Artificial intelligence is a hot topic these days. Headlines touting its use in everything from drug discovery to the integration of voice-activated virtual assistants into the workplace seem to pop up daily. Along with these headlines comes no small amount of concern that this technology is on the verge of displacing millions of workers.
While fears of being replaced by a machine are understandable based on history–according to the Bureau of Labor Statistics, there are 7.5 million fewer manufacturing jobs in the US today than in 1980, and automation (along with China) is often blamed–the reality is, AI is being introduced into the workplace to augment and help existing employees do their jobs better, not replace them. The reason for this may be very straightforward: as impressive as it is (think self-driving cars or Google’s DeepMind beating human Go champions) the technology just isn’t capable of doing what humans can do.
At least not yet.
According to Deloitte’s 2019 Global Human Capital Trends report, while 41% of respondents said they are using automation “extensively” across many areas of their organizations, only 38% of respondents said they expect technologies like AI to eliminate jobs at their organizations over the next three years. And only 13% believe the numbers will be significant. This is down from previous years.
The Brookings Institute’s January 2019 report, Automation and Artificial Intelligence: How Machines are Affecting People and Places, suggests that things will get more complicated as time goes on. Even then, only about 25% of the working population are at high-risk of being replaced by AI-powered automation. They list office administration, production (without stating what kinds of production), transportation, and food preparation as some of the most vulnerable occupations.
“At first, technologists issued dystopian alarms about the power of automation and artificial intelligence (AI) to destroy jobs,” the authors write. “Then came a correction, with a wave of reassurances. Now, the discourse appears to be arriving at a more complicated understanding, suggesting that automation will bring neither apocalypse nor utopia, but instead both benefits and stress alike. Such is the ambiguous and sometimes disembodied nature of the ‘future of work’ discussion.”
But, for now, the technology is just not mature enough to create widespread jobs displacement, said Beena Ammanath, AI managing director at Deloitte Consulting.
“The best way to think about it,” she said. “If it takes a human less than 30 seconds to use their brain power to do a certain task, then it is most likely either already automated, or it’s going to be automated in the near future with AI and machine learning. In most cases, that’s where we are with AI. So, in a way, that’s actually enabling humans to remove some of the mundane and boring tasks from their daily work.”
That analysis of AI’s capabilities hasn’t changed much in the past few years, said Ammanath. What has changed is its proliferation. Human resources (HR), for example, is an early adopter of the technology. They are using AI to help ease the burden of everything from setting up new employees with computers and desk space to figuring out employee career advancement strategies to sifting through resumes.
Jobs are safe for now
None of these functions are supplanting existing employees. These decision-support applications, as they are called, are intended to relieve HR personnel of mundane, time-consuming tasks in the first instance, while augmenting their abilities to spot good candidates for open positions in the second. This frees up people to do more intellectually challenging activities like actually making decisions.
“A lot of the adoption, at least that I’m seeing right now, really is machines making decisions or recommendations,” said Greg Layok, managing director, Chicago Technology Digital Product & Transformation at the consulting firm West-Monroe. “But, still, the human is involved making the final call.”
This type of decision support has been integrated into customer resource management (CRM) systems for some time. AI in those instances helps customer service reps (CSRs) by recommending the next-best-offer and/or next-best-action they can use to resolve customer problems.
Another area that is getting attention from AI is robotic process automation, where AI is supplanting rules-based process automation to make it more useful and responsive to different situations. West-Monroe calls this intelligent process automation.
What is happening in some instances, however, is AI is taking on the easy tasks, leaving harder to solve problems for humans to tackle. This can lead to employee burnout, said Jeremy Wortz, a senior architect in West Monroe’s Technology practice.
“This is literally what a client told us, is that, ‘I have a loan processor who’s using intelligent process automation to process bank loans,’ he said, ‘and now, all of a sudden, all of the easy ones are automated because the AI can do that. And that was like a nice little break to keep pressing the button all day and, occasionally, I have a hard one.’ Now, all the hard ones are in front of them all day, and the mental fatigue really kicks in.”
AI still in its infancy
Although the fear of wholesale jobs displacement is real, the reality of workplace AI adoption is still in its infancy and being narrowly focused on specific use cases. Adoption curves also differ widely based on the industry in question, said Chris Havrilla, a senior manager at Deloitte.
“It differs from sector to sector,” she said. “So, in the financial industry where there is heavy documentation compliance regulation … you do see more advances coming in terms of better chatbots and so on. Where things need a more human touch … there are challenges not only with the technology itself but also the challenges around regulation.”
Integration and orchestration
The challenges to wider-adoption comes in two other forms, said Wortz. The first is integrating AI with existing systems and processes. To be effective (or even functional), AI often requires access to vast amounts of data. With data housed in multiple repositories spread out across large organizations, it is a real challenge to make it accessible.
A good example of how AI itself is being used to solve this problem is Kendra from AWS. It is a machine learning search platform designed to help enterprises make better use of their internal and external data stores like relational databases, SharePoint, Box, Salesforce, and cloud by indexing them and allowing users to pose natural language queries (instead of keywords) against the entire repository. The more people search, the better results get, according to the overview page on Amazon’s Kendra website.
The other challenge is orchestration, said Wortz.
“Oftentimes, when you see these solutions that are AI, when you open up the hood from a technical perspective, there’s actually a lot of things going on,” he said. “So, yes, there’s machine learning. But, at the end of the day, the only thing machine learning does is, it produces a … simple prediction. But the ability to orchestrate and connect that with other algorithms, other tools, to gain greater context, is really where we’re seeing the focus. It is the realization [by organizations] that that’s the right investment … so when this technology is ready, you’re ready to go.”
And that is really the point. As interesting and advanced as AI has become, today it is mostly just sorting through massive amounts of data to spot patterns, make recommendations for the next movie or pair of shoes, finishing sentences for you in Google’s Gmail, filling out forms, and answering simple queries from customers.
More advanced uses, like helping radiologist diagnose disease, are still being tested and won’t be ready for general use any time soon, said Mitchell Schnall, MD, chair of the radiology department at the Perelman School of Medicine at the University of Pennsylvania, in an article on DiagnosticImaging.com.
So, for now at least, AI is not displacing employees. And won’t any time soon. In some cases, however, companies are re-training their employees to train the AI the organization is deploying so it understands things like emotional intelligence. This is a good example of what is actually more likely to happen and has happened in the past as new technology is introduced into the workforce: jobs will change.
“There are some jobs that will be displaced, that just won’t be relevant as self-driving cars become more real, for example, or trucking,” said Beena. “There’s definitely a lot of organizations that, more than displacing the workforce, they think of ‘How do we re-skill them so that they can continue to tap into the domain, the core knowledge that exists within those employees?’ There’s a lot of value in the domain knowledge that these employees possess.”