Last month I moderated a panel on “The Impact AI on Jobs and Skills” at the annual 2025 MIT Sloan CIO Symposium. “As AI transforms industry after industry, its impact on the labor force is becoming a critical focus for business, educators, and policy makers,” read the panel’s blurb. “What will be the impact of AI on jobs and skills? Will AI mostly automate human jobs or will it augment human skills and make workers more productive? What human capabilities best complement AI shortcomings? How is AI reshaping workforce dynamics, skill requirements, and career paths. The panel will discuss these important questions as well as strategies on how to best prepare for the emerging age of AI.”
The impact of AI on jobs and skills is a very important subject. In 2022 Congress requested a study by the National Academies on the current and future impact of AI on the US workforce. The three year study was conducted by a committee of experts from universities and private sector institutions. The 140-page report, “Artificial Intelligence and the Future of Work” was released in November of 2024. It concluded that “the speed of technological progress is reshaping not just the tools but also the fabric of the workforce and societal structures. … The trajectories that Al-enabled futures might take can lead to outcomes of profound benefit or significant disruption.”
My fellow panelists were Isabella Loaiza, — postdoctoral associate at the MIT Sloan School of Management; Amita Goyal, — managing partner at the global consulting firm Zinnov; and Vagesh Dave, — global VP and CIO at the energy engineering company McDemott International. The week before the Symposium we held a video call and decided to focus our 40 minute panel on three key topics:
- What will be the overall impact of AI on jobs? Will AI automate most human jobs, or will it mostly augment and create different kinds of jobs that will require new human skills?
- How long will it likely take for AI to be widely deployed across economies and societies? Will it be decades, as has been the case with other major technologies for the past two centuries or will it happen significantly faster?
- What's the likely longer term evolution of AI? How will AI transform companies, education, health care, and our personal lives? Are we likely to achieve artificial general intelligence (AGI) when AI can do just about everything humans can do?
What will be the overall impact of AI on jobs?
Earlier this year, I wrote a blog based on “The EPOCH of AI: Human-Machine Complementarities at Work,” a research paper published by Loiza and MIT professor Roberto Rigobon in November of 2024. I asked Loaiza to summarize the key findings from their research.
She replied that a lot of of what has been written on the future of work, — both in the academic literature and in media articles, — has been focused on the question: “are machines going to automate most jobs?” And, while this question is important, it’s not the most important thing that we should be asking. We wanted to shift the focus from the machines to the humans by asking a different question: “what human capabilities complement AI shortcomings?”
After engaging in a series of interviews with a wide range of experts, Loaiza and Rigobon identified five groups of capabilities that enable humans to do work in the areas where machines are limited which make up the acronym EPOCH: Empathy and Emotional Intelligence; Presence, Networking, and Connectedness; Opinion, Judgment, and Ethics; Creativity and Imagination; and Hope, Vision and Leadership.
These are the capabilities we rely on when making decisions that don’t have a black-and-white answer but are instead based on human judgement, such as a physician diagnosing a complex ailment or a judge deciding on a sentence. These are the capabilities that will enable humans to complement and thrive alongside AI.
How long will it likely take for AI to be widely deployed across economies and societies?
In “AI and the Modern Productivity Paradox,” a 2017 research paper, economists Erik Brynjolfsson, Daniel Rock, and Chad Syverson compared the expected impact of AI to that of other historically transformative technologies throughout the Industrial Revolution. They concluded that, as has been the case with other such transformative technologies, AI will radically change the economy, but realizing its potential will require a large number of complementary co-inventions, investments, and regulations which will likely take considerable time.
There is a considerable difference of opinion between those that believe that AI is a normal historically transformative technology that will take time to deploy at scale and those who argue that AI is in a class by itself whose impact over just the next few years will be enormous.
I asked Amita Goyal how widely deployed is AI among the customers she consults with around the world, and how long she thinks that it will take to deploy AI at scale?
I would break down the question into three parts, she said:
How is AI getting deployed?
- Around 40% of technology companies are already deploying AI in the software development lifecycle, from gathering requirements, to design, development, testing, deployment and so on. This percentage will likely rise to 90% in a few years, but not everywhere in the world.
- A number of her customers are already leveraging AI to automate workflows in internal processes like HR, finance, and procure to pay.
- Others are leveraging AI to enhance the customer experience, although the ROI for this category is very hard to measure
- And, a few companies, mostly startups, are trying to leverage AI to impact the top line of the company.
Where is AI getting deployed? Not surprisingly, this varies widely depending on which region of the world you’re talking about.
What type of companies are already deploying AI? While just about all companies are talking about AI, there is a huge difference in the readiness to implement AI depending on whether we’re talking to a technology or software company or to a cement or consumer package goods company.
The vision, preparedness, and maturity are very different. For example, she was recently speaking to a software company. They are already using AI agents to sell their products and improve the operations of the company. It used to take them a few months to gather the requirements and create a customer solution. “Now they do it in two weeks.”
What's the likely longer term evolution of AI?
Some AI researchers have predicted that “the impact of superhuman AI over the next decade will be enormous, exceeding that of the Industrial Revolution.” But, in a recent essay, “AI Is a Normal Technology,” Princeton professor Arvind Narayanan and his PhD student Sashay Kapoor argued that we should treat AI “not as some looming superintelligence that’s going to go rogue and take over humanity, but as a type of technology like any other, — like electricity, like the internet, like the PC, — that have taken a period of years or even decades to fully diffuse throughout society.”
Vagesh Dave, has already been using AI in the complex engineering projects that he oversees as CIO of McDemott International. I asked him how AI is likely to evolve into the longer future.
He replied: “In Greek mythology, it is said that when Prometheus gave mankind the gift of fire he said that if you learn how to control this and use it carefully, it will do wonders for you. However, if you're careless and it becomes uncontrolled, it has tremendous destructive capability.”
“AI is a technology like that. If you look at the past 250 years, every 50 years, something comes along that transform how we live and work. AI is one of those things. In IT when you conduct a pilot to test different technologies, it’s almost like simply piloting a new flavor of ice cream. But with AI, it’s like piloting an experimental pharmaceutical drug which needs a longer trial period because the side effects may not be felt until you try it for months and months and months.”
“In my industry we are looking at engineering construction. We are looking at efficiency and safety.”
At that point I interjected that when Dave talks about engineering construction, he’s talking about working with energy companies that are looking for oil in the middle of the ocean, so making a mistake can be really dangerous.
“AI can do three things,” he said. First, it can analyze data at scale and speed; second, it can automate tasks. It could be a task with 100 steps, but if they're well-defined, AI can make decisions and automate them; and third, AI has the ability to look at past data without emotions and predict how to improve safety.”
Dave then told us a concrete, real example of how he’s already using AI agents to help improve the safety of workers. In a construction site, you want to make sure that everyone is wearing a hard hat and other safety equipment, and that they don’t get too close to heavy equipment like a forklift. You can ask the workers to watch out for each other, but they might miss something. Alternatively, you can have cameras around the construction site, train the AI agents to monitor the cameras for possible safety violations, and sound an alarm when one is detected.
Making sure that everyone is wearing a hard hat is not such a good job for a human because, for example, it’s rather boring and has limited career advancement potential. But, it’s perfect for an AI agent because the agent does what it’s been programmed to do and doesn’t get bored or tired.
And, Dave added, it’s very important for the safety of workers. “If you do this for months and save one life, it’s all worth it.”
Comments