McKinsey: Scaling AI Beats Fragmented Business Pilots

While AI dominates corporate strategy, a vast gap exists between pilot schemes and enterprise value. 

A report from McKinsey, which surveyed 1,000 senior and midlevel executives across 696 manufacturing and service-sector businesses, shows that almost 90% of organisations say they are at least experimenting with AI. However, just 7% report scaling it across the enterprise.

McKinsey says the pattern suggests that the impact of AI comes not from experimentation alone, but from integration into core operational processes. 

Rahul Shahani, McKinsey Partner and leader of the firm’s Manufacturing and Supply Chain Practice in North America, led the research. 

He says: “One of the clearest findings from our research is that manufacturers don’t realise the full value of AI simply by deploying the technology into operations that are sub-optimal.”

Isolated AI limits business performance

McKinsey’s report shows a performance advantage for companies which moved beyond isolated AI use and scaled across the enterprise. 

It found that, even with widespread disruption, reported by almost all of its respondents, companies with AI embedded across multiple functions generate nearly double the profit margins of peers using AI in only a few departments.

Among the companies which embedded AI across multiple functions, the difference was even more pronounced in capital returns. The report found that the three-year return on invested capital is more than five times higher for these firms.

McKinsey discovered that parts of advanced manufacturing are more consistent in deploying AI across functions. This consistency reflects years of investment in data, analytics and execution discipline. 

Most responses to the survey were from large organisations, with only 20% from companies with less than US$1bn in revenue.

Manufacturing productivity gains rely on operational excellence

The report highlighted that companies which embed AI across more functions report steadily higher productivity gains. Organisations limiting AI to a small set of use cases see far more modest results.

Leading companies increasingly move beyond experiments to embed AI in core workflows and link use cases directly to operational and financial outcomes. 

Advanced companies see significantly higher productivity and profitability than their peers, but operational excellence matters too. This relies on robust management and technical systems, a strong corporate purpose and well-defined operating principles and behaviours. 

McKinsey says companies with the best results were combining technology with these core operational elements. 

Companies which have built advanced technology into their operational excellence achieve higher productivity increases than companies relying mainly on manual or analogue systems.

The power of digital twin integration

Throughout its report, McKinsey analyses key examples of how manufacturers were embedding AI. 

One of the most relevant examples to its argument about operational excellence and AI integration is the Siemens facility in Nanjing, China. The site is a World Economic Forum Global Lighthouse Factory.

Rahul says: “The biggest gains come when companies pair these tools with strong management systems, clear operating principles and disciplined execution. Siemens’ Nanjing facility, part of the World Economic Forum’s Global Lighthouse Network, is a good example. 

“By combining digital twin capabilities with broader operational improvements, the company was able to significantly increase throughput. For manufacturers, the lesson is that technology matters, but the operating model around it matters just as much.”

High product variability and small batch sizes were putting constant pressure on throughput and delivery reliability at the Nanjing site. While the leadership team explored the use of digital twins, it resisted scaling them prematurely, McKinsey’s report says.

The site responded by tightening its operating backbone before expanding the technology. It integrated a manufacturing operations management system which governed data flows between virtual models and physical assets.

Teams at the Siemens site then validated simulations through structured routines before implementing changes. Clear decision rights were defined when human confirmation was required and leaders treated IT/OT integration as well as data standards as core operational disciplines.