Why Manufacturing AI Investments Fail Without the Right Foundation

AI in Manufacturing: Big Budgets, Small Results

Manufacturing leaders face a paradox. They recognize that artificial intelligence represents the future of production, with visions of highly automated factories running with minimal human intervention. They are backing this conviction with substantial capital, as 93 percent of surveyed Chief Operating Officers plan to increase digital and AI spending over the next five years. Yet despite these ambitious investments, only 2 percent have successfully embedded AI across all operations.

This disconnect, revealed in McKinsey’s December 2025 survey of 101 manufacturing COOs at companies with over $1 billion in revenue, points to a fundamental misalignment between where manufacturers are spending and where they need to invest for AI to deliver lasting value. The data suggests that many organizations are making a critical error: prioritizing flashy automation technologies while neglecting the unglamorous but essential foundations that determine whether digital transformations succeed or stall.

The Investment Surge and the Scaling Problem

The manufacturing sector’s commitment to AI is undeniable and accelerating. Over the past five years, one-third of surveyed companies spent less than 1 percent of cost of goods sold (COGS) on digital and AI initiatives. Looking forward, only 7 percent plan to maintain such modest investment levels. The remaining 93 percent intend to spend more, with nearly one-third committing at least 5 percent of COGS to these technologies.

This spending surge aligns with broader industry trends. According to the World Economic Forum’s Global Lighthouse Network, 90 percent of recent manufacturing technology applications now incorporate AI. The vision driving these investments is seductive: lights-out factories where robots build robots, with human workers monitoring operations remotely. Some advanced robotics facilities have already crossed this threshold.

Yet the survey reveals a sobering reality check. About two-thirds of respondents indicate their companies remain stuck at the exploration or targeted-implementation stage. Only one-third have achieved network-wide integration, and a mere 2 percent report full operational embedding. This suggests that most manufacturers are struggling to move from successful pilots to scalable, value-generating implementations.

Research from MIT’s Initiative on the Digital Economy supports this finding, indicating that only 10 to 15 percent of AI pilots in manufacturing successfully scale to production. The primary barriers are rarely technological. Instead, they involve organizational capacity, data readiness, and change management capabilities.

The Troubling Gap in Investment Priorities

What makes the McKinsey survey particularly illuminating is not where COOs are investing most heavily, but where they are investing least. The top priorities follow predictable patterns: shop floor automation and robotics lead the list, followed by AI-driven process optimization, factory operations and control systems, and simulation and scenario planning.

These choices emphasize visible, tangible technologies that promise immediate operational improvements. They continue multi-year investment trajectories in areas like robotics where many manufacturers have established expertise. They target familiar use cases that have been optimization goals for decades, including predictive maintenance, schedule optimization, and process improvement.

The problem lies at the bottom of the priority list. The three lowest-ranked investment categories are workforce enablement and augmentation, data and IT/OT infrastructure, and cybersecurity. These foundations determine whether AI deployments can scale safely and sustainably across an enterprise. Without robust infrastructure, organizations cannot reliably collect, process, and act on the data that AI systems require. Without skilled and empowered employees, even the most sophisticated tools remain underutilized. Without strong cyber protections, connected factories become vulnerable attack surfaces.

This prioritization creates a dangerous mismatch. COOs are essentially building increasingly sophisticated digital structures on unstable foundations. The approach might generate impressive pilot results and compelling proof-of-concept demonstrations, but it undermines the ability to achieve the network-wide scale that justifies major capital commitments.

The Human Capital Paradox

The workforce dimension reveals a particularly striking contradiction. When asked to identify the biggest challenges in implementing AI, half of surveyed COOs cite the need for cultural shift as a major impediment. Nearly as many point to reskilling needs, making these the two most frequently mentioned barriers alongside business process reimagination.

COOs clearly recognize that people issues represent critical obstacles. Yet workforce enablement ranks dead last in their investment priorities. This disconnect suggests a significant blind spot in strategic planning. Leaders acknowledge the problem in principle but fail to allocate resources proportionate to the challenge.

Academic research on technology adoption provides context for why this matters. A comprehensive study published in the California Management Review analyzed hundreds of digital transformation initiatives and found that projects with substantial upfront investment in workforce capability building were 2.6 times more likely to achieve their performance targets than those emphasizing technology deployment first.

The explanation lies in how organizational learning actually occurs. Workers cannot simply be handed new AI-powered tools and expected to use them effectively. They need structured opportunities to understand how these systems work, experiment with their applications, and develop judgment about when to trust algorithmic recommendations versus when to apply human expertise. This learning process requires time, coaching, and psychological safety to make mistakes.

SQM, the Chilean lithium mining company highlighted in the McKinsey article, demonstrates this principle. The organization’s AI models enable frontline employees to make fine-grained, real-time production decisions optimizing output while minimizing resource consumption. Company leaders explicitly credit training investments as critical to this success, enabling workers to use technologies effectively and continuously improve their application.

The pharmaceutical manufacturer example provides additional evidence. That organization provided coaching to more than 25 leaders and managers while involving over 100 frontline employees in agile development sprints. These investments shifted organizational culture toward greater dynamism and responsiveness. The results included labor productivity gains exceeding 10 percent and successful internal recruitment for about a dozen new digital and analytics roles.

Contrast these success stories with the survey finding that workforce enablement receives the lowest investment priority. The implication is clear: many manufacturers are setting themselves up for expensive failures by deploying sophisticated technologies into organizations unprepared to use them effectively.

The Infrastructure Challenge

The technical foundations present similar concerns. Forty-six percent of surveyed COOs report limitations in their data or IT/OT systems. Nineteen percent identify outdated infrastructure as a specific problem, while 18 percent cite poor data quality. Even when use cases are proven effective, a quarter of companies struggle to build reusable and scalable applications.

These infrastructure weaknesses create multiple problems. Poor data quality undermines AI model accuracy, leading to unreliable predictions and eroding user trust. Outdated systems lack the computational capacity and connectivity required for real-time analytics. Siloed architectures prevent the integration necessary for enterprise-wide deployment. The inability to build reusable applications means that each new implementation requires starting from scratch, eliminating economies of scale.

A Gartner study on manufacturing IT infrastructure found that companies with modern, integrated data architectures achieved AI scaling rates four times higher than those with legacy systems. The research emphasized that infrastructure investments deliver returns through enabling multiple use cases rather than supporting any single application.

The pharmaceutical manufacturer case illustrates this dynamic. Site leaders recognized that legacy IT/OT systems were too siloed to scale AI investments. By establishing three integrated data platforms connecting about a dozen IT systems and more than 150 Internet of Things sensors, they created unified infrastructure for flexible digital applications. This foundation enabled parallel deployment of multiple high-impact use cases, accelerating transformation by increasing overall equipment effectiveness by ten percentage points while halving unplanned downtime.

Yet infrastructure ranks near the bottom of COOs’ investment priorities. This suggests that many organizations are attempting to scale AI applications across fragmented, incompatible systems. The approach might work for isolated pilots but becomes increasingly untenable as deployments expand. Each new site or production line requires custom integration work, dramatically increasing implementation costs and timelines.

The Governance Deficit

The McKinsey survey reveals another critical gap: governance frameworks. A Manufacturing Leadership Council survey found that the majority of manufacturing organizations lack AI-specific key performance indicators. Where such targets exist, however, nearly two-thirds of companies meet or exceed them. This suggests that robust governance represents one of the most powerful differentiators in realizing AI’s potential.

Effective governance provides several critical functions. It establishes clear accountability for outcomes rather than just activities or outputs. It creates feedback loops enabling rapid course correction when implementations veer off track. It ensures resource allocation aligns with strategic priorities rather than getting captured by competing internal interests. It provides transparency that builds stakeholder confidence and organizational buy-in.

Research from the Harvard Business School examining enterprise technology investments found that governance quality predicted project success better than technology choice, budget size, or leadership support. Projects with well-defined KPIs, regular value reviews, and clear accountability structures delivered returns averaging 40 percent higher than comparable initiatives lacking these elements.

Yet fewer than half of surveyed manufacturers have established such frameworks for their AI investments. This creates risk that substantial capital commitments lack the discipline and rigor required to generate commensurate returns. Without clear metrics, organizations cannot distinguish successful deployments from failures until significant resources have been consumed. Without accountability structures, problems get obscured or rationalized rather than addressed.

Three Steps Toward Sustainable Scaling

The McKinsey article outlines a three-step framework for turning AI investment into lasting impact. Each step addresses dimensions that current investment priorities underemphasize.

First, redesign production by reimagining business processes end-to-end rather than simply automating existing workflows. This means creating new operating models that take full advantage of technology capabilities. Implementation road maps should prioritize investments according to lasting business value rather than technological sophistication.

The consumer goods company example demonstrates this approach. Despite vast differences in scale, layout, infrastructure, and culture across legacy production sites, leaders discovered common issues including rising changeover losses and high resource consumption. Recognizing these patterns enabled reconceiving the factory network as an integrated production system. This insight guided pragmatic solutions emphasizing mature technologies like improved sensor deployment and digitalized standard operating procedures rather than chasing cutting-edge applications.

Second, build scalable technology on IT backbones engineered for interoperability. Minimum viable architectures based on common data products, open interfaces, and industrial-grade pipelines improve data availability and quality for enterprise-scale deployment. Judicious use of third-party applications balanced with selective bespoke advanced models provides cost efficiency alongside customization.

This requires deep understanding of current data architecture and willingness to make targeted infrastructure investments before scaling applications. The pharmaceutical manufacturer established integrated platforms connecting disparate systems and sensors, creating unified structures for flexible digital applications. With this foundation, the site could codify proven use cases into reusable capabilities including shop floor apps for operational efficiency tracking, schedule optimization, and augmented reality reducing changeover delays.

Third, drive scale and adoption through reskilled people, renewed culture, and revamped operating models supporting organization-wide implementation. At leading manufacturers, workforce reenergization starts by establishing AI-optimized capabilities for process reinvention, partnering with HR to deliver tailored training at scale.

The pharmaceutical manufacturer provided coaching to over 25 leaders and managers while involving more than 100 frontline employees in agile sprints. These practical moves shifted culture toward greater dynamism and responsiveness, generating labor productivity gains exceeding 10 percent while filling about a dozen new digital roles mostly with internal candidates.

The Build-Buy-Partner Balance

Almost three-quarters of surveyed COOs expect to pursue hybrid build-buy-partner operating models that curate technology partner ecosystems while building internal strengths, often through centers of excellence. This approach recognizes that no organization can develop all required capabilities internally.

Global Lighthouse Network members are rapidly rebalancing capabilities, developing increasing shares of AI solutions in-house while retaining vendor partnerships for cutting-edge applications. HVAC manufacturer Qingdao Hisense Hitachi worked with university and automation partners to develop highly precise machine-vision-based positioning systems that reduced production cycle times by 22 percent and changeover times by two-thirds.

This balanced approach requires clear strategic thinking about which capabilities provide competitive differentiation versus which can be effectively outsourced. It demands governance processes ensuring that partner relationships transfer knowledge rather than creating dependencies. It needs talent strategies developing internal expertise while accessing specialized external skills.

Yet executing this balance effectively depends on the foundational investments many COOs are underweighting. Without skilled internal teams, organizations cannot effectively evaluate vendor claims, integrate third-party solutions, or learn from partner collaborations. Without robust infrastructure, external applications cannot connect to internal systems. Without strong governance, partner ecosystems become fragmented rather than coherent.

Learning from Digital Transformation Failures

The patterns visible in this survey echo broader research on digital transformation failures across industries. A comprehensive PwC study analyzing over 1,000 digital initiatives found that 70 percent failed to achieve their objectives. The primary causes were remarkably consistent: inadequate change management, insufficient skills development, poor data quality, legacy infrastructure constraints, and weak governance.

These findings align closely with the gaps identified in manufacturing AI investments. The implication is that manufacturers risk repeating well-documented mistakes from other sectors’ digital transformation efforts. The substantial capital commitments planned over coming years could generate disappointing returns if foundational weaknesses remain unaddressed.

Some high-profile manufacturing failures illustrate these risks. Boeing’s production challenges with the 787 Dreamliner partly stemmed from implementing sophisticated manufacturing technologies across a supplier network lacking the integration, skills, and processes to use them effectively. General Electric’s ambitious Industrial Internet initiatives struggled when legacy infrastructure and organizational resistance impeded scaling beyond pilot sites.

These cautionary examples share common threads: impressive technology deployments undermined by insufficient attention to enablers. In each case, leaders recognized problems eventually but only after substantial value destruction. The lesson is clear: identifying foundational gaps early and investing to address them prevents far costlier problems later.

Recommendations for Manufacturing Leaders

The survey findings suggest several concrete steps for COOs seeking to maximize returns on AI investments:

• Conduct honest readiness assessments across people, process, technology, and governance dimensions to establish an accurate maturity baseline.
• Rebalance investment portfolios to address foundational weaknesses before scaling applications broadly (infrastructure, workforce development, and governance).
• Establish clear KPIs and value reviews that emphasize business outcomes, including leading indicators (adoption, data quality) and lagging measures (productivity, uptime).
• Pair workforce investments with deployments by co-designing training alongside application development to build skills in context.
• Treat infrastructure as a strategic enabler by modernizing data architectures to unlock multiple use cases and compound returns over time.
• Learn from proven practitioners through lighthouse networks, industry associations, and academic partnerships to reduce trial-and-error costs.

Conclusion

The manufacturing sector stands at a critical juncture. Artificial intelligence offers genuine opportunities to transform production economics through unprecedented automation, optimization, and responsiveness. The technology has matured sufficiently that successful applications are proven across multiple industries and use cases. Manufacturing leaders recognize this potential and are committing substantial capital to capture it.

Yet the McKinsey survey reveals a troubling pattern. While investment in AI technologies accelerates dramatically, many organizations are underinvesting in the foundational capabilities that determine whether sophisticated tools translate into sustainable competitive advantage. Workforce enablement, infrastructure modernization, and governance frameworks receive insufficient priority despite clear evidence that they represent critical success factors.

This misalignment creates significant risk. Billions of dollars could be spent deploying AI applications that fail to scale beyond impressive pilots. Organizations might achieve isolated successes while struggling to capture enterprise-wide value. Competitors investing more strategically in foundations could leapfrog ahead despite smaller technology budgets.

The path forward requires discipline and patience. Manufacturing leaders must resist the temptation to chase every promising technology while neglecting less glamorous but essential enablers. They must balance the desire for rapid competitive positioning with the reality that sustainable transformation requires organizational readiness. They must recognize that the constraint on AI value creation is rarely the technology itself but rather the people, processes, and infrastructure that determine how effectively organizations deploy and improve it.

For COOs willing to make this shift, the opportunity remains substantial. The companies that get the balance right will achieve the vision of highly automated, AI-optimized production that so many leaders desire. Those that continue prioritizing technology over foundations risk learning expensive lessons that others have already paid to understand. The choice belongs to manufacturing leaders, and the survey data suggests that many need to reconsider where they place their bets.