Why the US is at risk of losing the AI talent and productivity war

The hardest thing to manage is change. I wrote that line more than a decade ago in an article about the “XPocalypse,” Microsoft’s end-of-life deadline for Windows XP. My argument then was that the real crisis was not obsolete software. It was the shortage of technically literate professionals capable of guiding organizations through inevitable transitions.

More than a decade later, the names have changed. The lesson has not.

Y2K defined the pattern. The risk was real, but disaster was avoided because skilled people did the work. When nothing happened at midnight (1999-2000), many assumed the threat had been exaggerated instead of recognizing that it had been managed. Windows XP became the next version of the same problem. The operating system stayed embedded in retail, banking, healthcare, energy, law enforcement and defense systems long after it should have been retired. The vulnerability was real, but the larger lesson was mostly missed: organizations let technical debt pile up until a deadline turns it into a crisis.

Is agentic AI actually breaking the enterprise SaaS business model?

Now we have the “SaaSpocalypse.” Headlines warn that agentic AI is breaking the SaaS business model, lowering software valuations and making entire categories of enterprise tools obsolete. Investors are reacting; analysts are talking about “FOBO,” Fear of Becoming Obsolete, and organizations are again asking whether they are ready for what comes next.

The disruption is real. AI agents can now automate workflows that once required dedicated software tools and teams of human operators. The per-seat pricing model that powered two decades of SaaS economics is under pressure. But the apocalyptic framing misdiagnoses the problem. SaaS is not dying. It is bifurcating.

Platforms requiring precision, auditability, complex state management and regulatory accountability, such as financial systems, healthcare records and compliance infrastructure, will remain essential. What is collapsing is the undifferentiated middle: horizontal tools that AI agents can replicate cheaply and at scale.

The organizations most exposed are not simply those using the wrong software. They are those who outsourced technical judgment along with technical execution. They bought SaaS as a substitute for internal capability, accumulated organizational debt and now lack the human capital to navigate a transition that is fundamentally about people and process.

The old taxonomy still applies: people, process and technology. Technology serves business functions. Processes create efficiency. Qualified people sustain both. But the pace of technological change continues to outrun the education system’s ability to produce experienced professionals with current skills.

AI has widened that gap. Data engineers now design orchestration infrastructure that determines whether AI produces value or liability. Security practitioners must govern autonomous agents acting on behalf of enterprises. Business leaders need enough technical fluency to make build-versus-buy decisions in a market changing in real time.

These are not narrow technical tasks. They are the applied outputs of serious STEM education grounded in a business context, professional standards and sustained practice. We are still not producing enough people who have those skills.

How is the growing STEM education gap threatening AI leadership?

The numbers are sobering. The United States now produces fewer than 820,000 STEM graduates annually, representing about 20% of all degrees awarded. China produces approximately 3.57 million STEM graduates each year, about 40% of its university degrees. At the doctoral level, the gap is sharper. In 2000, the United States awarded 17,830 STEM PhDs, compared with China’s 7,520. By 2022, China awarded more than 50,970 STEM doctorates, over 50% more than the 33,820 awarded in the United States.

This matters directly to AI leadership. Countries building the strongest STEM pipelines today are positioning themselves to define the architecture, governance and standards of AI systems tomorrow.

How can we solve the AI talent shortage and rebuild the IT profession?

More than a decade ago, I argued that IT must be treated as a profession, not merely a resource. Finance, medicine, law, engineering and accounting all have formal professional pathways, standards and institutional support. Information technology underpins nearly every critical function of modern society, yet still lacks equivalent professional frameworks.

The AI transition makes this more urgent. As AI absorbs routine execution, the humans left in the loop must be more capable, not fewer. Their role is shifting from implementation to governance, from configuration to architecture, from maintenance to judgment. That requires better preparation, stronger incentives and professional recognition.

The United States still leads in private AI investment, but it has not matched that commitment with investment in the human capital needed to sustain it. China has embedded AI degree programs across more than 500 universities and integrated corporations directly into research and workforce pipelines. India’s AI upskilling surge is driven heavily by corporate sponsorship, with employers treating workforce education as strategic investment. The European Union has committed significant public funding to AI talent development and cross-border STEM mobility.

The United States has examples worth scaling. North Carolina’s AI Academy at NC State, built with more than 100 corporate partners, combines university credentialing with applied workplace training. North Carolina A&T, the nation’s leading producer of Black engineers, is partnering with NVIDIA and the Office of Naval Research to expand AI and cybersecurity talent. Texas has committed heavily to doctoral research infrastructure through the Texas Institute for Electronics, linking universities, government and industry around semiconductor and defense technology priorities.

These models show what a national strategy should look like: public investment, corporate sponsorship, university research capacity and continuous pathways from undergraduate study through doctoral work. But they remain exceptions. Corporate PhD fellowships from leading technology companies are valuable, but they are filters, not pipelines.

The technology sector has long harvested talent from a pipeline it does not adequately fund, then wondered why the pipeline runs short. That model is no longer sustainable. Federal and state governments must create the policy environment, including tax incentives, credentialing reform, research funding and visa frameworks, that makes corporate STEM investment structurally attractive rather than reputationally optional.

The SaaSpocalypse will pass, as Y2K and the XPocalypse passed, because capable people will do the work. The headlines will move on. The underlying shortage will remain.

What I called for in 2014 still stands: STEM education, paired with business, information management and finance, must become a sustained national infrastructure. Not as a reaction to this disruption, but as preparation for the next one.

The hardest thing to manage is change. The next is learning from it.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?