Operator Horizonte require a data product for reliable billing, commercial indicators and care supported by AI. The committee calls for a proposal that can be approved and that engineering can prove. The whole case is fictitious and solved by discussion without access to a console.
The incumbent and Restrictions
There are 10 000 client logs and 12 400 contracts. A person can have several contracts. Two systems use different keys and some phone numbers are shared. The date of birth is missing from 700 records. It's not a required data for this product. Instead, 120 contracts have no valid reference to the client. One dashboard calls customers "assets" with any contract and another requires an end-of-day contract.
The proposal should respect these conditions as appropriate:
- Filling: processing only valid contractual and client relations; to preserve and resolve the rejections.
- Analytics: to publish "client at the close" with definition, date of cut and reconciliation.
- Customer Care: show an authorized agent only what is necessary for the assigned case.
- AI: to base responses on adopted and existing documents; To scale ambiguities to a person.
- Consent: To distinguish aims, cancellations and validity. Do not infer permission from marketing from an active contract.
- Operation: Fictional goal of freshness of 30 minutes for contractual changes, agreed recovery and attributable expenditure by product.
The latency, volume and thresholds are supposed to be educational. The actual assessment requires them to be measured with the specific sources and requirements of each organisation.
Your Proposal for the Committee
Develop your proposal using the DAMA → Oracle matrix. For each decision, indicates owner, monitoring, evidence and response to non-compliance.
| Area | Question to be responded to |
|---|---|
| Data Government | Who approves the definition and exceptions of use? |
| Data architecture | What roles do Oracle AI Data Platform and Oracle Autonomous AI Lakehouse play, and which connections must you validate? |
| Modeling Data and Design | How do you represent client, contract and consent without duplicating meanings? |
| Storage from Data and Operations | How do you detect delays, restore a version and attribute costs? |
| Security of Data | What does each profile see and how do you check its effective permissions? |
| Data integration and interoperability | How do you detect duplication, reattempts and late changes? |
| Documents Management and Content | How does an existing document get to AI and how does it stop using? |
| Data Master and Reference | What attributes link and what conflicts require human review? |
| Data Warehousing and Business Intelligence | What is the indicator’s grain, and how do you avoid multiplying amounts when joining tables? |
| Metadata management | How do you identify consumers affected by a change of contract? |
| Quality of Data | Why do you block 120 contracts and do not discard 700 customers without date of birth? |
No commercial names are evaluated. A good response connects the goal with the mechanism and recognizes the conditions to prove that it works.
Solution Reasoning
To see a possible proposal and its boundaries
The Commercial owner approve "an active client at close" as a person with at least one contract valid at the date of cut. The steward records the definition and its exceptions. Engineering models source keys, correspondences and views A shared phone does not become a universal identifier.
Oracle AI Data Platform co-ordinates ingestion and transformations with permissions by role. Oracle Cloud Infrastructure GoldenGate is an option if the origins and freshness target warrant CDC; a lot can also fulfill if its duration and frequency have been shown. The freshness control compares the brand of origin with the publication, including delays and reattempts. The last loaded data does not guarantee that all changes are present.
The quality standards are implemented about the data: 120 contracts without valid client will go to quarantineand owner resolution. 700 customers with no date of birth stay if they meet product rules. require an unnecessary data would add friction and exposure. The result of that particular rule is 12 280 contracts with valid reference to 12 400, approximately 99,03 %. That doesn't prove that the other quality dimensions are satisfied.
Oracle Autonomous AI Lakehouse Serves SQL structures with explicit grain. Inputs and amounts aren't directly attached to all conjunctions to compute income, because that relationship can multiply rows. The client indicator uses the rule of existence of an existing contract and reconciles with a known sample. Oracle Analytics Cloud publishes an approved definition in a semantic model Reusable by analysis. The analyst checks the filters and aggregations of the consultations generated with assistance from AI and reconciles the indicator with the sample.
Security A dedicated platform operator, engineer, steward, analyst, care agent and auditor. Each access is checked from its effective identity, including inheritance and credentials of external connections. Oracle Data Safe can prepare protected copies of supported bases. The row or column controls are applied to the engine that supports them. are not attributed by analogy to all catalogs. An extension about Oracle AI Data Platform for classification and protection keeps human approvals and evidence of its logic.
The owner documentary published only commercial conditions currently in place and approved at the knowledge base. It keeps an inventory of versions, withdrawal procedure and re-ingestion verification. The assistant needs permission to recover content and for any subsequent actions. RAG does not give these authorisations.
Metadata record definitions, identifiers, liability and connections. The Oracle AI Data Platform lineage It shows sources and consumers of the caught artifacts. The team reviews these units before changing a table and retains versions, executions and approvals to rebuild the decisions of the period.
Operations and FinOps they review executions, warnings, recovery and tagged spending. A budget alert opens up a review of consumption. It doesn't stop the platform by itself. An agreed process decides what work to reschedule without losing evidence or breaking up with priority services.
Oracle Fusion Cloud Enterprise Data Management is evaluated when master domain management, its approvals and match / merge warrant its requirements and scale. It's not required to put them in to solve every duplication of a table. The architecture retains the DAMA criterion while changing the tool.
Acceptance and Closing criteria
Assign from zero to two points to each criterion: zero if missing, one if mentioned and two if explain with a verifiable example.
| Criterion | Evidence of a complete response |
|---|---|
| Responsibility | Owner and steward identified and derogations with monitoring |
| Meaning | Entities, aim, validity and grain consistent with examples and indicators |
| Protection | Reportable actual accesses II and document treatment by context |
| Reliability | Rules, quarantine, freshness, recovery and impact of changes with evidence |
| Feasibility | Applicable controls, verified requirements and assigned expenditure |
A pedagogical goal is eight of ten pointswithout omitting protection and liability. That's not an official certification threshold. Repeat the explanation with a change: a client revoks marketing while maintaining an existing contract. You have to be able to show what use changes, what operation continues and what evidence both decisions will prove.
He goes back to the sections of the course if some criterion of the heading remains unjustifiable. The closure of the workshop is the ability to uphold and revise a proposal and not the memorization of a list of products.