The vision of AI in manufacturing is seductive: “lights out” factories that are so heavily automated that they almost run themselves, with human workers monitoring operations from an off-site control center. Indeed, a few of the most advanced robotics factories have already passed a crucial line, with robots building robots.
That’s the future that so many COOs desire, according to our global survey of more than 100 COOs at manufacturers with at least $1 billion in revenues (see sidebar, “Our methodology”). Companies are raising their bets on digital and AI technologies.
Over the past five years, one-third of respondents say that their companies spent less than 1 percent of the cost of goods sold on digital and AI. But when asked about their plans for the next five years, only 7 percent of respondents plan to keep their investment levels that low. The other 93 percent will spend more, with almost one-third intending to spend at least 5 percent.
This commitment is visible through much of the manufacturing world. Among the latest applications that factories have submitted to join the Global Lighthouse Network (GLN) of technologically advanced manufacturing sites,1 fully 90 percent of tech use cases now incorporate AI.
What’s far less clear is whether that money will be well spent. COOs’ survey responses reflect broader uncertainty about the time and expense that may be required for AI to reach the scale where it can generate real value. About two-thirds of respondents indicate that their companies’ AI implementation is still at the exploration or targeted-implementation stage (Exhibit 1). A mere 2 percent say that AI is now fully embedded across all operations.
So far, COOs’ priorities look much as one would expect. The use cases rising to the top of the list—such as predictive maintenance, schedule optimization, and process improvement—are familiar ones that manufactures have targeted for decades.
What’s striking, however, isn’t where COOs are investing the most but where they are investing the least. At the bottom of COOs’ collective priority list are several fundamental people and tech elements that are critical to get right for AI to meet its promise of generating sustained productivity improvement.
That suggests that many COOs may be overlooking serious risks to their investment plans. And, given how much more they are likely to spend, the omissions could have serious long-term consequences.
Short-term progress, but at what long-term cost?
At first glance, COOs’ AI priorities look like the right ones. When asked where they expect digital and AI to deliver the greatest impact, COOs emphasize the fundamentals: expanding production capacity, boosting labor productivity, improving quality, and increasing end-to-end visibility. To minimize wasted effort on pilots, their road maps call for scaling a focused portfolio of five to 12 use cases by 2030. Target use cases, such as factory scheduling, digital performance management, and digital twins, are consistent with spending plans tilted toward shop floor automation, robotics, and the systems that control factory operations.
These choices can appear especially attractive when they continue investments that many manufacturers have made for years, as in robotics. Others are particularly important for reaching scale quickly, as with AI-informed scenario planning and data-driven decision support.
Yet there’s a flaw. As powerful as these investments can be, companies may be ignoring some of the basic levers and enablers that allow an organization not only to build useful AI-based tools but to deploy and keep improving them as the business evolves. On average, respondents give the lowest prioritization to workforce enablement, IT/OT (IT and operational technology) infrastructure, and cybersecurity (Exhibit 2). These foundations help determine whether AI deployments can scale safely and sustainably. Without robust infrastructure, skilled and empowered employees, and strong cyber protections, even the most advanced automation or optimization tools risk stalling at pilot stage or exposing the enterprise to new vulnerabilities.
Paying attention to these factors is paying off for companies such as Chilean miner SQM, one of the world’s largest producers of lithium. Its AI models are enabling frontline employees to make fine-grained, real-time production decisions that optimize output while minimizing water and energy use. SQM leaders point to the company’s training investments as critical to this success. The continuing education enables workers to use the technologies effectively—and keep improving with them.
The stumbling blocks COOs can already see
COOs are already aware of some of the critical problems that they face. When we asked respondents to name the biggest challenges in implementing AI in operations, two of the three relate to people. Fully half of respondents cite the need to shift their culture as a major impediment, and almost as many point to reskilling needs (Exhibit 3).
Legacy processes compound the issue. Too many manufacturing workflows remain rigid and optimized for past technologies, making it hard to reinvent operations around AI-first ways of working.
On the tech side, the foundations are far from solid: 46 percent of surveyed COOs report limitations in their data or IT/OT systems, with outdated infrastructure (19 percent) and poor data quality (18 percent) further slowing progress. Even when use cases are proven, a quarter of companies struggle to build applications that are reusable and scalable.
A Manufacturing Leadership Council survey of its membership has found a similarly mixed picture on governance, with the majority of respondents saying that their organization lacks AI-specific KPIs.2 Yet where such targets are in place, nearly two-thirds of companies meet or exceed them—suggesting that robust governance is one of the most powerful differentiators in realizing AI’s potential (Exhibit 4).
How manufacturers can get AI on track
To turn investment into impact, COOs will need to own the AI value agenda as part of the broader COO productivity mandate. That means treating AI not as a series of experiments but as a rewired performance engine—anchored in KPI-tied targets, backed by ring-fenced funding, and tracked through a regular cadence of value reviews.
Companies that hold themselves accountable for outcomes are far more likely than peers to see their AI systems progress from pilot to profit. Rewiring the enterprise means taking three essential steps (Exhibit 5).
Redesign production
Redesigning production means reimagining business processes from start to finish to create a new operating model that takes full advantage of technology. The resulting implementation road map strictly prioritizes investments according to the lasting business value that they can create—not the wow factor of the underlying technology.
This approach to redesigning production enabled a consumer goods company to upgrade its sprawling network of legacy production sites, which had vast differences in scale, layout, tech infrastructure, and management culture. Despite the variability, the same basic issues kept cropping up across sites, such as rising production losses during product changeovers and high energy and water usage. Discovering these commonalities helped leaders reconceive of their factory network as an integrated production system. It also helped guide the company toward pragmatic solutions that emphasized mature technologies, such as improved deployment of sensors and digitalized standard operating procedures.
Build scalable technology
The second step is to build scalable technology on an IT backbone engineered for interoperability and application development that emphasizes reusable capabilities. A minimum viable architecture—based on common data products, open interfaces, and industrial-grade pipelines—can improve data availability and quality for enterprise-scale application deployment. Judicious use of third-party functional applications and a few highly bespoke advanced models can provide a good balance of cost efficiency and customization.
Achieving scale requires an in-depth understanding of the operation’s current data architecture and a willingness to make targeted investments. A large, complex production site for a global pharmaceutical company, for example, recognized that its legacy IT/OT systems were too siloed for the company to scale its AI investments. By establishing three integrated data platforms that connected about a dozen IT systems and more than 150 Internet of Things sensors, site leaders created a unified structure for flexible digital applications.
With that foundation in place, the site could then design for scale by codifying proven use cases into reusable capabilities—including multiple apps for the shop floor that now support live operational-efficiency tracking and schedule optimization, as well as augmented reality that can reduce delays in product line changeovers. Deploying multiple high-impact use cases in parallel accelerated the site’s transformation by increasing overall equipment effectiveness (OEE) by ten percentage points while halving unplanned downtime. The plant is now on target to more than double its production volume in less than three years.
Drive scale and adoption
Finally, success in scaling AI depends on reskilled people, a renewed culture, and a revamped operating model that supports adoption across the entire organization. At leading manufacturers, reenergizing the workforce starts by establishing an AI-optimized capability for process reinvention that partners with HR to deliver tailored training at scale. The pharmaceutical manufacturer provided coaching to more than 25 leaders and managers at its Lighthouse site and involved more than 100 frontline employees in agile sprints. These practical moves shifted the site’s culture to be more dynamic and responsive. As a result, the site saw labor productivity gains of more than 10 percent and filled about a dozen new digital and analytics roles—mostly with internal candidates.
Almost three-quarters of surveyed COOs say that they expect to pursue a hybrid build-buy-partner operating model that curates an ecosystem of tech partners and simultaneously build internal strengths, often through centers of excellence. At the consumer goods company, a center of excellence structure has helped the operations leadership align internal stakeholders more quickly and effectively to further accelerate homegrown AI know-how.
Some of the GLN’s advanced production sites are rapidly rebalancing their capabilities, with network members developing more and more AI solutions in-house. Nevertheless, even these companies retain vendor partnerships for cutting-edge solutions. HVAC (heating, ventilation, and air-conditioning) manufacturer Qingdao Hisense Hitachi Air-Conditioning Systems, for example, worked with university and automation partners to develop a highly precise machine-vision-based positioning system that reduced production cycle times by 22 percent and changeover times by two-thirds.
For manufacturers, turning AI investment into lasting competitive advantage takes more than the right automation technologies or production-monitoring tools. It means building new human and infrastructure capabilities as they completely rethink how a production network operates and what it can achieve.
