I've spent my career at the intersection of machine learning, operations research, and real-world decision-making — as a researcher at MIT, and as the founder of Ikigai, where we set out to prove that AI could do more than describe what's happening in a business. It could tell you what's going to happen, and what to do about it.
That's the promise of decision intelligence. And it's harder than it sounds.
Most enterprise AI today is retrospective. It analyzes what happened, surfaces patterns, and generates recommendations. That's genuinely valuable, but it's not enough.
The decisions that matter most in any organization aren't about what happened last quarter. They're about what happens next: whether to adjust inventory ahead of a demand shift, whether a supply chain disruption requires rerouting now or can wait, whether a pricing decision made today will hold up under three different demand scenarios.
Answering those questions requires something most AI systems don't have: the ability to simulate. To model a business as a dynamic system, stress-test decisions against multiple futures, and recommend actions that are robust across scenarios — not just optimal for the most likely one.
This is what Ikigai was built to do. And joining Celonis is what allows us to do it at an unprecedented scale.
Celonis built the foundation that decision intelligence needs to work.
Decision intelligence without operational context is like a flight simulator without a real aircraft model. You can run the simulation, but the outputs only matter if the underlying representation of the system is accurate. Celonis's Context Model is the most sophisticated real-time operational model of enterprise processes that exists. It knows how work actually flows across systems, where the exceptions are, where the friction lives, and how the business behaves under different conditions.
When you combine that with Ikigai's forecasting and simulation capabilities, something qualitatively different becomes possible. AI that doesn't just understand your operations, but can look forward into them. That can say: here are three scenarios, here's what each costs you, here's the decision that holds up best across all of them.
That's not a feature. That's a new category of capability for the enterprise.
For organizations already working with Celonis, the addition of decision intelligence means your operational data — the processes, the patterns, the exceptions already captured in the platform — becomes the input to a forward-looking intelligence layer. The work you've done to map and optimize your processes now feeds directly into the ability to simulate and improve them before changes are made.
For organizations new to Celonis, the Context Model now offers something end-to-end: from understanding how your business runs today, to designing how it should run tomorrow, to operating it with AI agents that act on that intelligence in real time.
And for the broader industry, I believe this marks a crucial turning point. The question enterprises have been asking — "we've invested in AI, why aren't we seeing the returns?" — has a structural answer. AI without context can't understand your business. And AI without decision intelligence can't improve it. The Celonis Context Model, with decision intelligence at its core, addresses both.
In my role as Chief Scientist, Enterprise AI at Celonis, I'll be focused on three things:
I'll be sharing more in the weeks and months ahead on the science behind decision intelligence, on what we're building, and on the broader questions I think the industry needs to answer as AI moves from experimentation to operation.
The missing layer of Enterprise AI is here. The work of making the most of it starts now.
Keen to learn more about Decision Intelligence? Meet Devavrat and some of the Ikigai Labs team at our upcoming Process Intelligence Day in New York City on June 4.
