Executives across industries have poured billions into artificial intelligence systems. Yet many watch their expensive tools sit idle or deliver disappointing results. The missing piece sits in plain sight. Companies that treat workers as partners in the AI rollout see real gains. Those that don’t watch productivity stall and morale drop.
Communications of the ACM laid out the pattern clearly in a report published today. Only 7 percent of AI spending goes toward people and training. The other 93 percent buys technology. https://cacm.acm.org/news/investing-in-workers-to-work-with-ai/ That split explains why so many initiatives falter. Employees fear replacement. They lack the skills to prompt models effectively. Without guidance they misuse tools or avoid them entirely.
Accenture attacked the problem head on. The firm sent a memo to senior employees. Use AI tools regularly or risk missing promotion. The message landed. Partners with OpenAI, Anthropic and Palantir supplied the platforms. Training turned anxiety into adoption.
Citigroup went further. Its “Citi AI” program trained more than 182,000 colleagues. Over 70 percent now use the tools. They generated 21 million prompts in 2025 alone. Shobhit Varshney, head of AI at the bank, described the goal. “To ensure our employees are armed with future-ready skills we are committed to teaching them how to best use these tools to increase their productivity.” A module called Prompt like a Pro taught clear communication with models. One well-crafted prompt, Varshney noted, “can transform a simple query into a powerful analysis.”
Law firms face their own reckoning. The billable hour model clashes with AI that completes research in minutes. Andrew Johnson, CIO at Brownstein Hyatt Farber Schreck, saw the tension. “AI is incredibly disruptive to that model. So much can be accomplished if the technology is used correctly.” The firm required training through AltaClaro. More than 90 percent of its 250 attorneys and legal staff finished the four-hour courses that combined study, assignments and group practice. Johnson offered a blunt warning. “It’s almost going to be perceived as malpractice to not use AI.”
But. Training must fit the context. Generic courses produce generic results. PerfectServe, a healthcare communications company, built monthly series tailored to its 400 employees. Modules covered specific tools, prompting techniques and safety. More than half participated. Laura Suarez, vice president of AI, watched attitudes change. “People are hungry to continue to upskill. The bigger thing is a behavior shift and seeing what is possible. AI is meant to assist folks and let them focus on more high-value tasks.”
Prisma took a similar path. The e-commerce firm embedded reassurance in its AI mission statement. Focus on “doing administrative tasks faster, not AI is taking your job.” Allison Georgoulis, vice president of e-commerce, said the language mattered. “We are building this assurance into the language of our enterprise-wide AI mission and policies. We see this investment as an advantage to our employees.” The company created an AI council of champions and worked with a partner to build three internal chatbots. Alignment with HiTRUST security standards gave employees confidence the tools were safe.
Alexey Karnaukh, cofounder of LinkBuilder agency, started training for a simpler reason. “We saw real fear in the team. AI is by no means a replacement, but a reinforcement.” That message, repeated across successful programs, reduces resistance. Workers who understand the complementary role embrace the technology.
The numbers tell a cautionary tale. A Smartcat report of 200 enterprise leaders found 58 percent rely on self-serve learning or offer no formal AI training at all. https://www.smartcat.com/global-growth-report-2026/ Meanwhile AI budgets swell. Fortune reported global spending on track to hit levels that reflect a 44 percent increase in 2026 while training budgets grow just 5 percent. Average learning time per employee actually fell from 47 to 40 hours. https://fortune.com/2026/03/17/ai-economy-workplace-investment-human-potential-competitive-advantage/ Sixty percent of knowledge workers say they have received no formal training on the AI tools they now use daily.
This mismatch carries consequences. A National Bureau of Economic Research paper examined firms that invest heavily in AI. They shift toward more educated workforces with higher shares of STEM and IT skills. Hierarchical layers flatten. Junior roles grow while middle management shrinks. The pattern favors those already equipped with strong technical foundations. https://www.nber.org/system/files/working_papers/w31325/w31325.pdf
Policy makers see the gap too. An opinion piece in Communications of the ACM called for broad labor measures. Programs from the U.S. Department of Labor offer adult job training at low or no cost. South Carolina gives tax credits to firms that hire apprentices. The authors argued such incentives should expand beyond traditional trades into digital skills. https://cacm.acm.org/opinion/generative-ai-requires-broad-labor-policy-considerations/
Recent analyses reinforce the point. The Atlantic Council warned in May that corporations must earn worker trust to capture returns on AI spending. Without buy-in, productivity gains remain theoretical. Transparent communication about skill shifts and early involvement of domain experts improve outcomes. https://www.atlanticcouncil.org/blogs/geotech-cues/to-realize-returns-on-their-ai-investments-corporations-must-consider-their-workers/
Some organizations act on the evidence. Walmart partnered with Google to train 1.6 million employees in AI fundamentals. Amazon committed more than $1.2 billion to upskill workers for technology-enabled roles. Mastercard built an internal talent marketplace that uses AI to match employees to growth opportunities. SAP folded continuous learning time and wellbeing supports into the workweek. IBM doubled down on entry-level hiring, betting that young workers trained in durable skills will drive the next wave of innovation.
McKinsey researchers examined the broader picture. Demand for AI fluency and technical skills rose sharply between 2023 and 2025. The consulting firm concluded that investing in workers and their skills, not just in technology, decides whether human potential expands or narrows. Leaders who engage directly with AI, invest in human capabilities and balance gains with responsibility produce better results. https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-skill-partnerships-in-the-age-of-ai
The invisible labor that powers AI also deserves attention. A February report from Communications of the ACM highlighted the low-paid data raters and labelers who train models for firms such as Scale AI and Sama. These workers, often paid as little as $2 per hour, remain hidden behind subcontractor arrangements. Antoine Casilli, a researcher cited in the piece, argued that forcing companies to employ them directly and pay fairly would raise costs but create incentives for better business models. https://cacm.acm.org/news/the-invisible-labor-force-powering-ai/
So the choice sharpens. Firms can chase short-term savings by cutting training while scaling AI. Or they can accept the higher upfront cost of structured programs, governance frameworks and continuous learning. The second path demands more from leaders. They must redesign roles, measure new kinds of productivity and communicate honestly about how work changes.
Evidence accumulates that the second path pays off. Employees who receive contextual training become champions. They spot opportunities executives miss. They save time on repetitive work and redirect effort toward judgment-intensive tasks. Organizations that flatten hierarchies while raising educational requirements create workforces better suited to AI collaboration.
Yet most companies still bet against their own people. They spend on models and infrastructure but starve the human layer that makes those systems valuable. The gap between AI hype and actual results will persist until that changes.
Citigroup’s prompt volume and PerfectServe’s behavior shifts offer concrete proof. Training works when done right. It reduces fear. It builds competence. It turns potential disruption into measured progress. The firms that master this combination will set the standard for the next decade. The rest risk watching their expensive technology deliver far less than promised.
WebProNews is an iEntry Publication