Artificial intelligence deployments are failing at scale, and the reason, according to Celonis, is not the technology itself but the absence of context.
Patrick Thompson, SVP of Customer Transformation at Celonis, said the newly launched Context Model is built around precisely that problem.
"Without context, without a context model in your process intelligence, AI struggles to work. If you look at studies by MIT, you'll find 85–90% of AI fails because they have no context - they just have some raw data in a data warehouse, and agents hallucinate all over that. Whereas if you contextualise the data and create a digital twin of how your operations are working, now all of a sudden AI is intelligent," said Thompson.
The Massachusetts Institute of Technology's GenAI Divide: State of AI in Business report, released last year, found that despite USD $30-40 billion in enterprise generative AI investment, roughly 95% of AI pilots fail.
Celonis was founded in 2011 as a process mining company, focused on bringing visibility into business processes rather than operating in the AI space. Thompson said the platform has since evolved into the distinct category of process intelligence.
"Process intelligence is becoming so relevant is because of AI. Celonis ... didn't intend to necessarily play in the AI space, but to bring visibility to processes, and that's process mining. Now the platform has evolved to a group of systems have come together to create process intelligence, almost to some degree creating the magic quadrant that Gardner has created for us."
The expanding sector is reflected in Gartner's addition of a Process Intelligence Magic Quadrant in 2026. He said the move reflects the market's evolution beyond process mining into broader process intelligence capabilities.
He frames customer transformation around five strategic pillars that organisations typically pursue at the executive level: digital transformation, supply chain transformation, shared services or GBS transformation, customer experience, and operational improvement. The role of platforms such as Celonis, he said, is to enable those initiatives, not simply to deploy technology.
The platform offers three temporal views of business operations: the past, present, and future. Users can examine historical process data down to individual user IDs, timestamps, and durations; assess what is happening in real time against a designed process or "happy path"; and run forward-looking simulations to model what would happen if processes were restructured or AI components were introduced.
A separate capability, called Process Adherence Management, monitors whether teams are following newly designed processes after a transformation project has been completed - addressing the gap between what an organisation designs and what it actually does. The solution serves as the backbone, enabling users to identify value opportunities by comparing their event data against a "should-be" process model in a guided manner.
Thompson will be a keynote speaker at the Sydney and Melbourne Process Intelligence Days on July 21 and 23. His sessions will focus on how customers see the platform strategically and transformatively, and how both global system integrators and technology partners can leverage it effectively.