Knowledge from your sources
Connect file shares, SharePoint, scans and readable databases. Originals remain the authoritative sources. A designated reviewer can correct uncertain OCR results.
NXTGN AI CORE
AI using your knowledge and terminology.
NXTGN AI CORE brings AI to your company knowledge without sending sensitive data to external language models. Local open-weight models, an ontology of your domain and an in-memory knowledge graph answer questions about contracts, support cases and assets. Every answer has evidence, and access rights apply at every data level. AI CORE runs on your servers or as SaaS in Germany and is in productive use at customers.
Explore scope and workflow ↓KNOWLEDGE GRAPHConnect assets and components
- SiteNorth plantcontains ↓
- Production lineLine 3contains ↓
- AssetPumpcontains ↓
- ComponentImpeller
NamesLaufrad · Impeller · Pos. 14Source informationManual · maintenance instruction
Scope
What we deliver for you
Ontology, taxonomy and knowledge graph
AI CORE derives a domain model from your knowledge. Terms, aliases and relationships are prepared, and your specialists approve them. You can upload an existing ontology or define your own. The knowledge graph runs in memory and answers accordingly fast.
Terminology and language rules
Maintain preferred names, synonyms, abbreviations and prohibited terms. The terminology supports answers and text checks. A terminology specialist can work without administrative access.
Keywords and meaning
AI CORE combines exact and semantic retrieval. Multilingual embeddings connect questions to sources in other languages without translating the collection. Retrieval expansions and original evidence remain traceable.
Security at every data level
AI CORE respects access rights at every data level. Permissions from the sources are checked before content is retrieved and synchronised every five minutes by default. Unclear permissions mean no access. Your intellectual property stays protected.
Local models instead of token costs
AI CORE works with local open-weight models on your servers or as SaaS in a data centre in Germany. Sensitive data goes to no external language model, and there are no token costs from a model provider.
Agents on your knowledge
Agents search, compare and analyse, for example for contract questions, in support or on assets. They deliver reports with evidence.
Architecture
Your knowledge base stays with you, even when Copilot asks.
A question can start in the AI CORE portal or in Microsoft Copilot. It always passes through the same security layer. Only the question, answer and evidence leave your environment.
Your company or a data centre in GermanyYour data sources
AI CORE portalSearch and dialogue · diagnosis · terminology · handwriting · administration
Interface for assistantsMCP, always included, for Copilot and other AI assistants
Security layerIdentity · access profile · permission trimming · audit
NXTGN AI COREOrchestration · agents · policies · evidence required
Hybrid retrievalStructured query · semantic search · graph search
Structured knowledgeFacts and master data
Vector indexMeaning and similarity
Knowledge graphOntology and relationships
IngestionExtraction and OCR
Local language modelEmbeddings and answers
Sign-inEntra ID or OIDC
- ERP and business systems
- CRM
- Data lake
- SharePoint
- File shares
Crossing the boundary
Question, answer, evidence
Stays with youData, index, knowledge graph
Optional: Microsoft cloud
Microsoft Copilotin Teams, SharePoint or CRM
Copilot Studio agentOrchestration
Language modelphrases the answer
Entra IDshared sign-in
Approach
How your domain model is built
You do not need to bring an ontology. AI CORE proposes one from your knowledge. If you already have one, we use it.
- 01Connect knowledge
Documents, contracts, manuals and databases with their permissions.
ResultA knowledge base with access rules
- 02Recognise terms
AI CORE finds terms, synonyms and abbreviations and prepares aliases.
ResultA proposed glossary
- 03Derive the ontology
Classes and relationships are proposed. You can upload or define your own ontology.
ResultA domain model of your company
- 04Review and approve
Your specialists confirm, add or reject.
ResultApproved company knowledge
- 05Questions and agents
Answers, analysis and agents work on the knowledge graph.
ResultFast answers with evidence
Comparison
Cloud AI or AI CORE
Today
- Sensitive data goes to external language models
- Costs rise with every token
- The model does not know your terminology
- Source permissions get lost on the way
With NXTGN
- Data and models stay with you or in Germany
- No token costs from a model provider
- Ontology and knowledge graph know your terms
- Access rights apply at every data level
In use
Productive at customers, piloted by large companies
AI CORE is in productive use at customers, and several large companies are testing it in pilot projects. Typical questions concern contracts, support cases and knowledge about assets. Your knowledge stays in your company or in a data centre in Germany.
- Operation
- On your servers or as SaaS in Germany
- Models
- Local open-weight models
- Domain model
- Derived automatically or your own ontology
Your company does not have to adapt its terminology to a model.
Purchasing and maintenance may name the same part differently. An internal abbreviation may mean different things at different sites. A larger language model alone does not solve this. AI CORE connects these names to the domain context defined by your experts.
Glossary
A concept can have several names. Your team defines the preferred name and adds synonyms, abbreviations and prohibited terms. AI CORE uses this terminology when retrieving and phrasing information.
Taxonomy
A taxonomy classifies terms and objects: an impeller is a component, a pump is an asset. This hierarchy helps retrieve information from related subordinate areas.
Ontology & knowledge graph
The ontology describes the types of things in your domain and how they relate. The knowledge graph represents concrete links: this component belongs to this asset, and this manual describes it. In AI CORE, the domain model consists of the glossary, taxonomy and confirmed relationships.
AI CORE can propose relationships from source content. Your specialists approve or reject them. AI does not define company-specific meaning without review.
Ask in German and find the answer in an English manual.
AI CORE combines keyword and semantic retrieval. Part numbers remain searchable exactly; differently phrased questions are also matched by meaning. Multilingual embeddings connect questions and sources across languages without translating the source collection.
The evidence remains the original. A translation can appear alongside it as a reading aid using maintained terminology. It does not silently become a new source. AI CORE also indicates when a match was found through a synonym or relationship.
Answers, analysis and diagnosis using the same knowledge base.
Analyse data
For analytical questions, the language model plans the query and the database performs the calculation. Results include evidence tables and stated exclusions. ERP or CRM integration requires readable tables or views and appropriate permission information.
Support service and diagnosis
Your specialists maintain diagnosis paths for recurring problems. AI CORE guides users through them and records whether a case was resolved, escalated or stopped. Draft knowledge articles are published only after human approval.
Monitor changes
A schedule or source event can start an agent run. It searches, compares or analyses data and supplies an evidenced report within the recipient’s access scope. These AI CORE agents report; they do not modify business data. Process agents with write access are a separate integration service.
The product description distinguishes included features from demonstrable functionality. A demo checks the current version against your requirements. Product description and feature status ↗
How your teams can use AI CORE
NXTGN AI CORE is a standard product. Knowledge sources, terminology and access are configured for your business. The knowledge graph, terminology checks, analysis and diagnosis use the same knowledge base. Models and data remain on a dedicated instance.
01
Contract questions
Anyone who needs to know which notice period, liability or price clause applies reads through many contracts and amendments.
What we deliver
AI CORE recognises contract types, parties and clauses and answers questions with the exact passage. Confidential contracts are visible only to those allowed to see them.
What you gain
Fast answers for purchasing, sales and legal, without contract data leaving the company.
How we measure value
Search time per contract question, answers with a source passage.
02
Terminology across departments
Purchasing, service and engineering use different names for the same component. Internal abbreviations and legacy terms hinder retrieval and coordination.
What we deliver
Your specialists maintain names and relationships. AI CORE connects the glossary, taxonomy, knowledge graph and sources. Matches indicate whether a synonym or relationship was used.
What you gain
Knowledge remains usable when another term is used or an experienced colleague leaves. Domain definitions are traceable and can be corrected.
How we measure value
Terminology coverage, correctly linked sources, unresolved names and cross-team questions.
03
Evidenced analysis
A question spans several documents or database rows. One text match cannot provide a comparison or total.
What we deliver
For analytical questions, the model plans and the database calculates. AI CORE shows evidence and excluded records. Access and permissions are defined for each integration.
What you gain
Your team can trace results to the source. Missing information stays visible rather than being replaced by a convincing estimate.
How we measure value
Traceable evidence, reviewed results, disclosed exclusions and comparison or aggregation errors.
04
Diagnosis knowledge and recurring reports
Not everyone knows the next check after a fault. Changes to source content are often discovered incidentally.
What we deliver
Specialists maintain diagnosis paths. AI CORE guides users through them and records outcomes. Reporting agents retrieve, compare and analyse on schedules or source events. They do not modify business data.
What you gain
Service knowledge becomes available to less experienced employees. Owners receive evidenced reports and see resolved or escalated cases.
How we measure value
Resolution and escalation rates, knowledge gaps, professional approvals and report completeness.
Clarify before delivery
Your specialists own terminology and relationships. Evidenced answers do not replace professional review. A demo shows current functionality; we assess source quality, integrations, access and operational requirements for your installation.
Is AI CORE just search with chat?
No. Retrieval is one function built on a structured knowledge base. The glossary, taxonomy and knowledge graph represent terminology and context. Analysis, diagnosis paths, terminology checks and reporting agents use the same foundation.
Who maintains the ontology and terminology?
AI CORE prepares terms, aliases and an ontology from your knowledge. You can also upload or define your own ontology. Your specialists approve what applies, so an unreviewed AI guess never becomes company knowledge.
Is the product custom-built?
The product core is fully standardised. Your sources, terminology and settings are configurable. Additional custom integrations are agreed separately.
Where does AI CORE run?
On your servers or as SaaS on servers in Germany. Models, embeddings and OCR run there, not at an external model provider.
Are there token costs?
Not from a model provider. AI CORE uses local open-weight models. Costs depend on the instance and operation, not on the number of questions.
Can AI CORE be used in Microsoft Copilot?
Yes. AI CORE always includes an MCP interface. Copilot and other AI assistants ask their questions through it. The answer comes back with evidence, and the knowledge base stays with you.
Can the included agents modify systems?
No. AI CORE agents retrieve, compare and analyse. They supply reports with evidence. Write operations in CRM, ERP or business applications require separately designed process integration.
See AI CORE using your terminology and sources.
We show how the domain model, multilingual retrieval, evidence and access controls work together. We then discuss the requirements for your installation.
