With Gartner predicting that by 2028 an average global Fortune 500 enterprise will have over 150,000 agents in use, it’s clear that agents have become a massive engine for productivity. Their proliferation in the enterprise means that the calculus around measuring their efficacy changes. AI value is less about how productive agents are and more about how much they can transform the business.

How should AI value be measured?

AI value is generally measured in terms of productivity. McKinsey’s 2026 survey of 1,719 executives found that 80% say AI has improved their productivity. Productivity isn’t the entire picture, however.

“You might have a mechanical engineer on a plant floor who’s spending more time in spreadsheets than diagnosing equipment,” says Harry Murphy, Vice President of Presales for Applications, Data and AI (Americas) at Siemens. “You want the people running these processes to actually be doing the job they’re paid to do.” In other words, automate the spreadsheet and the engineer gets hours back.

But speeding up one step doesn’t mean that the entire process changes. That’s where transformational value comes in.

A CIO can report that everything runs well and still have no answer when the CEO asks what changed. And the CEO is asking. Gartner’s 2026 survey of 469 CEOs and senior executives found that 80% expect AI to force a high or medium degree of change in their operational capabilities.

Presently, the response to CEOs is bleak: Only 6% of the executives in McKinsey’s survey can point to significant impact agents have on earnings.

How do you create transformational value with AI agents?

Creating transformational value with agents isn’t a question of quantity. It’s a question of the quality of agents. Producing higher quality agents starts with a foundation. Murphy lives in a house built in 1844, with a fieldstone foundation, the kind where shifting stones create cracks in the plaster. “It would have been very satisfying to just plaster over those cracks,” he says. “But the main problem remains the foundation.”

In organizations, those “cracks” manifest in business processes in different ways:

• Handoffs that no one can trace
• Legacy systems that can’t integrate easily
• Manual workarounds that hide broken workflows

To Murphy, agents built in silos or designed merely to automate an existing step in a process are the plaster that organizations use to fix the ostensible problem. Yet the foundation is what needs to be addressed.

A sturdy foundation for agents requires two things: underlying contextualized data that accurately reflects business reality, and an orchestration layer that lets humans and agents interact with it. Only when agents are grounded in that reality can they move past surface-level productivity and orchestrate operational change.

Productivity vs. transformation: Why a chatbot wasn’t enough

Productivity ROI is about doing tasks faster; transformational value is about eliminating the broken processes behind them. Take a global microscope manufacturer: when a piece of lab equipment breaks in the field, the solution to replace it sits across four systems:

1. CRM records
2. Product catalogs
3. Service manuals
4. Materials data

Field engineers hopped between all four to resolve the ticket.

The obvious fix was a retrieval-augmented generation (RAG) chatbot that looked up answers in the service manuals. It was fast, but field engineers were still getting incomplete answers when they would ask which parts fit certain microscopes. The engineers stopped trusting the chatbot’s results. The optics manufacturer essentially plastered over the crack.

“If you’re looking for a product to implement that operational change, I think you are barking up the wrong tree,” says Murphy.

The company rebuilt the foundation instead with a knowledge graph — built with Graph Studio — that spanned all four sources, with a shared ontology that encoded relationships that retrieval had missed.

Graph Studio provides the deep contextual foundation. But a foundation on its own doesn’t help the engineer in the field. “I can’t live in a foundation,” as Murphy puts it. So the company built an app for them using Mendix, an agentic development platform.

Mendix turns that context into action. Organizations use it to coordinate work end-to-end, not just workflows inside a single application but also agents and systems across what the company already owns. Orchestration, integration, governance and agentic development sit within a single integrated development environment.

Together, as part of Siemens’ Intelligence Center X, an agentic enterprise system, Graph Studio and Mendix bridge the gap between knowledge and execution.

The result: The manufacturer didn’t just get a slightly faster way to look up service manuals (productivity ROI). They fundamentally changed how field engineers operate by unifying their data foundation and orchestrating it through an app (transformational value).

Transformational AI value comes from a solid foundation

If your agents are siloed automation — whether it’s 5, 15, or 150,000 — those productivity gains are likely to slip. As Murphy says of faulty foundations, “The agent you build this quarter could be obsolete the next.” The walls are just going to keep cracking, no matter how much plaster you spackle onto them.

True transformational value doesn’t come from more agents, it comes from a foundation of clean, contextualized data that represents the business and an orchestration layer that conducts it. That is how you turn AI from productivity patches into the operation-changing impact your CEO is looking for.

Read more about process orchestration with Mendix.