There’s a universal truth I’ve learned during my time in the corporate world: NO ONE wants to admit their operations are held together by workarounds and prayers. It makes sense. When you build a process over years, you grow attached to it. So when someone shows up with a new tool or tech to ‘optimize’ your workflow, all you hear is:

“Your baby is ugly.”

That’s, in a nutshell, the uphill battle I fought many times driving digital transformation at companies like Merck, GE HealthCare, or Johnson Controls.

Now, the good news is that over the past year or so, that defense mechanism has started to crumble under the rise of AI. As leaders are rushing to agentify operations, they’ve come to realize that without a solid process foundation, AI is doomed to fail.

(If you want to put a number on it: 76% of 1,600+ global business leaders we surveyed said broken or unoptimized processes hold their AI initiatives back.)

Acknowledging the problem is only the first step though. So for anyone in the thick of a process/AI/digital transformation, I wanted to share some learnings: what I’d repeat, and what I’d never do again.

One quick disclaimer: To make it concrete, I’ll mostly talk about how I scaled Celonis at Merck (read about my biggest Do’s and Don’ts here). One, because it was one of my most successful career gigs: we went from being a “science project” to a core pillar of the AI strategy that’s still thriving after I left ($75M in value unlocked so far). And two, because I’m convinced that the operational context Celonis gives you is the key to make AI worth your investment (so much so that I work for them now). Let’s dig in.

1. Reframe the blame

In a company as big as Merck, the biggest barrier to change usually is corporate politics.

Every executive wants to protect their kingdom. When you walk into a room with a process mining tool that maps reality, showing every bottleneck and why it’s happening, people easily get protective. You CANNOT approach these conversations with blame, or you’ll get kicked out of the room real fast. Here’s the narrative that worked for me:

"Look, your team is running as fast as they can. But they're up against our legacy systems. That’s not your fault. But there are things we can improve pretty quickly so your people stop doing grunt work.”

That’s the way to turn skeptics into active co-conspirators.

2. Stop doing things the way you’ve always done them

I’ll be honest: After so many years in the trenches, I was convinced that the bigger the problem, the more time, people, and money it should take to fix it. When I was asked to help figure out why we were shipping products but didn’t bill them, I did what I’d always do. I built a project plan that said it would take three different teams six months and a few hundred thousand dollars to fix it. In my mind, that was the “cost of doing business.”

But once I discovered Celonis (and really understood what the platform was capable of), we realized we could do it faster, better, and for a fraction of the cost. In less than a month, a single person on my team (not 3 teams!) built the app end-to-end. Nice extra: We didn’t just find the unbilled shipments; we built the logic to ensure they never slipped through again.

This was the start of our journey to unlock $20M across Order-to-Cash and Source-to-Pay — and tens millions more, enterprise-wide, after that. My proudest moment was when I started getting people from all corners of the business coming to me on a daily basis asking me to solve their problems with Celonis because they'd heard such great things about it.

3. Go where it ‘hurts’ first

Starting in the organization I was in (GBS), seemed like the logical first step. We plugged Celonis into our Source-to-Pay initiative — a project with C-suite attention that had an ambitious 40% cost-takeout target.

Now, don’t get me wrong, I’m pretty proud of what we achieved there. But given what I know today, I’d say:

Take a look at the messiest, most customized corners of your core business first.

Why? Because GBS cost reductions don't really change how the business competes. If you want to scale, go straight to the operations where your business actually wins or loses in the market.

For us, that was the clinical trial study setup process. Merck had a massive pipeline of blockbuster drugs coming down the line. To handle that influx, our clinical trials needed to run faster.

By plugging Celonis directly into ClinOps’ custom tools, we mapped the process, uncovered where hand-offs would stall, and saw how we could save millions of dollars right off the bat.

Finding that kind of value in what is just a tiny fraction of the entire ClinOps life cycle proved to the wider C-Suite that we were barely scratching the surface with Celonis — and gave my team much broader scope going forward.

4. Do POVs. Many. Fast.

When a department head claimed their workflow was perfect, I’d initially tried to ‘pitch’ my way into their heart. Don’t recommend it, it’s just a waste of your time and theirs. Instead, just tell them to give you their flat files. By loading flat files into Celonis, you can show their real process bottlenecks. Seeing is believing, and it usually makes even the biggest skeptics switch to your side.

My team got so skilled in building these POVs that we could churn them out within one to two days. Doing these fast will help you to keep momentum going once you get your first big wins under your belt. (These days, the Celonis Platform also has agents that build these POVs even faster than we did.)

5. Get a value architect

Now this point is not just specific to Celonis: when you’re scaling a transformation program across complex domains like clinical ops or the pharmaceutical supply chain, external consultants can only take you so far. Partners can cover data engineering all day, but they lack the internal authority to break into new territory. You need someone who knows the culture, can talk to the C-Suite, and knows where the ‘process bodies’ are buried.

The single best decision I made was to bring a Global Process Owner from the business into our CoE. She knew the stakeholders, she understood the issues of our custom systems, and she had the personal credibility to knock on doors and earn trust.

The best proof that what we were doing was working? My entire team of 30 people shifted their focus to being Celonis-first before we’d inject any RPA bots or AI agents.

6. Move into the AI office

If your transformation remains buried inside a single function like GBS or Finance, your impact will eventually fizzle out. I wasted more time than I’d like to admit fighting people’s initial reactions: “You’re in Finance, you don’t understand our processes. Go away.”

The ultimate game-changer for our team was moving our CoE directly into a centralized, enterprise-wide AI Office.

It took me seven years to get to $27 million in value. Once we moved under the AI Office, it took the team just one year to leap to $75 million in value.

I’ll let that sit with you.

My ultimate advice to everyone in DX/AI/process transformation

I could go on and on about what worked for me and what didn’t. But the baseline is this: Be bold, run your proofs of value fast, find internal champions and, above all: Stop calling a broken process beautiful.

Thanks for reading. :)