Going beyond pilots with composable and sovereign AI | MIT Technology Review

You need to enable JavaScript to view this site.

Skip to ContentMIT Technology ReviewFeaturedTopicsNewslettersEventsAudioMIT Technology ReviewFeaturedTopicsNewslettersEventsAudioSponsoredArtificial intelligenceGoing beyond pilots with composable and sovereign AIAI scaling is hindered by fragmented enterprise infrastructure in a constantly shifting technology ecosystem. A new architectural paradigm of composable, sovereign AI can help enterprises move past pilot purgatory.
By MIT Technology Review Insightsarchive pageJanuary 19, 2026In partnership withUniphore Today marks an inflection point for enterprise AI adoption. Despite billions invested in generative AI, only 5% of integrated pilots deliver measurable business value and nearly one in two companies abandons AI initiatives before reaching production. DOWNLOAD THE ARTICLE The bottleneck is not the models themselves. What’s holding enterprises back is the surrounding infrastructure: Limited data accessibility, rigid integration, and fragile deployment pathways prevent AI initiatives from scaling beyond early LLM and RAG experiments. In response, enterprises are moving toward composable and sovereign AI architectures that lower costs, preserve data ownership, and adapt to the rapid, unpredictable evolution of AI—a shift IDC expects 75% of global businesses to make by 2027. The concept to production reality AI pilots almost always work, and that’s the problem. Proofs of concept (PoCs) are meant to validate feasibility, surface use cases, and build confidence for larger investments. But they thrive in conditions that rarely resemble the realities of production. Source: Compiled by MIT Technology Review Insights with data from Informatica, CDO Insights 2025 report, 2026 “PoCs live inside a safe bubble” observes Cristopher Kuehl, chief data officer at Continent 8 Technologies. Data is carefully curated, integrations are few, and the work is often handled by the most senior and motivated teams.
The result, according to Gerry Murray, research director at IDC, is not so much pilot failure as structural mis-design: Many AI initiatives are effectively “set up for failure from the start.” Download the article. by MIT Technology Review InsightsShareShare story on linkedinShare story on facebookShare story on emailPopularWe’re learning more about what vitamin D does to our bodiesJessica Hamzelou10 Breakthrough Technologies 2026Amy NordrumThe great AI hype correction of 2025Will Douglas HeavenChina figured out how to sell EVs. Now it has to deal with their aging batteries.Caiwei ChenDeep DiveArtificial intelligenceThe great AI hype correction of 2025Four ways to think about this year's reckoning.
By Will Douglas Heavenarchive pageWhat’s next for AI in 2026Our AI writers make their big bets for the coming year—here are five hot trends to watch.
By Rhiannon Williamsarchive pageWill Douglas Heavenarchive pageCaiwei Chenarchive pageJames O'Donnellarchive pageMichelle Kimarchive pageMeet the new biologists treating LLMs like aliensBy studying large language models as if they were living things instead of computer programs, scientists are discovering some of their secrets for the first time.
By Will Douglas Heavenarchive pageAn AI model trained on prison phone calls now looks for planned crimes in those callsThe model is built to detect when crimes are being “contemplated.”
By James O'Donnellarchive pageStay connectedIllustration by Rose WongGet the latest updates fromMIT Technology ReviewDiscover special offers, top stories,
upcoming events, and more.Enter your emailPrivacy PolicyThank you for submitting your email!Explore more newslettersIt looks like something went wrong.
We’re having trouble saving your preferences.
Try refreshing this page and updating them one
more time. If you continue to get this message,
reach out to us at
customer-service@technologyreview.com with a list of newsletters you’d like to receive.The latest iteration of a legacyFounded at the Massachusetts Institute of Technology in 1899, MIT Technology Review is a world-renowned, independent media company whose insight, analysis, reviews, interviews and live events explain the newest technologies and their commercial, social and political impact.READ ABOUT OUR HISTORYAdvertise with MIT Technology ReviewElevate your brand to the forefront of conversation around emerging technologies that are radically transforming business. From event sponsorships to custom content to visually arresting video storytelling, advertising with MIT Technology Review creates opportunities for your brand to resonate with an unmatched audience of technology and business elite.ADVERTISE WITH US© 2026 MIT Technology ReviewAboutAbout usCareersCustom contentAdvertise with usInternational EditionsRepublishingMIT Alumni NewsHelpHelp & FAQMy subscriptionEditorial guidelinesPrivacy policyTerms of ServiceWrite for usContact uslinkedin opens in a new windowinstagram opens in a new windowreddit opens in a new windowfacebook opens in a new windowrss opens in a new window

The article discusses the challenges enterprises face in scaling AI initiatives beyond initial pilot projects, highlighting a critical gap between proof of concept (PoC) success and operational impact. Despite significant investments in generative AI, only 5% of integrated pilots achieve measurable business value, while nearly half of companies abandon AI projects before reaching production. This stagnation is attributed to fragmented enterprise infrastructure, which struggles to adapt to the dynamic nature of AI technologies. The authors argue that the root issue lies not in the models themselves but in the surrounding systems—limited data accessibility, rigid integration frameworks, and fragile deployment pathways—that hinder scalability. To address this, they advocate for a shift toward composable and sovereign AI architectures, which promise to reduce costs, preserve data ownership, and enhance adaptability in an evolving technological landscape. This transition, anticipated by 75% of global businesses by 2027 according to IDC, represents a structural reimagining of how enterprises design and implement AI solutions.

The article critiques the inherent limitations of traditional AI pilots, which often operate in controlled environments that do not reflect real-world complexities. PoCs are typically designed to validate feasibility and build confidence for larger investments, but they thrive in conditions that are artificially insulated from production realities. As Cristopher Kuehl, chief data officer at Continent 8 Technologies, notes, these pilots function within a “safe bubble” where data is meticulously curated, integrations are minimal, and work is handled by highly motivated teams. This setup creates an illusion of success that fails to account for the messy, interconnected challenges of scaling AI across enterprises. Gerry Murray, research director at IDC, emphasizes that many AI initiatives are “set up for failure from the start” due to structural misdesign, rather than technical shortcomings. The article underscores that the problem is not the models or algorithms but the lack of flexible, resilient infrastructure capable of supporting AI’s complexity and adaptability.

A key argument in the piece is that composable AI offers a framework for modular, interoperable systems that can dynamically assemble and reconfigure components to meet evolving business needs. This approach contrasts with traditional monolithic architectures, which are inflexible and costly to modify. Sovereign AI, on the other hand, emphasizes data ownership and control, ensuring that enterprises retain autonomy over their datasets while leveraging external tools and models. Together, these paradigms aim to address the scalability bottleneck by enabling seamless integration of AI technologies with existing systems, reducing dependency on rigid pipelines, and minimizing the risks associated with deploying untested solutions. The authors suggest that this shift is not merely a technical upgrade but a cultural and strategic imperative, requiring organizations to rethink how they conceptualize AI as a core driver of innovation rather than a peripheral experiment.

The article also highlights the growing urgency for enterprises to adopt these new architectures, given the accelerating pace of AI development and its increasing integration into business operations. The authors reference IDC’s forecast that 75% of global businesses will transition to composable and sovereign AI by 2027, reflecting a broader industry recognition of the need for more agile and sustainable solutions. This transition is framed as a response to the limitations of current AI strategies, which often prioritize short-term gains over long-term viability. By prioritizing modularity and data sovereignty, enterprises can mitigate the risks of obsolescence and ensure that their AI systems remain aligned with evolving technological and market demands. The piece also touches on the importance of collaboration between AI developers, data scientists, and business leaders to align technical capabilities with strategic objectives, emphasizing that successful scaling requires more than just technological innovation—it demands a holistic reevaluation of organizational processes and priorities.

Another critical point raised in the article is the need to address the disconnect between AI research and enterprise implementation. While advancements in large language models (LLMs) and retrieval-augmented generation (RAG) have generated considerable excitement, their practical application in business settings remains constrained by infrastructure and operational challenges. The authors argue that the focus should shift from showcasing AI’s potential to demonstrating its tangible value through scalable, real-world deployments. This requires investing in tools and frameworks that facilitate seamless integration with existing workflows, as well as fostering a culture of experimentation and continuous improvement. The article suggests that enterprises must move beyond isolated pilots and instead build ecosystems where AI can be iteratively refined, tested, and deployed across departments and functions. This approach not only enhances the likelihood of success but also ensures that AI initiatives contribute meaningfully to organizational goals.

The piece also touches on the broader implications of AI scalability for industries and economies, noting that the failure to overcome these challenges could stifle innovation and limit the transformative potential of AI. By adopting composable and sovereign architectures, enterprises can create more resilient systems that adapt to changing conditions without requiring complete overhauls. This flexibility is particularly important in sectors where AI applications must navigate complex regulatory environments, such as healthcare, finance, and logistics. The authors stress that data sovereignty is a critical component of this strategy, as it allows organizations to maintain control over their information while still benefiting from external AI capabilities. This balance between autonomy and collaboration is seen as essential for fostering trust, ensuring compliance, and maximizing the value of AI investments.

In addition to technical considerations, the article highlights the importance of organizational readiness in achieving successful AI scaling. Many enterprises lack the internal expertise, governance structures, and cultural alignment necessary to support large-scale AI initiatives. The authors recommend that organizations invest in upskilling their workforce, establishing clear AI governance policies, and fostering cross-functional collaboration to bridge the gap between technical teams and business units. They also emphasize the need for leadership buy-in, as top-down support is crucial for overcoming resistance to change and ensuring that AI strategies are integrated into broader business objectives. By addressing these human and organizational factors, enterprises can create the conditions necessary for AI to thrive beyond the pilot phase.

The article concludes by framing composable and sovereign AI as a necessary evolution in the enterprise AI landscape, one that addresses both technical and strategic challenges. It calls for a shift away from fragmented, siloed approaches toward more integrated, adaptable systems that can support the growing complexity of AI applications. The authors suggest that this transition will require ongoing investment in infrastructure, talent, and innovation, but the long-term benefits—such as increased efficiency, reduced costs, and enhanced competitive advantage—are deemed well worth the effort. By embracing these new paradigms, enterprises can move beyond the limitations of current AI strategies and unlock the full potential of artificial intelligence as a transformative force in business and society.