In a world of infallible artificial intelligence, computers could do most of our work for us. They could diagnose our illnesses in a second. Robots and autonomous vehicles could shop and deliver our groceries. Systems could ensure we don’t break our budgets. AI could operate our transit — planes, trains and cars — without human assistance, and even make our dinner.
That’s the vision of many AI enthusiasts. But the current reality is that while there has been progress, humans are still required to do most jobs. An AI could introduce problems to the workplace, creating risks for workers, their employers and customers, some experts say.
Today, AI can power grocery store robots that change how stores get stocked, speed up vaccine production and generate creative ideas. But the latest advancements raise important questions for workers: How much of our jobs depend on humans? Can technology replace us?
AI won’t entirely replace humans any time soon, industry experts and companies investing in the technology say. But jobs are transforming as AI becomes more accessible.
“Every job will be impacted by AI,” said Pieter den Hamer, vice president of research who covers artificial intelligence at market research firm Gartner. “Most of that will be more augmentation rather than replacing workers.”
Companies have been using AI for years to help crunch large amounts of data to produce insights for their businesses. Some blue-collar jobs have used AI-powered machines to help with warehouse inventory.
White-collar jobs are likely to see the biggest impact near-term, den Hamer said, as AI can be applied at a relatively low cost compared with deploying a fleet of autonomous trucks, for example.
Banking and finance
What’s happening: Large banks have been using AI to improve back-end operations, cybersecurity and power chatbots for faster customer response.
Royal Bank of Canada said it’s testing generative AI to help build software faster. AI can help developers find code they can repurpose for new products or write basic new code, said Martin Wildberger, its executive vice president of innovation and tech.
[Quiz: Did AI make this? Test your knowledge.]
Financial firm Capital One said AI and machine learning are central to its engineering workforce. The bank holds AI and ML patents for fraud detection and natural language processing.
AI advances in the field: Several banks are aiming to offer more personalized financial products and advice, increase the speed of fraud detection to alert customers instantly and remind people of specific bills or spending.
Abhijit Bose, a Capital One senior vice president, said AI could soon monitor transactions to offer more personalized financial advice, insights on spending and saving or quick alerts on deviations from normal spending habits — something as simple as an outlier tip percentage.
Morgan Stanley recently began testing chatbots powered by OpenAI’s GPT-4 with 300 advisers to help them easily pull up research and data. The firm plans to open it up to its 16,000 advisers in upcoming months.
But financial institutions are cautious. AI could introduce risks such as frustrating customers with too much automation, breaking privacy laws aimed at protecting customer’s personal financial data and potentially discriminating against people with lower income.
How jobs might change: RBC is asking workers across functions to become familiar with using AI tools, Wildberger said. It could provide customer service representatives with summaries of complex cases based on previous interactions. And business teams could automate some processes to be more efficient.
“We really focus on the productivity side of tech,” Wildberger said. “Can we automate something to free up [employees’] time?”
Capital One said it’s hiring AI and machine learning engineers, but it’s also upskilling current engineers. Bose said the company has already trained more than 100 engineers through its six-month program.
Health care and pharmaceuticals
What’s happening: Many hospitals use electronic medical records, an area that may benefit from AI for organization and analysis, said Hatim Rahman, an assistant professor at Northwestern University’s Kellogg School of Management who studies AI’s impact on work. And drug development can involve analyzing hundreds of millions of data points, another area where AI could help.
AI advances in the field: Johnson & Johnson sped up the trials of its coronavirus vaccine by using AI to identify hot spots including where variants emerged, said Jim Swanson, executive vice president and chief information officer. It also can help narrow the focus on molecules and identify targets for drug discovery or accelerate image analytics to determine drug effects. And AI supports the manufacturing process for a personalized blood cancer treatment that modifies patients’ own cells.
Swanson said AI also helps guide physicians through procedures like surgeries with augmented reality. As the physician works, it provides guidance on the best next steps. It also helps with reporting adverse events related to drugs by scanning the latest medical literature and flagging reports that need to be reviewed or accelerate image analytics to determine drug effects.
The University of Kansas Health System recently rolled out a generative AI app to more than 140 hospitals. The app, from health-tech company Abridge, records audio of a patient interview, transcribes it, then summarizes important elements to automatically fill out a patient’s medical chart.
“The joy of medicine is helping people get to better health, not the clerical activity,” said Gregory Ator, the health system’s chief medical informatics officer and surgeon. This “just streamlines documentation.”
Generative AI can introduce errors, though, which could be problematic for health-care providers. Abridge highlights parts where reliability of the transcript may be decreased so people can review it, CEO and cardiologist Shiv Rao said.
Some health-care professionals are using AI for cancer screenings, medical imaging and predictions to better detect problems. Google is working with partners such as the Mayo Clinic to validate AI that could automate part of the planning process in radiation treatments for cancer, help mobile ultrasound devices detect early stages of breast cancer or provide vital maternity data without a sonographer and power tuberculosis screenings. But it will probably take years before the technology is ready for professional use, Google said.
If relied on too heavily, AI errors in medical processes could have life-altering consequences.
That’s why Tammy Mahaney, a Bay Area nurse and sonographer, said she checks readings provided by AI-enabled systems to ensure they match what she’s seeing. But she said the tools help her care for more patients.
During a missionary trip to an underserved community in the Galápagos Islands, Mahaney used Butterfly IQ+, an AI-enabled tool that helps perform an ultrasound, interpret it and automatically provide measurements and images on a mobile device such as an iPhone. With the ultrasound, Mahaney determined that a woman in her mid-40s was pregnant and not suffering from a tumor, as she had been told. Still, Mahaney said AI is just a tool.
“You always want to be cautious about diagnosis,” she said of AI. “The limitations are you don’t get the human interaction and instinct.”
How jobs might change: Rao said AI isn’t too far from being able to aid health professionals with decision-making.
“There will be a space for AI to be a thought partner,” he said, adding that the tech could help find the differentiator between two conditions.
In the future, AI could be tied to more devices and wearables for health, Swanson said. Johnson & Johnson aims to digitally upskill 10,000 additional employees this year so they can use the tech to forecast sales or improve operations. And it’s exploring how to use and combine data without bias.
What’s happening: One way big retailers use AI is to track the market price of items, which changes based on factors including supply and logistics, to ensure products are competitively priced. AI can help adjust prices of thousands of products across a store, said Ananda Chakravarty, vice president of research at market intelligence firm International Data Corporation. AI can also help forecast exactly when to drop prices to increase profits.
Retailers are also using AI to schedule workers based on a store’s need, automatically charge people for items with computer vision and recommend products to customers online, Chakravarty added.
AI advances in the field: Sam’s Club, which often serves as tech pilot for Walmart, debuted autonomous floor scrubbers late last year, which in addition to cleaning floors, use computer vision to scan shelves for missing items, low inventory or mislabeled products. The information gets sent back to an ecosystem that could change workers’ priority list. For example, they may need to unload and stock water next if scrubbers determine shelves are empty.
The retailer also uses AI in its virtual voice assistant called Ask Sam, which workers can use to quickly find prices, locate items or help customers. It hopes AI will soon help determine things such as how many croissants workers should bake and automatically alert them when the doughnut count is low, for example.
“We’re moving to where AI is going to be embedded in a lot of things so we can increase associate productivity and reduce friction for members,” Pete Rowe, Sam’s vice president of tech, said.
Looking ahead, retailers might use computer vision to automatically identify whether a customer is old enough to buy alcohol, Chakravarty said, adding that the tech is in early stages of adoption. Generative AI may also soon write product descriptions for thousands of products, said Christian Beckner, the National Retail Federation’s vice president of retail technology and cybersecurity. And AI could crawl social media to automatically design clothes or products based on trends, allowing retailers to get new items to market quickly.
But AI-enabled systems aren’t always well received by everyone. When Walmart rolled out robot cleaners a few years ago, some store associates complained about malfunctions and the time they tended to training the robots (Walmart said the bots appeared to have been well-received). And facial recognition systems have historically suffered flaws, often misidentifying people of color, which could lead to security unfairly targeting Brown and Black people.
How jobs might change: Workers’ jobs are likely to be dictated by what machines deem most important or risk losing money or efficiency for the store. Workers also will probably need to adjust to working with data and tech more frequently, Chakravarty said.
“You don’t have to be an expert, but you need to know how to interpret the data,” he said.
But more AI could mean more risks.
“The key concern would be … the risk of algorithmic discrimination or adverse consequences in how you treat different types of customers,” Beckner said. “There definitely needs to be a level of caution.”
Writing and marketing
What’s happening: One of buzziest forms of AI — generative AI — can produce digital images, conversational text, code and summaries of lengthy documents from a simple prompt. While it’s still in its early days, it has big implications for jobs that include writing, coding or promoting products.
AI advances in the field: Last year, software development platform GitHub debuted GitHub Copilot, a tool that uses OpenAI models to write code based on a user’s prompt. Copilot can suggest methods, unit tests, boilerplate code and complex algorithms, GitHub said.
Some writers are using generative AI tools like ChatGPT to co-write and illustrate books to sell on Amazon. And one legislator used it to help draft a law aimed at regulating AI. Companies such as Microsoft and Google are integrating generative AI tools so office workers can do tasks like write emails or create presentations faster within the apps.
Jonathan Nelson, senior digital marketing manager of growth for the American Marketing Association, said marketers are experimenting with ChatGPT to write articles, including optimizing them for search engines, though they’re not yet publishing those items.
“You have AI write a 1,000-word article, and then go through and edit it to make it sound human again,” he said. “It’s a framework for articles.”
Jeff MacDonald, social strategy director at ad agency Mekanism, said he uses generative AI to brainstorm images for illustrators and designers. He also uses it to scrape TikTok comments and analyze reactions, ideas, and similarities and differences between brands.
But he often uses other tools to double check AI-generated items, as it can make things up or get things wrong, and he avoids using them in finished products. Some AI companies are being sued for scraping copyrighted materials.
“If [AI companies] lose these lawsuits … there’s no saying they can’t go after a brand that used copyrighted imagery,” he said.
How jobs might change: Generative AI tools could help workers become more productive, especially with content creation, said den Hamer of Gartner. That might mean using AI for a first draft and social media posts to solve simple problems or provide summaries of complex topics.
Nelson said though much is still experimental, marketers have a sense that they’ll soon work with AI if they aren’t already — even if it’s just to help determine the success of a campaign. But he said it will be important for the industry to keep human creativity front and center.
“If everyone relies too much on one or two AI [tools], and it operates the same way, do you end up with rampant sameness where nothing stands out?” he said.