How ClearOps makes service parts planning decisions understandable, verifiable, and actionable
A €20 spare part missing at the wrong moment can bring a €500,000 machine to a standstill. For manufacturers, preventing that downtime depends on thousands of daily service parts decisions: will the right part be in the right place when it's needed? The natural assumption is that better forecasting solves this, and planning algorithms have indeed gotten genuinely good at predicting hard-to-model aftermarket demand.
However, in building a planning application for this exact problem, we found that prediction accuracy wasn't where things actually stalled.
The Bottleneck Was Trust
Planners routinely encountered recommendations that looked correct but unusual. Instead of acting on them, they stopped to check: opening the ERP system, reconstructing inventory positions, manually rebuilding reasoning the planning system had already run.
An accurate recommendation a planner can't quickly verify costs about the same as a bad one, because the planner still has to do the verification work either way.
Planning systems don't only fail when they're wrong. They also fail when they're right but unexplained.
What Unexplained Recommendations Cost
Manual investigation takes planners away from higher-value work, but the bigger cost is what happens when uncertainty changes their behavior. A recommendation that's difficult to validate gets overridden. Safety stock gets increased defensively. Emergency shipments get triggered because the planner doesn't have enough confidence in the system's output.
These aren't irrational reactions. A planner is ultimately accountable for keeping parts available and customers' machines running. The result is a costly combination of more manual work, unnecessary inventory, emergency orders and stockouts.
That's why explainability isn't simply a user-interface improvement. If AI is going to make planning decisions, the reasoning behind those decisions needs to be available when the planner needs it.
Building on Celonis Context
This became the focus of our work during the 5th Global Ecosystem Hackathon, which challenged teams to build real-world applications combining the Celonis Platform with AWS technologies such as Amazon Bedrock. Our ClearOps team was named the EMEA regional champion.
We built ClearWhy AI, a working prototype that uses a multi-agent approach to explain the outputs of our service parts planning algorithms.
Instead of simply showing a planner the result, ClearWhy works through the reasoning behind the decision and turns it into an explanation a planner can understand and validate.
A Reorder Point of 13 units, for example, can be explained through the protection window, expected demand, safety stock, demand variability, holding costs and the required fill-rate target. An Order Quantity can be explained through its economic order quantity, constraints and current inventory position.
- The planner no longer has to reconstruct the reasoning from scratch.
- The algorithm produces the decision. AI explains the decision.
The bigger opportunity comes from connecting that reasoning to operational context.
Service Parts Planning is designed as a Celonis-native planning application, combining ClearOps' purpose-built forecasting and inventory optimization with the operational data and context available in the Celonis ecosystem.
Celonis provides the context. ClearOps turns that context into a planning decision. AI explains why that decision makes sense. This is how we see ClearOps extending the value of Celonis: from understanding what is happening in the operation to deciding what should happen next.
Why It Matters
The immediate goal is planner productivity: recovering a meaningful share of the time currently lost to manual investigation. More importantly, fewer investigations should mean fewer defensive overrides and fewer emergency orders, since both are typically reactions to uncertainty rather than to a bad recommendation.
That's the mechanism through which explainability is meant to translate into planning outcomes: not by making the algorithm smarter, but by making its recommendations easier for a planner to act on directly. Explainability turns the intelligence behind a planning recommendation into something a planner can understand, validate and act on, helping improve parts availability while reducing inventory costs and planning effort.
The Takeaway
Autonomous planning does not start with automation. It starts with trust, and trust starts with explanation.
That's the idea behind what we're building with Celonis: a Service Parts Planning application built directly inside Celonis, bringing together Process Intelligence, operational context, AI-powered planning, and explainability in one solution.
While planning applications already exist, this approach is fundamentally different. Instead of separating planning from the operational reality of the business, we use the context and data already available in Celonis to explain why a specific planning decision is being made.
This creates a new path from Process Intelligence to action: from understanding what happened, to deciding what should happen next, to explaining why. The result is planning that doesn't just generate recommendations, but gives planners the confidence to act on them.