inflated expectations’, and descending to the trough of disillusionment. Remember Ask Jeeves in the 1990s? For, ooh, all of five minutes, it was the web’s new toy. A search engine that could handle natural language questions was a novelty, but it was quickly overtaken by other engines.
CIOs and CMOs I talk to certainly seem to be having second thoughts. Generative AI, at least in its present guise, is surprisingly good at assembling content from flimsy requests but as Dr Johnson said of the dog walking on its hind legs, this may be a case of “it’s not done well… but you are surprised to find it done at all”.
Generative AI may be useful in creating reams of passable copy or a first pass of text that can later be edited and it’s sometimes very good at creating images, but it is even now stretched beyond snapping point. Already it’s been co-opted by a dodgy army of content farmers and spam merchants happy to churn out and disperse low-quality wordage on a ‘never mind the quality, feel the width’ basis.
Four out of five agree that AI and ML is necessary to keep their organisation competitive.
It’s also in a grey area when it comes to fair use and plagiarism and the well-known tendency for generative AI to ‘hallucinate’ like a student out of their depth in a pub argument, which could lead to issues of serving up misleading information, or even libel.
That said, given the rate that large language models are being created, the wide recognition of the issues outlined here and the race to add more features, it’s possible that this opinion will date badly and possibly in months rather than years. For now though, as one CTO said to me just today, “it's a case of you get what you pay for”.
Here come the worker bots
Generative AI is the arrowhead for modern AI and something that lets us all encapsulate the startling opportunities on offer but it’s far from being entirely representative of the movement as a whole. One perhaps more dull area that appears to have been obscured by the rise of ChatGPT et al is the highly effective area of robotic process automation.
Maybe because it’s relatively prosaic, RPA can be overlooked at times, but it has matured rapidly as a way to automate processes that machines are good at handling and which are conversely very dull to humans. Expense management, drafting legal documents and business contracts, handling mortgage approvals and credit card issuing are deathly dull, rote tasks. But bots don’t get bored, don’t make clerical errors and have no need for breaks and holidays. Even better, RPA is now a mature approach, so CIOs are aware of any hidden issues and can get on faster with releasing value.
Bots free up human beings to use their emotional quotient skills rather than squander their days on repetitive activities.
Some of the challenges are technical: integrating with other core applications for example. But very often (and this applies to AI and ML generally) they are soft issues.
Work is changing, and with it our notion of ourselves and how we apply our abilities and equip ourselves for a brave new world.
The elephant sitting on the table is of course the understandable fear that people may get automated out of their jobs. This is an issue for leaders who must persuade staff to reach a broader understanding of the benefits and limits of RPA. The short version of this narrative is that machines aren’t good at things that require human emotions, such as empathy, whereas we single-cerebellum, flesh-and-blood dullards struggle to compete with algorithmic speed and scale. Bots free up human beings to use their emotional quotient skills rather than squander their days on repetitive activities. But communicating this honestly and explaining potential repercussions for roles, the need for training, upskilling and so on, are required to buy support from the ground up for automation.
This has always been true of IT, so leaders need to be fearless and unapologetic here, as well as straight talking. AI will change the way humans work but not replace them.
Human+Machine will be an unbeatable combination that helps to bypass concerns over data privacy and assuage concerns over bots running amok. Just as typing pools, carbon copies, dictation secretaries and written memos sent via pneumatic tube networks no longer persist, we all have to accept that work is changing, and with it our notion of ourselves and how we apply our abilities and equip ourselves for a brave new world.
It’s blindingly clear that business leaders already see value. In HCM for example, the chances to accelerate successful recruitment, upskilling, talent management and resource controls can be a game changer.
According to recent research conducted by Vanson Bourne on behalf of Workday, four out of five agree that AI and ML are necessary to keep their organisation competitive, while two-thirds say that they have already seen benefits to efficiency and productivity. But concerns over ethics and privacy are pesky and persistent. The current state of play sees 94 percent of respondents investing and 83 percent maintaining or escalating their stakes.
Smart CIOs will be wise to take a bimodal approach that eschews the fashionable glitz and focuses both on quick wins and longer-term strategic differentiation. A bumpy road lies ahead as we sort through the technical, practical and ethical potholes, but the destination has a luminous summit.
AI will change the way humans work but not replace them.
