Managed AI in the USA: What almost everyone overlooks in the current AI hype

Too risky, too complex: The real reason why America's corporations are now outsourcing their AI

Public discourse on artificial intelligence usually revolves around spectacular language models, gigantic data centers, and the multi-billion-dollar arms race among the major tech giants. But away from this limelight, a quiet yet profound transformation is underway in the US economy. The focus is no longer simply on developing the models, but on their smooth, secure, and profitable operation in everyday business: so-called managed AI. Driven by an acute shortage of skilled workers, exploding operating costs, and an increasingly complex regulatory landscape, US corporations are relinquishing control of their algorithms to specialized service providers. This article examines why the real AI boom is taking place in the seemingly insignificant maintenance and operations business, debunks the myths surrounding common analyst figures, and reveals who is truly profiting in this new economy of artificial intelligence.

Why America's corporations have long since relinquished control of their own algorithms to third parties

The public debate about artificial intelligence in the United States revolves almost exclusively around language models, data centers, and the race between the major platform providers. What's being overlooked is a market that is perhaps even more economically significant than the competition for the most powerful model: the operation of these systems in day-to-day business operations. Managed AI, the ongoing, contractually secured management of productive AI applications, has transformed within just a few years from a niche topic to one of the fastest-growing segments of the American service sector. Anyone who wants to understand where the US economy is headed in the coming years must ask not only who builds the best models, but also who keeps them running, monitors them, secures them, and is responsible for them in a crisis.

From experiment to mandatory operation: How the market is reorganizing

The difference between a one-off AI implementation and a continuously operating system may sound technical, but it is economically crucial. Implementation consulting ends with the handover of a functioning prototype. Managed AI begins precisely where this handover ends and encompasses monitoring, the ongoing operation of training and inference systems, governance processes, cost control, the orchestration of autonomous agents, and incident response – all under contractually agreed service levels. This shift is no accident, but rather the logical consequence of a maturation process: Companies that initially experimented with pilot projects now face the challenge of operating production systems permanently, securely, and economically, and very few have the internal personnel to accomplish this alone.

This very skills gap, combined with increasing regulatory pressure and the sheer complexity of continuously managing AI costs and model drift, is driving the demand for outsourced operations. Hyperscalers like Amazon, Microsoft, Google, IBM, and Oracle are increasingly producing their own managed inference offerings, while system integrators and traditional consulting firms are taking over operations, governance, and industry-specific use cases. A hybrid ecosystem is emerging between these two camps, in which the platform provides the computing power and the service provider assumes responsibility for smooth, compliant operation.

Billions in tailwinds from Washington and Wall Street

The political and economic climate of the past twelve months has further accelerated this trend. In June 2026, the government signed an executive order to promote advanced AI innovation and security, which provides for voluntary pre-approval of particularly high-performing models and establishes a clearinghouse for AI cybersecurity, coupled with guidelines from the Cybersecurity and Information Security Agency (CISA) for federal agencies and critical infrastructure. As early as December 2025, the government had attempted, through another executive order and a specially created legal department, to enforce a uniform, minimally burdensome federal framework while simultaneously defending itself legally against differing regulations in individual states. For companies operating AI systems in multiple states, this coexistence of federal policy and a growing patchwork of state laws creates significant legal uncertainty, which in turn fuels the demand for specialized governance services.

In parallel, market analysts are painting a picture of explosive growth. A forecast published in May 2026 puts global AI spending for the current year at around $2.59 trillion, an increase of 47 percent compared to the previous year, with AI services alone expected to account for approximately $585 billion. The analysts responsible explicitly identify 2026 as a turning point, at which the investment burden shifts from the capital expenditures of hyperscalers to the operating budgets of the companies themselves. A market analysis presented in August 2026 supports this thesis with a remarkable observation: As early as 2025, spending on the ongoing operation of AI models, i.e., managed inference, exceeded spending on training new models for the first time, reaching $23.1 billion compared to $16.3 billion. This shift marks the transition from a research-driven to an operational economy.

Who really pays: A look behind the analysts' advertising brochures

As tempting as clear-cut rankings may be, the available data doesn't stand up to close scrutiny. There are no official, standardized statistics on which industries in the United States spend the most on operating AI systems. Different market research firms arrive at conflicting results, depending on their definitions and methodologies. One study sees the financial and insurance sector clearly in the lead, with a share of just over a quarter of the US market for autonomous AI agents. Another study, based on venture capital perspectives, identifies healthcare as by far the largest spending segment in the field of generative AI, with an estimated volume of around $1.5 billion in 2025 alone. A third analysis, based on official surveys from the Census Bureau, shows that—measured by pure usage rates—the information sector leads with just under 40 percent and the financial sector with about 34 percent, while healthcare lags significantly behind in actual operational application.

These discrepancies are not accidental, but rather reflect a fundamental problem of definition. Those who talk about spending on generative AI often conflate software licenses, one-off consulting projects, infrastructure costs, and the actual operation. Those who speak of usage rates are measuring something different than those who inquire about contract volumes for outsourced operations. Furthermore, official surveys by the Census Bureau, conducted between the end of 2025 and mid-2026, show that overall AI usage among American companies is only 18 to 22 percent and is heavily concentrated in large companies with more than 250 employees, where the rate rises to approximately 37 percent. Therefore, there is no question of widespread penetration of AI in the American economy. The demand for professionally operated AI has so far been a phenomenon of large corporations and knowledge-intensive industries, while small and medium-sized enterprises (SMEs), despite aggressive marketing from many providers, offer hardly any reliable purchase figures.

Managed AI Platform - Image: Xpert.Digital

Here you will learn how your company can implement customized AI solutions quickly, securely and without high entry barriers.

A managed AI platform is your all-inclusive, worry-free solution for artificial intelligence. Instead of dealing with complex technology, expensive infrastructure, and lengthy development processes, you receive a ready-made solution tailored to your needs from a specialized partner – often within just a few days.

The key advantages at a glance:

⚡ Rapid implementation: From idea to ready-to-use application in days, not months. We deliver practical solutions that create immediate added value.

🔒 Maximum data security: Your sensitive data stays with you. We guarantee secure and compliant processing without sharing data with third parties.

💸 No financial risk: You only pay for results. High upfront investments in hardware, software, or personnel are completely eliminated.

🎯 Focus on your core business: Concentrate on what you do best. We take care of the entire technical implementation, operation, and maintenance of your AI solution.

📈 Future-proof & scalable: Your AI grows with you. We ensure continuous optimization and scalability, and flexibly adapt the models to new requirements.

More information here:

• Managed AI Platform

Regulation as a hidden growth engine for operational service providers

A key, often underestimated driver of demand for professionally managed AI services lies within the regulatory landscape itself. Banks and insurers have been subject to strict model risk management requirements for years, which regulators explicitly apply to generative AI and autonomous agents as well. The stock market regulator has announced cybersecurity audits for 2026 that will explicitly include AI controls. In healthcare, the principle is that any processing of protected patient data by an AI system is subject to the same data protection requirements as any other processing. In practice, this means that the use of cloud-based language models requires a formal agreement with the provider regarding the handling of this data. While there is no separate AI healthcare law, existing regulations effectively act as a barrier to entry, forcing many hospitals and insurers to outsource the operation of their AI systems to specialized, demonstrably compliant service providers, rather than managing it themselves.

Even at the federal level, a comprehensive AI law is still lacking. Instead, policymakers govern through implementing regulations and a legislative framework presented in March 2026, while in practice, the voluntary framework of the National Institute of Standards and Technology for the risk management of AI systems has established itself as the de facto standard. This is supplemented by the international standard ISO 42001, which companies increasingly present to their own customers as proof of proper governance. The European parallel is interesting: In the summer of 2026, the European Union, as part of a so-called "Digital Omnibus," postponed key high-risk obligations of its AI law to the end of 2027 and August 2028, respectively. This postponement provides short-term relief for American corporations with European operations, but hardly changes the fundamental need for governance services in the US domestic market, because separate, sometimes stricter, sectoral regulations already apply there.

The big names in the business world and their quiet power shift

A look at the latest financial figures of major consulting and IT service companies confirms a shift in the balance of power within the industry. At one of the largest global consulting firms, its business with ongoing services grew to over nine billion US dollars in the third fiscal quarter of 2026, significantly faster than its traditional, project-based consulting business. During the same period, one of the leading technology companies reported very high order intake related to AI, while its traditional consulting business stagnated – a pattern repeated by other competitors. In August 2026, a long-established technology company and a leading provider of language models announced a partnership aimed at jointly rolling out secure enterprise solutions; a sign that the line between model developers and operations service providers is increasingly blurring.

Indian IT service providers and American systems integrators are increasingly winning multi-billion-dollar contracts, particularly in the financial and healthcare sectors. The question of whether services are provided locally in the United States or via offshore centers in India plays a significant and, to date, insufficiently documented role for sensitive, regulated industries. A major transaction involving an IT service provider and a large American health insurer, revealed in July 2026 and valued at over US$500 million, exemplifies how closely traditional IT outsourcing and the new operation of AI systems have become intertwined.

The elephant in the room: Do most AI projects really fail?

Few figures have dominated the debate about the economic benefits of artificial intelligence over the past year as much as the claim that 95 percent of all generative AI pilot projects failed to generate a measurable return on investment. This figure, originally from a study associated with the Massachusetts Institute of Technology, has been repeated in countless media reports and sales presentations without any independent verification of the underlying methodology, sample size, or precise definition of zero return. Critics point out that numerous vendor and consulting studies simultaneously report cost savings of 25 to 40 percent in selected, well-managed projects, which stands in stark contrast to the supposed near-universal failure of the technology. A far more mundane explanation is more likely than a blanket failure of the technology: The difference between an impressive demonstration project and a genuinely profitable, sustainably operated system lies not in the model itself, but in the lack of an underlying operational model. This gap between demo and production is the very commercial breeding ground on which the entire managed AI industry thrives.

Between outsourcing and in-house development: The silent trench warfare of corporations

Not every company is following the trend toward outsourcing. While surveys of executives paint a picture in which the vast majority of American companies view external operating partners as important or even crucial for scaling autonomous AI agents, a counter-narrative is emerging. Large, financially strong brands are increasingly building their own custom AI systems instead of off-the-shelf software, driven by the conviction that strategically critical capabilities should not be delegated to third parties. The available data does not yet allow for a reliable quantification of the actual proportion of internally operated systems compared to outsourced operations, forcing the public debate into a rather unhelpful black-and-white logic, even though the reality in most companies is likely a hybrid approach.

A related problem, increasingly discussed in expert circles, concerns the actual operating costs. Many companies have so far failed to systematically link the ongoing costs of computing power and model usage to the actual productivity gains, giving rise to a new field called AI FinOps, which is dedicated exclusively to the financial management of AI operating costs. Market observers now cite this issue as one of the core value propositions of professional operations service providers, alongside continuous quality assurance, embedded governance, and policy-driven automation.

Labor market: The real cause of the outsourcing wave

Behind the decision to outsource the operation of AI systems to external service providers often lies simply a lack of internal specialists. Labor market analyses from spring 2026 show that healthcare, one of the sectors with the greatest anticipated need for AI, is among the least prepared for this transformation. Conversely, company surveys from the third quarter of 2026 report that the actual weekly use of AI tools by employees is particularly high in knowledge-intensive sectors such as professional services and software companies, while it is significantly lower in healthcare, manufacturing, and education. This discrepancy between demand and expertise explains why the demand for operational services is increasingly shifting from pure model development to operational and architectural capabilities: The shortage is less about specialists who can train new models than about those who can keep existing systems running securely, economically, and in compliance with regulations.

What is publicly discussed and what actually matters

What's remarkable is the discrepancy between what receives attention on social media and in public debate and what is actually decided in companies' procurement processes. The public discussion focuses almost exclusively on data centers, electricity consumption, and the sheer scale of infrastructure investments by large platform companies, while the fundamental question of who is actually responsible for the smooth operation of the deployed systems in hospitals or bank branches receives little attention. At the same time, new offerings are emerging at the lower end of the market, often marketed by startups, for so-called managed AI employees. These employees are intended to relieve small and medium-sized enterprises, such as dental practices or local service providers, of simple tasks like appointment scheduling and customer support. How viable these business models actually are beyond mere marketing promises remains unclear, as reliable revenue and retention figures are lacking.

Also striking is a growing skepticism toward non-American models and offshore operations centers when it comes to confidential corporate data. This preference for domestic, transparent operational structures is likely to become a key differentiator in regulated industries in the medium term for providers who can credibly demonstrate that they process sensitive data exclusively within the United States.

A bet on control rather than creativity

The economic story of artificial intelligence in the United States is currently being told in two chapters. The first, which has garnered significant public attention, concerns increasingly powerful models and gigantic infrastructure investments. The second, potentially more economically consequential, chapter concerns who will operate these models responsibly, securely, and economically in everyday life once they have left the experimental stage. Available data suggests that the real value creation is increasingly shifting to this second, far less glamorous area, driven by a shortage of skilled workers, regulatory pressure, and the simple realization that an impressive prototype is far from a profitable business model. For providers or companies seeking to succeed in this market, the question is not so much which model is the most intelligent, but rather which organization can keep it running reliably, compliantly, and cost-effectively. This, and not the size of the next language model, is likely to ultimately determine who truly profits in this new economy.

Despite the dynamic nature of the market, caution is still advised. The available market figures are primarily derived from commercially driven studies by large consulting and analysis firms, whose definitions of managed AI vary considerably and rarely disclose what proportion is actually attributable to pure operations as opposed to software, infrastructure, or one-off consulting services. To date, no reliable, methodologically consistent, and independently verified ranking of the industries with the greatest demand for managed artificial intelligence exists. Anyone making sound investment decisions based on such rankings should be aware of this methodological imprecision and critically examine the underlying assumptions of each individual study before treating them as established facts.

Konrad Wolfenstein

I would be happy to serve as your personal advisor.

You can contact me at wolfenstein∂xpert.digital or

Just call me on +49 7348 4088 965 .

LinkedIn