AI Is Poised to Revolutionize Work — Or Wreck It

Global corporate spending on artificial intelligence solutions is on track to approach $200 billion by the end of this year, with more than a quarter of organizations already deploying generative AI tools. But even as 60% of HR leaders say AI is a top priority, nearly half of AI companies with AI projects had abandoned most of them in 2025. That’s more than just wasted investment — it’s a reputational risk and strategic failure for the leaders who backed them.

The stakes are clear: AI has the potential to transform HR by streamlining operations, improving decision making, and elevating the employee experience. But too often, it’s implemented without clear goals, oversight, or alignment with how people actually work. The result? Tools that go unused, systems that amplify bias, and organizations that are left scrambling to recover from costly missteps.

The real problem isn’t the technology. It’s how it’s being introduced — without enough transparency, oversight, or understanding of how people actually work. As a result, many HR leaders find themselves in the same trap: racing to adopt AI without the right strategy, safeguards, or support systems in place.

Still, when AI is done right — with a clear purpose, ethical guardrails, and human-centered design — it can be transformative. For HR, the opportunity isn’t just to automate. It’s to lead. Because the real revolution isn’t AI itself — it’s how organizations choose to work alongside it.

AI’s Meteoric Rise — and Predictable Fall

AI’s rapid rise over the past two years has catapulted it from back-office algorithms to boardroom obsession. It’s now seen as the magic bullet for everything from hiring and onboarding to performance management and culture-building. HR has been especially fertile ground for adoption, with AI automating administrative tasks and freeing up teams to focus on strategy, experience, and employee well-being.

But as with every revolutionary technology before it — fax machines, PCs, email — AI is already falling into a familiar pattern. First comes the hype. Then comes disappointment. Then slowly, the real value emerges. It’s a cycle well known to analysts and leaders alike, and it offers a helpful lens for understanding where AI sits today.

Today, AI is riding a wave of high expectations, but many HR teams are discovering it’s harder to realize the promise than expected. Without a clear strategy, tools can go unused, systems may fall short, and leaders may find themselves adjusting course midstream. The challenge isn’t to stop the cycle — it’s to navigate it carefully, avoiding common pitfalls such as chasing tech without purpose, working in silos, ignoring ethics, and skipping change management.

An EY survey released in December found that 54% of senior leaders felt like failures as AI leaders. Additionally, 53% said their employees felt exhausted and overwhelmed by the pace of new AI information and developments, and 65% said they had trouble motivating their workers to accept AI technology. Clearly, the AI workplace revolution isn’t going to be an easy one.

Where AI Goes Wrong

Despite its promise, AI adoption in HR is already littered with missteps, many of them predictable. Automating dysfunction and ignoring ethics often go hand in hand. These failures usually stem from poor training data or flawed HR processes — problems that AI doesn’t fix, but rather magnifies. It’s the latest version of an old programmer’s lament: Garbage in, garbage out.

Built to Fail

The biggest early adopter AI disaster in the HR space blew up when Amazon used its own early version to screen job applications, only to find that the AI prioritized resumes from men and penalized any mention of “women,” as in “women’s chess club,” and downgraded graduates of two women’s colleges. In 2018, Reuters revealed that the system had been trained using earlier applications from the mostly male candidates in the tech industry. The project was promptly scrapped, with just a small portion salvaged for redevelopment. that the large language model had been trained using earlier applications from the mostly male candidates in the tech industry.  The project was promptly scrapped with just a small portion salvaged for redevelopment.

Similarly, but much more recently, tutoring company iTutorGroup settled a $365,000 suit brought by the Equal Employment Opportunity Commission (EEOC) in 2023 after its AI recruiting software automatically rejected more than 200 candidates who were women ages 55 and older or men ages 60 and older. “Even when technology automates the discrimination, the employer is still responsible,” then-EEOC Chair Charlotte Burrows said.

Going beyond discrimination into outright illegality, a Microsoft-powered chatbot launched by New York City’s government in 2024 advised entrepreneurs and business owners they were legally allowed to take part of their workers’ tips, and that it was perfectly legal to fire workers who complained about sexual harassment. As of late 2024, according to CIO.com, “the chatbot remains online,” accompanied by a pop-up with the dire warning to verify any answers and avoid using them “as legal or professional advice.”

These AI risks remain a long-running issue that continues to flare up even after a decade of cautionary cases. More recently, a 2023 suit against software provider Workday charging that its AI technology discriminated against applicants on issues of age, race, and disabilities was expanded in May, going from one single plaintiff to a collective action with four additional plaintiffs. 

AI Errors Aren’t Just External Risks

Another risk of disaster can result from a lack of ongoing monitoring of AI, especially generative AI solutions which, in addition to fines, lawsuits and reputational damage, can wreak ongoing havoc with an organization’s internal operations, said Pragya Gupta, chief product and technology officer at isolved HCM, a provider of HR, payroll, and benefits solutions. Gupta recalled how, after launching an applicant tracking solution, isolved added functionality to generate job descriptions for recruiting.

“One day, the AI decided to include salary information in the job description,” she said. “It was not part of our prompts. It just decided that one day it was a great idea.”

An isolved client had recently acquired a subsidiary where the employees were being paid much less than the salary ranges listed in the new AI-generated job descriptions, and it didn’t take long for the underpaid workers to notice.

“The customer was saying, ‘Oh my God, we have a mini-revolt within our company right now.’ It created a major issue within the company,” Gupta said. “Regular audits that inspect and adapt the process must be part of deploying technology. Those are either automated checks that say, ‘Hey, make sure there’s no salary information before you publish’ or there are humans looking [at and] approving those job descriptions.”

Build Confidence Through Collaboration

Gupta stresses that successful AI implementation requires focusing on more than the capabilities of technology and considering how AI is going to be adopted and integrated into daily work.

“If you decide to bring AI on, what does your adoption road map look like?” she asked. “Yes, you now have a smart car, but do you even know how to drive that car? The key is helping employees adopt the technology that’s being offered. It’s not going to change their world overnight, but it’s going to be this gradual process of integration until you get to the point where people say, ‘I don’t know how to live without it’ because that’s how they operate now.”

Gupta offered the example of an AI-powered candidate matching application that scanned resumes, sorting through 1,000 applications to find the 10 best candidates. The recruiters doubted the process, however, and felt they still needed to sort through the entire batch of resumes. The solution was to allow recruiters to grade the resumes selected by AI with a plus, if they thought it was a good match for the job, or a minus if the match was wrong. That not only trained the AI to become more accurate but gave recruiters confidence that the resulting matches were correct.

“In the beginning, we didn’t see an increase in usage but a dip,” Gupta said. “As we introduced that feature, we started seeing the usage gradually go until we saw close to a 38% reduction in time to hire. People were like, ‘Yep, these 20 matches are great. I’m going to call them, I’ll find my match and make the hire.’”

Don’t Just Add Tools — Rethink Work

Previous new technologies have typically created better tools to handle existing tasks. PCs were initially deployed as highly improved typewriters and calculators that used word processing and spreadsheet software to accelerate productivity, improve accuracy, and raise the level of work being created. Those initial uses grew into networks that created new means of communication, connectivity, and collaboration.

For decades in publishing, computers were simply considered new-and-improved typesetting machines rather than the world-changing digital content platforms they are now. Similar missteps can occur when organizations deploy AI as a simple tool, rather than integrated as part of an open-ended reimagination of work processes.

“There’s an organization that I’m not going to name that spent over $700,000 on buying Microsoft Copilot licenses and training everyone on how to use Copilot, and no one uses it,” said Jason Averbook, senior partner and global HR digital transformation & AI leader at Mercer. “The challenge with AI is that we’ve jumped right to technology and left the human behind.”

“We work with an organization that has 17 chatbots, including one for compensation and one for payroll, and they’re asking, ‘Why aren’t our employees using these chatbots?’ It’s because they don’t know which one to use, because the chatbots don’t talk to each other,” Averbook said. “It would be like going to a specific Google for the weather, another Google for sports, another Google for research on bears, and another research on Google for black shirts.”

Thinking of AI as simply a new technological tool means that organizations will fail to fully integrate their AI technology into rethinking the daily workflows of the people expected to use AI. And that’s the real AI disaster unfolding at this very moment, experts say: a wave of AI deployments where the technology is simply being ignored and going unused.

“AI adoption’s not going to happen unless we have what I think of as ‘AI embodiment,’ which means my people embody a way of working and accept a way of working,” Averbook said. “They’re not just going to flock to a tool just because I say, ‘We now have AI.’ I have to embody a different mindset, which is that the way that we work now allows me to be my best self.”

That shift — from deployment to embodiment — is where the real work begins.

AI Is HR’s Big Moment

HR’s role in AI adoption starts with a deceptively simple question: What real problems are we trying to solve?

Too often, companies roll out flashy AI tools without a plan — copying competitors or chasing hype, said Julia Stiglitz, CEO and co-founder of the AI learning platform Uplimit. She recalled one client who was looking to expand its customer education program, where AI wasn’t part of the initial conversation. As Uplimit developed a solution, it turned out that AI offered a platform for scaling the process.

“It has to start with solving an actual problem,” Stiglitz said. “When you’re demonstrating tools, things can look really flashy or cool or interesting, but if it’s going to be a technology that is lasting within your company, that grows within your company, and has an impact within your company, it needs to be solving a real pain point.”

Vaccine manufacturer Moderna took a bold step in that direction in May, announcing that its IT and HR departments would merge as the company redesigns its teams to leverage roles best handled by technology and others best handled by humans. The aim is to eliminate some roles, reimagine others, and create entirely new functions.

This realization — that AI is not a technology but an entirely new way of working — puts HR leaders squarely at the forefront to lead this evolution in the nature of work, said Naama Manova-Twito, founder and CEO of MarkeTeam.

“AI is going to change everything we know about workforce planning and the landscape of organizational charts,” Manova-Twito said. “The organizations that are going to win will be the ones finding the right balance between the right AI agents and the right people to work alongside and manage those agents. HR leadership is going to be the single most important position in a company. HR leaders will have to be able to have a bird’s-eye view of the company, the budget, the scaling needs, and be able to manage workforce planning with hybrid teams.”

The greatest potential AI disaster to avoid now seems to be squandering the opportunity to embrace the creation of hybrid teams of humans and AI agents into a concept many experts are calling human-machine teaming, where humans and AI machines team up to drive business outcomes. AI fails when the focus is on how to use AI tools without developing the necessary digital mindset.

Brian J. O’Connor is an award-winning writer, journalist, and editor specializing in explanatory reporting on business, finance, economics, and technology, as well as executive ghostwriting. He earned nine national writing awards as the financial columnist for The Detroit News and is a contributing writer to The New York Times, MarketWatch, and other national outlets, as well as secretary of the nonprofit Cygnet Institute of Personal Financial Literacy.