
Decision Intelligence
Root Cause Intelligence
From anomaly to explainable action
Your system flagged that something went wrong. But detecting a failure is just the beginning. Understanding why it happened is what matters.
Root Cause Intelligence identifies the most likely cause with supporting evidence and recommends next steps to resolve the issue.
Built for Microsoft and Databricks environments, using native platform capabilities wherever possible.
The Problem
A Failure alert pops and you immediately think, what do we do now?
The Problem
Your Dashboard Says the Number Moved. What Happens Next?
Organisations have dashboards, alerts, anomaly detection and even monitoring, yet there are operational failures.
Although the failure alerts are automated, someone still has to manually work out why it happened. And what to do about it next? This process can take weeks, that delays resolution and impacts the operations.
Why Did it Happen?
Does Anything Need to be Done?
What Evidence Supports that?
What Should We Do Next?
How it Works
Same Path. One Traceable Decision
Every anomaly becomes a case. Every case follows the same path and every step leaves a record.
HOW IT DECIDES
Rules for Known Paths, Agents for Deeper Investigation
Use rules for known paths and AI agents for complex investigations or combine both to identify root causes.
Deterministic
A known and controlled path where the investigation follows predefined rules.
-
Explicit business rules and decision paths
-
Predictable and repeatable behaviour
-
Traceable, evidence backed decisions
-
Operational and compliance-sensitive decisions
Agentic
An open and complex path where the path to the answer is not fully known.
-
Open ended investigation
-
Many possible sources and hypotheses
-
Anomaly types that change over time
-
Interactive evidence gathering within approved tools and data
Hybrid
A combination of known and complex investigation paths, using the approach best suited to each situation.
-
Deterministic decisioning where the path is known and controlled.
-
Agentic decisioning where the path is open and complex with no set rules for investigation.
-
Highlight human intervention where needed.
WHAT YOU RECEIVE
A Decision You Can Check, Not Just an Answer
Each investigation produces a decision card: the outcome, the evidence behind it, how confident the system is and which route reached it.
No Action Required
Sales dip at Store 108 matches a planned range change. Evidence: range file, planogram date. Confidence: high. Route: deterministic.
Human Escalation
Cost spike in a production workspace. Deployment and usage logs conflict. Confidence: low. Route: agentic — escalated to platform owner.

REFERENCE ARCHITECTURE
Runs on the Platform you Already have
Root Cause Intelligence is a solution pattern implemented with native Microsoft or Databricks services in your tenant, on your data, publishing into your channels. There is no separate Cloudaeon runtime to buy or operate.
HOW TO GET STARTED
Start with One Operational Failure
Begin with a recurring operational failure that costs you time and effort to investigate.
Where it leads to
A defined starting point. Room to scale
Sprint → Production → Scale → Managed Service
First engagement clarifies:
-
Selected failure, metric & required data
-
Business expert input & evaluation criteria
-
Next step scope for production use
WHERE IT APPLIES
One Pattern. Many Operational Exceptions
The investigation pattern stays the same. The data, rules and actions are specific to each operation.
Retail
Failure
Product sales fall below expectations
Investigate
Stock availability, delivery and promotion data
Potential Response
Check shelf availability or follow up a delayed delivery
Manufacturing
Failure
Production output falls below expectations.
Investigate
Equipment downtime, maintenance history, material supply and quality holds.
Potential Response
Schedule maintenance or reallocate materials.
Supply Chain
Failure
A delivery service level target is breached.
Investigate
Supplier performance, inventory, carrier activity, warehouse operations and demand.
Potential Response
Expedite a shipment, identify an alternative source or rebalance stock.
Financial Services
Failure
Failed payments increase unexpectedly.
Investigate
Upstream system changes, transaction processes and relevant operational evidence.
Potential Response
Escalate the issue to the responsible team with supporting evidence.
FAQs
Root Cause Intelligence uses operational alerts, KPI anomalies, threshold breaches, forecast deviations, and other exceptions identified by your existing monitoring systems or data science models. It enriches these events with relevant business context, historical data and operational records to investigate possible causes. Data can be accessed through your existing data platform and approved integrations.
Root Cause Intelligence is designed to work with your existing data environment, with Databricks serving as its reference platform. It uses governed data access and Databricks capabilities to investigate anomalies, gather evidence and recommend actions. The implementation approach depends on your current architecture, data sources and integration requirements.
Root Cause Intelligence provides evidence backed recommendations to support operational decisions. Business teams retain oversight and cases with low confidence can be escalated for human review. Deterministic rules handle known investigation paths, while AI agents investigate more complex cases within approved data and tool boundaries. The level of human involvement depends on the decision policy and operational requirements.
The initial engagement identifies a specific operational failure, the relevant metrics and data sources and the business expertise needed to investigate it. It also establishes evaluation criteria and defines the next steps towards production use. The scope is tailored to the selected use case, existing data environment, investigation requirements and agreed outcomes.
After the initial engagement, Cloudaeon and your team can define the implementation scope for production use. This includes determining the appropriate investigation approach, validating the required data access, establishing evaluation criteria and planning integration with existing operational systems. The next phase depends on the findings, technical requirements and priorities established during the initial engagement.
Yes, it does, but detection alone is not the core of the solution. It consumes anomalies and events from your existing data science models, BI alerts, monitoring or operational systems. Where no detection exists, Cloudaeon adds a simple detection step. The real value lies in what happens after an anomaly is identified.
Not on its own. Known anomaly types follow deterministic business rules. A constrained AI agent is leveraged only when the investigation path is open-ended and it has access only to approved tools and data. Every route follows the same decision policy, evidence and confidence are checked before an outcome is returned.
It escalates the case to a person. Accuracy thresholds are agreed with your team. Cases below those thresholds go to a designated owner with the evidence gathered so far. The solution does not manufacture an answer when the evidence is insufficient.
Yes. The same solution approach can be implemented using the native services of each ecosystem, including Microsoft Fabric, Azure and Foundry or Databricks. It can also work across both environments where required.
The cost and duration depend on multiple factors. Like the operational failure being investigated, the data sources involved and the complexity of the required investigation. The initial engagement defines the use case, relevant metrics and data, business input and evaluation criteria. Cloudaeon then establishes the scope, timeline and cost for the next phase based on these requirements.

