Enterprise AI ROI Stalls as Industrial Copilots Proliferate

The headline number that didn't move

Nearly ten years into its tracking of corporate AI use, McKinsey's latest global survey finds broader adoption but not much more bottom-line impact. In The state of AI in 2026, 37 percent of respondents attribute at least some EBIT impact to AI. That is roughly the same share as the year before. 3 The survey describes organizations scaling AI across the enterprise, pushing it into more business functions, and using a widening mix of tools, including chatbots and agentic coding assistants. 3

The report also points to a new source of friction: cost. About 20 percent of respondents say AI-related operating expenses, including token costs, have limited how much they use AI. 3 Most still plan to spend more. 3 McKinsey also notes a gap between the benefits individual workers get from their own AI use and what companies manage to capture at the organizational level. 3

In short, more AI is being deployed, but the share of firms reporting enterprise-level financial results has stayed flat.

Meanwhile, the copilots keep coming

The industrial software market is moving quickly anyway. Siemens introduced nine AI-powered industrial copilots in 2026, built into platforms such as Teamcenter, Polarion and Opcenter. These cover product lifecycle management, compliance and manufacturing operations. 4 Its collaboration with NVIDIA aims at an "Industrial AI Operating System" spanning design, engineering, manufacturing and supply chains. 4

At Realize LIVE 2026, Siemens described agentic capabilities across Designcenter, Opcenter, Simcenter and Teamcenter. 1 In its operations software, for example, agents are meant to find the root causes of product and production problems and recommend fixes, including shop-floor reconfiguration. 1 The event also included a PepsiCo keynote on productivity gains, Supplyframe's $50 million investment in Xometry, and an integration with VolitionAI that brings component availability, risk and cost data into Designcenter. 1

Siemens is not the only vendor in this race. Microsoft, Schneider Electric, Honeywell, ABB and IBM are each competing with different mixes of automation expertise, enterprise software, cloud infrastructure and AI. 4 A vendor-authored ranking of ROI-focused platforms names SAP, Oracle, IFS, Microsoft and Infor, each with its own approach to embedding AI. 2

Where the sources agree and where they diverge

All four accounts agree that AI is moving from experiment to embedded workflow. They differ in how much they say about results. The two industry-side pieces describe a market in a "new deployment phase." 4 The Siemens coverage focuses on capability and customer stories rather than measured financial outcomes. 1 McKinsey's survey data is the main independent check, and it shows that broader deployment has not yet produced more widespread EBIT impact. 3

The IFS blog takes a middle position. It acknowledges that "ROI is uneven" and cites McKinsey figures: 88 percent of organizations using AI in at least one function, only about a third scaling it enterprise-wide, and 23 percent scaling agentic systems. 2 These figures appear to come from the prior edition of the survey rather than the August 2026 release, so they are best treated as a baseline. The blog also makes the economic case for industrial AI. It cites Siemens research estimating that unplanned downtime costs the 500 largest companies about $1.4 trillion a year, roughly 11 percent of revenue. It also cites Deloitte's finding that poor maintenance can reduce an asset's productive capacity by 5 to 20 percent. 2 The prize, it argues, is won or lost in the gap between piloting and scaling. 2

Readers should keep in mind who is speaking. IFS ranks itself among the top platforms. 2 The copilot market piece is a press release from a market research firm. 4 Both have reasons to frame the moment as an inflection point.

What it means

One reading is that buyers face a familiar pattern: adoption is easy to show, but measurable returns lag behind. Copilots are now built directly into the systems engineers and plant managers use every day. That should make AI easier to use, but convenience is not the same as cash flow. McKinsey's flat EBIT figure suggests most organizations have not yet redesigned processes enough to turn individual productivity into enterprise results. 3

Industrial use cases may be among the better candidates for closing that gap, because their targets can be measured. Downtime, maintenance losses and production defects have clear costs. 2 Agentic root-cause analysis on the shop floor is the kind of tightly scoped application where value can be traced directly. 1

Cost is likely to become the deciding factor. If one in five organizations is already rationing AI because of operating expenses 3, vendors adding more copilots will need to show where the savings come from, not just how many features they offer. The next survey will show whether the current wave of embedded agents raises that 37 percent or simply adds to the bill.