This piece was penned by Celonis President Carsten Thoma as part of The Importance of AI Roadmaps, Celonis’ inaugural article collection with the Massachusetts Institute of Technology’s (MIT’s) Sloan Management Review. Since 1959, Sloan Management Review has been one of the world’s leading sources for research-backed guidance on business leadership and digital transformation.
It’s difficult to overestimate the impact of AI on business. The predictions — from trillion-dollar GDP boosts to radically reshaped workforces — are both mind-blowing and convincing. On the scale of the industrial revolution and digitization in centuries past, AI will define business in the twenty-first.
But AI’s potential is not inevitable. Only one in twenty enterprise AI pilots are achieving measurable P&L impact, reports MIT NANDA. There’s a huge gap between AI’s hype and real-world execution.
At its heart, AI is a decision engine. Give AI the data and context it needs to know how the business runs, and you empower it to autonomously make smart decisions that produce value. Give AI insufficient data and it will make flawed decisions.
The data that AI needs lives in the systems that underpin enterprise processes: in systems of record like ERPs, data lakes, emails, spreadsheets, and so on. Hidden in all of this is the blueprint for how the business works.
In the past few decades, “digital transformation” and “enterprise modernization” have driven companies to continuously bring in more systems, upgrade them, and migrate to new ones. There was a time when the best-run businesses could benefit from processes in a rigid system. But that was before the internet, and long before the cloud, automation, and AI.
Now, processes and value chains, like the supply chain, run across dozens or hundreds of pieces of technology. Eventually, the systems started to dictate the process. Suddenly, companies had to settle because of system limitations.
How do businesses expect to deploy AI on top of this? We dream of autonomous agents in our supply chain that predict and react to disruptions before they happen.
But if we ever hope to build them, businesses need to free their processes from their systems. In fact, 58% of business leaders are concerned the current state of their processes may limit what they can do with AI.
Once enterprises solve this problem and are able to build AI agents on business understanding, that is step one to achieving the much sought-after agentic, autonomous enterprise.
Then, orchestration will be critical: the ability to monitor, fine-tune, and steer a workforce of agents at scale. Just as HR has developed as a practice for developing and supporting human labor, AI orchestration will do the same for agentic workers.
The promise of AI is real, but not guaranteed. Mastering AI depends fundamentally on mastering business processes and their context. The companies that do this successfully are primed to enlighten Enterprise AI and unlock its full value.
Just like the great enterprise revolutions of centuries past, those who succeed will be the ones who truly know their business, and therefore how to transform it for the future.
Celonis has partnered with MIT Sloan Management Review to share three article collections with readers, free of charge.
This first collection of the series shares original writing from MIT academics on a variety of topics related to AI roadmaps, including:
Celonis and MIT SMR’s second and third article collections will premiere in early 2026.
