Imagine you're running a company and everyone keeps telling you that artificial intelligence is the future. You need AI to stay competitive, but there's a catch: building AI systems from scratch requires millions of dollars, rare technical talent, and years of trial and error. What if there was another way?

This is the dilemma facing executives across every industry today. And according to groundbreaking research from universities in Germany, the Netherlands, and the United Kingdom, many companies are solving it by essentially renting AI capabilities from tech giants like Google, Microsoft, and IBM rather than building everything themselves.

The study, which tracked nearly 400 major American corporations over a decade, reveals a surprising pattern about who adopts these "AI as a service" solutions, why they do it, and whether it actually pays off. The findings challenge some conventional wisdom about competitive advantage in the age of artificial intelligence.

The Build Versus Rent Dilemma

Think of AI boundary resources as the digital equivalent of hiring a specialized contractor instead of training your own team. These are cloud-based services that companies can plug into their operations through application programming interfaces, or APIs. Need a chatbot for customer service? There's an API for that. Want to analyze customer behavior patterns? Another API handles it. Looking to automate quality control in manufacturing? Yet another service exists.

The appeal is obvious. Instead of spending years assembling a team of data scientists, building computing infrastructure, and training AI models on your own data, you can access sophisticated AI capabilities almost instantly. It's like having a world class expert on call whenever you need one.

But here's where it gets interesting. Conventional business strategy suggests that core capabilities should be developed in-house to maintain competitive advantage. If AI is truly transformational, shouldn't companies be building their own systems rather than relying on the same services their competitors can access?

The research team wanted to understand this puzzle. They analyzed data from S&P 500 companies between 2010 and 2019, manually tracking which firms adopted AI services, what they used them for, and how it affected their performance. What they discovered paints a nuanced picture that defies simple answers.

Two Paths, Two Purposes

The researchers found that companies use AI services in fundamentally different ways, and these differences matter enormously.

The first type they called AI boundary resources for process improvements. These are services used internally to make operations more efficient. Think of a logistics company using AI to optimize delivery routes, or a manufacturer using computer vision to detect defects on assembly lines. The AI works behind the scenes, helping the company run better but not directly touching customers.

The second type they labeled AI boundary resources for product improvements. These become part of what customers actually experience. When you talk to a voice assistant built into a smart thermostat, or when a shopping app recommends products based on AI analysis of your preferences, you're interacting with this type of AI service.

This distinction turns out to be crucial because different kinds of companies are drawn to each type for very different reasons.

The Knowledge Paradox

Here's something counterintuitive: companies that already have strong AI expertise are more likely to adopt AI services for improving their internal processes.

You might expect the opposite. Why would a company with significant AI knowledge need to rent capabilities from others? Shouldn't they build everything themselves?

The explanation lies in how AI development actually works. The research showed that companies developing their own AI patents, a solid indicator of internal AI capabilities, were 16% more likely to adopt external AI services for process improvements. Having a Chief Information Officer in the top management team increased adoption likelihood by a striking 75%.

The reason is integration complexity. Using AI services for process improvements requires deep technical knowledge because you need to connect external AI capabilities into your existing workflows. You need to prepare the right data inputs, integrate the AI's outputs into your decision processes, and understand when the AI's predictions can be trusted.

Companies with strong internal AI teams can evaluate which external services genuinely complement their capabilities and which are redundant. They can avoid the pitfalls that might trap less sophisticated adopters. In essence, knowledge enables them to be smart shoppers in the AI marketplace.

This challenges a common assumption that external AI services are mainly for companies playing catch up. Instead, the data suggests they're valuable complements even for AI leaders, allowing them to expand their capabilities faster than building everything from scratch.

External Pressure and Customer-Facing AI

The story flips when it comes to AI services that touch customers directly. Here, internal capabilities matter less than external competitive pressure.

The research identified two key pressure points. First, when digital startups enter an industry, established companies feel the heat. Digital natives like Uber, Airbnb, or direct-to-consumer brands are built around data and AI from day one. They scale rapidly and set new customer expectations.

When the proportion of digital ventures in an industry increased by one standard deviation, companies became 29% more likely to adopt AI services for product improvements. These services offer a quick way to match the AI-powered experiences that digital disruptors provide.

Second, when industry peers become more sophisticated in their AI use, companies feel compelled to keep up. The researchers measured this by analyzing how frequently companies mentioned AI concepts in their official business descriptions. When peer AI sophistication increased by one standard deviation, adoption of AI services for products jumped 52%.

This pattern makes sense. Imagine you're a hotel chain and your competitors start offering AI-powered personalization, dynamic pricing, and chatbots that handle customer service in multiple languages. You need to respond quickly. Building these capabilities in-house could take years, but plugging in external AI services might take months.

Interestingly, these external pressures don't significantly influence adoption of AI for internal processes. When competition heats up, companies focus on what customers see, not on back-office efficiency gains.

Does It Actually Work?

The critical question for any business decision is whether it improves performance. After all, renting AI capabilities instead of building them could create dangerous dependencies on powerful technology providers.

The research team tracked three types of outcomes: operational efficiency measured by operating costs per employee, revenue growth measured by sales per employee, and stock market performance.

For AI services used to improve internal processes, the results were positive across the board. Companies saw their operating costs decrease by about 7% over the following two years. The AI automation and better decision making delivered real efficiency gains. Perhaps more tellingly, stock prices increased by roughly 2 percentage points, suggesting investors valued these efficiency improvements.

For AI services used in products, the picture was more mixed. Companies saw sales increase by about 7% over two years, indicating that customers appreciated the AI-enhanced offerings. However, stock prices didn't budge significantly.

Why would investors shrug at sales growth? The researchers suggest several concerns. First, using external AI at the customer interface creates strategic dependency. If customers love a feature powered by Google's AI, but Google raises prices or changes terms, the company is vulnerable. Second, if many competitors adopt the same AI services, the advantage disappears. Third, the ease of adoption means benefits may be temporary.

The Hidden Costs of Convenience

The research reveals a subtle but important trade-off. AI services that are easiest to adopt may be least defensible as competitive advantages.

AI services for product improvements typically have what researchers call "low resource interdependence." They're designed as relatively self-contained modules that can be plugged into products without extensive integration work. A voice assistant API, for example, handles speech recognition, natural language understanding, and response generation as a complete package.

This convenience is precisely what makes them attractive when competitive pressure mounts. But it's also what makes them available to everyone. There's little that's unique or hard to replicate about using the same voice assistant API that dozens of other companies access.

By contrast, AI services for process improvements have "moderate resource interdependence." They require significant work to integrate into existing workflows, which demands internal AI expertise. This integration complexity creates a barrier that makes the capability harder for competitors to replicate, even though the underlying AI service is available to anyone.

Think of it like cooking. Using a pre-made sauce is easy and anyone can do it, but the result tastes similar regardless of who makes the dish. Developing your own sauce recipe requires culinary knowledge and experimentation, but the result can be distinctive. In the middle ground, taking a high-quality base ingredient and adapting it with your own techniques requires skill but can produce something special faster than starting from scratch.

The CIO's Strategic Role

One of the study's most practical findings concerns organizational leadership. Companies with a Chief Information Officer in their top management team were dramatically more likely to adopt AI services for internal process improvements but not for customer-facing products.

This pattern reveals how CIOs think strategically about AI adoption. With their technical expertise and strategic responsibilities, CIOs recognize opportunities to capture efficiency gains while avoiding potential pitfalls.

CIOs are particularly attuned to risks. Using external AI services involves data security concerns, potential vendor lock-in, and operational dependencies. These risks are more manageable for internal processes than for customer-facing applications where vendor problems directly impact customer experience.

Moreover, CIOs understand that AI services for improving internal processes can deliver substantial value without creating concerning dependencies in strategically critical areas. Optimizing warehouse logistics with external AI is different from having your customer relationship managed by someone else's AI.

This suggests that companies serious about strategic AI adoption should ensure technology leadership has a voice in top-level decision making. The presence or absence of a CIO in the executive suite shapes how companies navigate the build versus rent decision.

Lessons for the AI Era

This research arrives at a crucial moment. Artificial intelligence is transitioning from a specialized technology to a general-purpose capability that every company needs to master. The question is no longer whether to adopt AI but how.

The findings suggest several principles for navigating this landscape. First, external AI capabilities and internal AI development are complements, not substitutes. Companies with strong internal AI capabilities can use external services more effectively, creating combinations that are more valuable than either alone.

Second, context matters enormously. The right strategy for customer-facing applications differs from the right strategy for internal operations. There's no one-size-fits-all answer to the build versus rent question.

Third, competitive dynamics shape adoption patterns. When digital disruptors enter your industry or when peers advance their AI sophistication, the pressure to adopt external AI services for customer-facing applications intensifies. Recognizing these patterns can help companies respond strategically rather than reactively.

Fourth, immediate performance gains don't always translate to long-term strategic advantage. Services that boost sales may not impress investors concerned about sustainability and competitive positioning.

The Broader Transformation

The rise of AI as a service represents a fundamental shift in how business capabilities are sourced and developed. It's part of a larger trend toward modular, cloud-based infrastructure where companies assemble capabilities from multiple providers rather than building everything internally.

This shift has profound implications for competitive strategy. When key capabilities become available as services, traditional sources of competitive advantage may erode. The challenge becomes not just accessing AI capabilities but combining and integrating them in distinctive ways.

It also has implications for the AI industry itself. The research documented 415 distinct AI services from 239 providers available through just one API directory. This ecosystem enables innovation by lowering barriers to AI adoption, but it also creates concentration of power among major AI providers.

Policymakers and regulators are starting to grapple with these dynamics. Questions about data ownership, algorithmic transparency, and market power in AI services will become increasingly important as more companies depend on external AI providers.

What This Means For You

If you're a business leader, this research offers concrete guidance. Assess your AI strategy not as a binary choice between building and buying but as a portfolio decision. Where do you need distinctive capabilities that provide competitive advantage? Where can external services deliver good enough results faster and cheaper?

Invest in internal AI literacy even if you plan to use external services extensively. The research shows that AI knowledge makes companies better consumers of AI services. You need the expertise to evaluate providers, integrate services effectively, and avoid costly mistakes.

Pay attention to competitive dynamics in your industry. The entry of digital natives and the AI sophistication of peers are reliable signals that you may need to accelerate AI adoption in customer-facing applications.

Consider the composition of your leadership team. Do you have executives with the technical expertise and strategic vision to navigate AI decisions? The presence of technology leadership at the executive level correlates with more strategic AI adoption.

If you're an AI service provider, understand that different customer segments have different needs. Companies with strong internal capabilities may be your most sophisticated customers, looking for services they can integrate into complex workflows. Companies under competitive pressure may prioritize speed and ease of use for customer-facing applications.

The Future of AI Strategy

As AI continues its rapid evolution, the landscape of available services will expand dramatically. Large language models like GPT-4, computer vision systems, robotics control systems, and countless specialized AI capabilities are becoming available as services.

This proliferation creates new opportunities but also new complexities. How do you choose among dozens of similar services? How do you avoid creating a tangled web of dependencies on multiple AI providers? How do you maintain strategic coherence when critical capabilities are sourced externally?

The companies that thrive will likely be those that develop sophisticated capabilities in AI orchestration. Rather than mastering AI development from scratch, they'll master the art of selecting, combining, and adapting external AI services to create distinctive value.

This requires a new type of organizational capability that bridges technical understanding, strategic thinking, and operational excellence. It's not just about knowing AI exists; it's about knowing how to wield it effectively in specific business contexts.

A Balanced Approach

The research ultimately points toward a balanced approach to AI adoption. Neither pure internal development nor pure reliance on external services appears optimal. The most successful companies will likely be those that thoughtfully combine both, using internal capabilities to identify opportunities, evaluate options, and integrate external services in distinctive ways.

This balance will shift over time and vary by context. As AI services become more sophisticated and easier to use, the pendulum may swing toward greater external reliance. But as AI becomes more central to competitive advantage, companies may pull more capabilities back in-house.

What's clear is that AI adoption is not a one-time decision but an ongoing strategic process. The companies that understand this, that build the organizational capabilities to navigate these decisions thoughtfully, will be best positioned for success in an AI-driven future.

The question isn't whether to embrace AI. Everyone must. The question is how to do it strategically, balancing the speed and capability advantages of external services against the control and competitive advantage of internal development. This research provides valuable guidance for navigating those tradeoffs.

Publication Details

Published: 2025
Journal: European Journal of Information Systems
Publisher: Taylor & Francis Group
DOI: https://doi.org/10.1080/0960085X.2025.2473952

Credit and Disclaimer

This article is based on original research published in the European Journal of Information Systems by researchers from the University of Kassel (Germany), University of Groningen (Netherlands), and University of Mainz (Germany). The content has been adapted for a general audience while preserving scientific accuracy. For complete methodological details, comprehensive statistical analyses, full datasets, and in-depth theoretical frameworks, readers are strongly encouraged to consult the original peer-reviewed research article through the DOI link provided above. All scientific findings, data interpretations, and conclusions presented here are derived directly from the original publication, and full credit belongs to the research team and their institutions.