From shared services to AI strategy - how Merck built its process intelligence foundation

When Steve Carpenter first encountered Celonis at SAP Sapphire in 2014, the process mining platform was still in its early days - before it moved to the cloud, when it was "just process mining." But that initial glimpse into the technology's capabilities stayed with him. Three years later, when he joined Merck as Executive Director of GBS Digital Services, he discovered the pharmaceutical company was already planning to implement Celonis. Carpenter recalls:

I wasn't part of the decision, but I immediately got on board and said I'd love to work with this capability.

The timing was perfect. Merck had just established a new Global Digital Services organization and was facing a fundamental challenge: how to drive 5% productivity improvements year over year without understanding how work actually got done. Carpenter explains:

We wanted to understand, rationalize, and harmonize processes, and we realized we couldn't do that without a tool that would give us visibility into how things actually work. That's really why we went down the Celonis path in the beginning, in 2018.

The visibility problem

Merck's challenge wasn't unique, but its scale made it particularly complex. As a global pharmaceutical company, the organization had accumulated layers of customizations, workarounds, and process variations across different markets and business units. Traditional enterprise software provided transactional data but lacked the process context that revealed why bottlenecks occurred and where improvements would have the biggest impact.

The initial focus was deliberately narrow. Carpenter's team started with SAP-based processes, looking at procure-to-pay, order management, and supply chain activities. This provided a foundation for understanding how process intelligence could work within Merck's existing technology landscape while demonstrating concrete value to skeptical stakeholders. Carpenter says:

Initially, it was very much focused on SAP. But ultimately, we broadened our scope. We got access to the EMS [Enterprise Management System] and started going after subject areas and process areas that had nothing to do with SAP.

The platform's real potential became clear once Merck moved from an on-premise to cloud deployment. This unlocked access to Action Flows and other capabilities that enabled the organization to not just understand processes but actually do something about them. Carpenter cites a "ship not build" problem that would have taken three different teams four to six months and significant budget to solve using traditional approaches”

We had somebody from Celonis come in, and in four weeks, we had an end-to-end solution based on just using the Celonis capability.

Moving beyond the IT department

For the first several years, Carpenter's shared services organization operated as a center of excellence for Celonis. But as the technology proved its value, leadership pushed him to expand beyond finance and GBS:

I was knocking on doors with my suitcase full of Celonis, asking, 'Hey, do you want to buy some Celonis?' across the organization—manufacturing, commercial, clinical operations.

This grassroots approach yielded some success, with Carpenter's team proving value and securing investment for new use cases. But he quickly realized that in Merck's organizational structure, IT was better positioned to drive enterprise-wide adoption. The breakthrough came when he moved the entire Celonis center of excellence into the Chief AI Office on April 1st. Carpenter explains:

We want to make sure we're using process intelligence to inform our AI strategy. If someone comes in and says, 'I want to build some AI around this process,' great—let's look at your process, understand it, figure out where AI can make a difference, and then measure it afterward.

This positioning proved crucial for gaining organizational buy-in. While business leaders often viewed process analysis as complicated and time-consuming, they were eager to invest in AI capabilities. By framing process intelligence as essential infrastructure for successful AI implementation, Carpenter could secure both attention and resources from senior executives:

People want AI. It's at the top echelons of our company - we're driving an AI strategy. People tend to think that understanding process is hard, so if we can pull this into the AI program and make sure they understand that we need to understand process before we do AI and make it successful, that's what's been driving home for everybody here.

From diagnosis to prescription

Merck's use of Celonis has evolved from traditional process mining focused on individual workflows to object-centric process mining that reveals how different processes interact and impact each other. This shift enables the company to create comprehensive digital twins of its operations rather than isolated views of specific activities. Carpenter explains:

We started [object-centric process mining] about three months after it was announced at Celosphere - we became a beta customer. We still have a ton of classic models, so we're trying to figure out an incremental and opportunistic way to move to OCPM [Object-Centric Process Mining] where appropriate. Anything new is using OCPM.

The platform now supports both operational improvements and strategic AI initiatives. One significant but confidential use case involves content optimization, where Merck is implementing digital twins to understand complex approval workflows involving medical and legal reviews. 

In clinical operations, the company is using process intelligence to understand study setup procedures and supply chain coordination for clinical trials. Carpenter says:  

We're looking at how we get supply chain materials into clinical trials, making sure materials arrive on time because they need to be patient-specific, dose-specific, with specific handling of its chain. 

The potential impact is significant:

Days and weeks could save millions and millions, and also get our medicines out to people and patients more quickly and help save lives. 

Proving value and driving adoption

Merck has implemented a rigorous approach to demonstrating process intelligence value that has been crucial for maintaining support and funding. Every use case is evaluated for potential return, with business units receiving analysis and optimization support in exchange for documented evidence of realized benefits.

The results speak for themselves. Carpenter's team has documented substantial savings from process improvements, with expectations of significant dollar savings as the program scales. But perhaps more importantly, the approach has begun changing how business units think about operations. Carpenter explains: 

We've started to incorporate more business stakeholders into the COE in a federated fashion, so we can leverage their expertise and understanding of the process to help identify where they have bottlenecks, waste, rework loops, and things they need to improve. 

This shift from centralized analysis to distributed ownership reflects a broader maturation in how Merck approaches process intelligence. Rather than having a separate team analyze processes and recommend changes, business units are increasingly using digital twins directly to identify improvement opportunities.

The success has been particularly notable in clinical operations, where a digital twin implementation identified significant efficiency opportunities within three months:

There's a 580-person team that does clinical trial study setup—creating the database used at a clinical trial. They found specific bottlenecks and issues, and they changed their process immediately. No technology at all—just changing process based on what they see in the digital twin.

The culture challenge

Despite the technical success, Carpenter identifies culture change as the most significant ongoing challenge. Many managers remain reluctant to pursue process improvements even when the potential value is clearly documented:

You can tell the managers that you have a super tool in place, but to change processes afterwards…

Carpenter recalls a conversation with a senior leader who initially resisted pursuing millions in identified procurement savings:

He told me, 'I have to do so many things to change that. The people, I have to drive initiative. I would have to sell two or three more of our products to achieve the value.'

Learning from these experiences, Carpenter adopted a bottom-up approach focusing on operational staff rather than management:

Find the people on the shop floor, not the managers, because they are the ones that have to work with the things that you're trying to achieve.

This strategy has proven effective. After two years, business unit leaders began independently presenting Celonis success stories to the executive board—a key indicator of genuine adoption rather than mandated compliance.

Looking ahead

With approximately 10% of Merck's processes currently analyzed through Celonis, Carpenter sees substantial opportunity for expansion. The company is exploring partnerships with distributors and other external organizations, though data security and access controls remain important considerations. He says: 

I know there are a couple of big US-based distributors we use who are also Celonis clients. At some point, I think it would make sense for us to exchange data outside our four walls, obviously making sure that's secure and protected.

For now, however, the focus remains on internal transformation. The integration with the Chief AI Office is yielding new use cases that combine process intelligence with artificial intelligence capabilities, moving beyond operational efficiency toward more strategic applications.

Carpenter's advice for other organizations starting their process intelligence journey emphasizes the importance of quick wins and executive support:

Go after those quick wins—that'll help capture the imagination and hearts and minds of the rest of the organization. The other thing is executive buy in - that's 1,000,000% critical. You've got to have people who can see the forest for the trees.

The Merck experience illustrates how process intelligence platforms can evolve from operational tools to strategic enablers of digital transformation. By grounding AI initiatives in detailed understanding of how work actually gets done, the company has created a foundation for scaling artificial intelligence across the enterprise while maintaining focus on measurable business outcomes.

As Carpenter puts it:

 It's hard because people think, 'I own the process. My process isn't bad - it's good. I've built this thing over time. So don't come in and question my process.'

But, it’s clear that with the right approach, technology, and organizational positioning, that resistance can be overcome.