How to scale Celonis (and any other technology): 10 do’s and don’ts from a transformation leader

Done right, scaling Celonis (or any other technology) from pilot to enterprise-wide is a complete career gamechanger. But, at least in my experience, the way there can be a political minefield.

During my years of driving digital transformation at global enterprises like Merck, GE HealthCare, or Johnson Controls, I’ve made my fair share of mistakes. But I also learned valuable lessons in what it takes to scale technologies in a way that will get even die-hard skeptics on board.

So here are my 10 do’s and don'ts — some uniquely about Celonis, drawing from my experience of scaling the platform back at Merck. And some broader lessons on Enterprise AI, from my current role advising enterprise and DX leaders from inside Celonis.

Don'ts

DON'T start with reporting first.
Let’s start with a Celonis lens. The single biggest mistake you can make is to underestimate what the platform is capable of. It's easy to build simple visibility dashboards, hand them over to teams, and celebrate. But if you use Celonis as a ‘fancier’ Power BI or Tableau report, you’re missing out on real P&L value. The platform has evolved far beyond just analyzing processes and seeing why things break in real time. It gives your tech stack (and yes, notably your AI agents) the operational context and intelligence to fix things before they break. Learn more about why the Celonis Context Model is the missing layer to make AI work here.

My lesson at Merck:
There’s a learning curve in understanding what Celonis is capable of. At Merck, my team and I spent a good 12 months trying to drag some legacy teams out of a reporting-first paradigm (Another spoiler: There’s absolutely zero ROI in pure reporting).

DON'T push IT away.
Don’t isolate Celonis to the business side. Given that Agentic AI is bulldozing into every company’s enterprise landscape, and that AI absolutely needs operational context (which you’ll get through Celonis extracting data from your entire system landscape, unifying, sequencing, and enriching it), IT is your best friend.

My lesson at Merck:
Having someone from IT (data engineers or data analysts) in your Center of Excellence should be non-negotiable to connect your data sources, build process data models, and keep your data pipelines clean. On top of that, you’ll want a governance trio for every function in which you want to run Celonis (or build AI use cases). Meaning a process lead, a functional lead, and an IT lead who can take ownership of Celonis within that process area and collaborate closely with your CoE. Across all the customers I speak with, this is what really makes the difference in getting stakeholders aligned and holding people accountable.

DON'T rely on a pure grassroots approach.
Grassroots movements are great for getting folks down in the weeds excited, but let’s be real: they rarely have the leverage, the authority, or the budgets to move the corporate needle.

My lesson at Merck:
Way back when we shifted 2,000 finance people from local markets into a global GBS model, Celonis definitely helped us grow organically by pointing out what wasn’t working yet. But we only truly scaled when we got the attention from the C-Suite, most notably our CFO and CIO. If I were to start all over with Celonis, I’d say this: Don’t just pitch to the business; but try to get Enterprise Architects and the CTO on your side from day one. Pairing a strong business sponsor (like the CFO) with the CIO is what boosted Merck’s total value realized from $20M to $75M in just one year.

DON'T accept functional silos.
When I started, my role was based out of Finance. The minute I tried to knock on doors in manufacturing or clinical operations, I’d get: “Go away, you’re from finance. You don’t understand our processes.” You have to break out of the functional cage early. Process is process is process — no matter in which department it runs.

My lesson at Merck:
Change the conversation. Make them understand: Process is process is process. Celonis doesn't care if it's reading SOPs tables in SAP, a Databricks data stream, or a custom internal workflow. I upskilled my 30-people team to run flat-file POVs. We didn’t wait around for IT; we told leaders, "Just hand us your flat files." Within days, we could show exactly what their real workflows looked like. Once people saw their bottlenecks laid out like this, we knew we’d won them over.

DON'T assume your training stops at “Go-Live.”
Technology innovations move faster than organizations can digest them. If you aren't continuously refreshing your team's skills on structural paradigm shifts — like moving from case-centric to object-centric mining — your CoE will soon become obsolete.

Do's

DO start with a business-critical use case.
Skip the generic visibility (for Process Mining) or pure automation (for AI) use cases. Instead, target a high-impact problem that keeps your C-Suite awake at night. If you solve a problem tied to strategic enterprise objectives, funding and resources will never be an issue again.

My lesson at Merck:
For us, that breakthrough moment happened in clinical operations. Merck had a huge pipeline of new blockbuster drugs, which meant our ClinOps team suddenly had to handle hundreds of upcoming clinical studies. We plugged Celonis directly into their custom tools to analyze the study setup process – where trial sites would track protocols, patient deviations, etc. By creating a digital twin of that process, we pinpointed where hand-offs didn’t work or people simply weren’t 100% efficient. We optimized how teams focused their time, and uncovered millions of dollars in hidden value right off the bat. “Finding millions in just a fraction of the ClinOps life cycle proved to the wider C-Suite that we were barely scratching the surface with Celonis — and gave the team much broader scope going forward.”

DO poach an internal ‘Value Architect.’
You can easily outsource data engineering and raw analytics. Partners can handle that all day long. What you can’t outsource is internal authority. You need someone who knows the culture, can talk to the C-Suite, and knows where the ‘process bodies’ are buried.

My lesson at Merck:
When we built Merck’s Procure-to-Pay model, the Global Process Owner (GPO) for Order-to-Cash was so blown away by what we mapped that I convinced her to come work for me. She became my first internal Value Architect. Having her on my CoE team was my secret weapon to knock down resistance, simply because she was a process expert (and six-sigma certified).

DO capitalize on “opportunistic funding.”
Keep your eyes on annual productivity targets and budgets to use them to your advantage.

My lesson at Merck:
At Merck, we found ourselves with a million-dollar budget surplus we had to spend by the end of the year or lose. We seized that window, brought in Accenture, and mapped our procure-to-pay lifecycle — from vendor creation to invoice settlement. And we used the leftover cash to up/re-skill all 30+ people on my team over the next 18 months to drive the platform the way it was meant to be driven.

Do as many POVs as you can.
When a department head claims their workflow is perfect, don’t argue. Just tell them, “Give me your flat files.” By loading flat files into Celonis, you can show their real process bottlenecks within days. Seeing is believing, and it usually makes even the biggest skeptics switch to your side.

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

DO fight for the “Pull” model.
Remember how I said top-down is the way to go? That’s true especially as you’re starting out, but your ultimate goal as a CoE leader is to hit the tipping point where transformation stops being a grueling “push.”

My lesson at Merck:
We hit that inflection point after we proved to our CFO that we could strip 40% out of invoice processing. Suddenly, the energy shifted. I didn’t have to hunt down executives anymore; they were lining up outside my door asking for help.