What d.AP can actually do.
Seven building blocks that connect ERP, CRM and analytics through one ontology and deliver its definitions via MCP, A2A and REST to every agent, every app, every dashboard.
Data becomes knowledge.
At the core of d.AP is an ontology-based Knowledge Graph. It connects CRM, ERP, and analytics through shared semantics — in real time, without copying data. You query entities and relationships, not tables.
- Federated, no ETL The data stays in the source systems; d.AP accesses it in place via adapters. You define per source how often it synchronises.
- One definition per term “Customer” is defined once, with identical meaning across CRM, ERP and analytics. No more contradictory metrics between systems.
- Readable by humans and machines Your business teams read the same ontology that Aluna queries. d.AP writes new facts back into the graph.
)
Model your own business vocabulary.
An ontology defines in RDF/OWL which objects, properties, and relationship exist in your company. The Workbench turns this into a versioned model: defined once, retrievable in every use case.
- Classes, relations, rules Modelled visually, validated live against RDF/OWL. Domain experts help maintain the model, not just Data Engineers.
- Sandbox instead of production You test model changes in isolation; the production graph stays untouched until you release.
- Git-based version control Every model change is a commit: diffable, traceable, individually revertible.
)
Ask every question you have. Get grounded answers.
Aluna, the d.AP AI agent, translates natural language into graph queries and delivers answers as a table, chart, entity graph or text — with source, logic and filters.
- Graph reasoning, not RAG guessing Aluna navigates the ontology — classes, relationships, constraints — instead of guessing from text snippets of similar-sounding documents.
- Every answer verifiable Source, applied filters and confidence are attached to every answer. No result without provenance.
- −70% tokens per query Schema-RAG loads only the relevant slices of the ontology instead of whole context windows. That cuts AI processing costs directly.
)
Turn any answer into a widget.
d.AP dashboards are live views onto the Graph, not a BI copy. Every metric carries its definition, and logic with it and updates as the source changes.
- Logic stored in the KPI Behind every metric sits its executable SPARQL definition, open to inspect — not hidden away in a spreadsheet.
- Drill down into the graph Clicking a cell jumps straight to the underlying entity. No switching into the source system.
- Built once, embedded via reference One widget is referenced into other dashboards and apps — no copies, no follow-up maintenance.
)
Every answer is fully explainable.
At d.AP, transparency is architecture, not a feature. Every answer breaks down into source, joins, filters and aggregation. No black-box reasoning.
- Provenance per value Clicking a figure shows the source system, table and timestamp. The lineage is recorded automatically as you go.
- Reasoning trace Every step from question to answer can be inspected: intent, generated SPARQL query, result.
- Export for audits Lineage and trace go straight to your auditors or the regulator — built in, not reconstructed after the fact.
)
One model. Every system.
The Knowledge Layer is an API service, not a UI. Agents, apps and BI tools query the same verified definitions — defined centrally, valid everywhere.
- MCP & A2A AI agents and copilots access verified knowledge via MCP/A2A — instead of relying on fragile RAG pipelines.
- REST · JSON Every application reads the same semantics via REST/JSON without rebuilding its own knowledge logic.
- Iceberg Data Products BI and analytics read Iceberg tables built on the same definitions — metrics no longer drift apart.
)
Defined once. Used everywhere.
Every metric becomes a reusable metrics query: versioned, with metadata and a refresh strategy. No more re-implementing the same KPI in ten different tools.
- By hand or via Aluna Business teams formulate metrics in natural language too; the library makes them versioned and reusable.
- Real-time or cached Refresh strategy per metric: seconds for live data, hours for expensive aggregations.
- One metric, many consumers The same definition feeds dashboards, APIs and agents — no redundancy, no drift.
)
Seven capabilities. Six outcomes.
The seven capabilities above are the technology. Now the effect: six properties a Knowledge Layer needs so agents work productively instead of merely answering.
- §GROUNDING
Copilots and agents draw business context from your knowledge graph — with source, rule and role — instead of guessing from individual prompts. The model doesn't guess the schema, it reads it.
- Every answer points to a verifiable source.
- One term, one definition — across all departments.
- §KNOWLEDGE GRAPH
Business logic across ERP, CRM, MES, IoT and documents lives in one queryable layer — modelled once, not rebuilt per tool. The source systems stay unchanged; d.AP federates on top of them.
- Modelled once, usable in every use case.
- No bulk transfer — federation in place.
- §PROVENANCE
Answers and actions are documented down to the source — role-based and granular. The audit trail is created while answering, not after the fact.
- Source, rule, role — retrievable for every statement.
- Compliance without a separate logging system.
- §ONTOLOGY
Your business logic is modelled once in the ontology and reused by every new agent or copilot — no re-engineering for each new application.
- New use cases in days, not quarters.
- One ontology, any number of consumers.
- §TIME-TO-VALUE
Go-live with five to six sources in three months; every further source in one to two weeks — not in project cycles.
- No big-bang migration, no data standstill.
- Scales at your pace, not at the vendor's release cadence.
- §STANDARDS
Model and data live in open standards — RDF/OWL, MCP, A2A. No lock-in to an LLM, cloud or database; the ontology can be exported at any time.
- Portable — today and in five years.
- Vendors are replaceable, the knowledge layer stays.
See d.AP in action.
In around 60 minutes we demonstrate d.AP live: Aluna translates questions into graph queries and answers with source, logic and filters — on a real example model, with no setup on your side.
)
)