From system start to the self-learning machine
Three expansion phases for the complete AI system on your own hardware. We see AI as a cyber-physical system: it captures data from machines and processes, evaluates it with specialised models and is designed to interact with its environment under control. Define the right scope for data integration, AI functionality and governance.
Scope overview
| Merkmal | Phase 1: System setup & productive start | Phase 2: BI & knowledge development | Phase 3: The self-learning machine |
|---|---|---|---|
| Use case & target group | |||
| Target group | SMEs / teams | Mid-sized companies & enterprise | Corporations & special requirements |
| Number of users | <= 10 users | 10 - 30 users | >= 30 users |
| Typical use cases | 1–2 prioritized use cases | Multiple use cases / departments | Complex domains, OT/edge, specialized processes |
| Time-to-Value | Fast (days) | Planned (a few weeks) | Project-dependent (complexity-driven) |
| Scalability | Limited | High | Unlimited |
| Roadmap & scaling plan | |||
| Cyber-physical system & self-learning machine | |||
| Capture: business, process and document data in one data pipeline | |||
| Capture: machine and plant data (OT/edge) | |||
| Evaluate: specialised models with measurable quality per use case | |||
| Approvals & outcomes are captured (human-in-the-loop) | |||
| Models tuned specifically to your use cases (fine-tuning / LoRA) | |||
| Act under control: guardrails & approvals as an environment for agentic AI | In development | ||
| Self-learning from approvals and outcomes | Goal, together with the customer | ||
| Deployment & operations | |||
| On-premise operation | |||
| Air-gapped / offline operation possible | |||
| Docker container deployment | |||
| Optional Kubernetes / multi-node | |||
| Defined offline update process | |||
| Backup/restore concept (runbook) | |||
| Monitoring / health checks | |||
| High availability (HA) / redundancy design | |||
| Hardware & scaling | |||
| All-in-one box (HW+SW preconfigured) | |||
| Scalable from workstation to server | |||
| Multi-GPU / high-load inference | |||
| Capacity planning (CPU/RAM/GPU/storage) | |||
| Data sources & integration | |||
| Index documents (e.g. PDF/DOCX/TXT) | |||
| Databases (basic connectors) | |||
| Extended connectors (e.g. ERP/CMMS/MES) | |||
| SharePoint / collaboration systems | |||
| OT/edge data integration (sensors/gateways) | |||
| Data harmonization / mapping (light) | |||
| Data harmonization (complex, multiple domains) | |||
| Data lifecycle: retention and deletion concept | |||
| Document processing & RAG | |||
| Standard RAG (chunking + embeddings + vector store) | |||
| Hybrid search / reranking | |||
| Source display / citations in answers | |||
| Structured extraction (tables/forms) | |||
| Quality metrics (recall/precision) for RAG | |||
| OPAIRS assistant | |||
| Standard OPAIRS (with database access) | |||
| Special OPAIRS (with domain knowledge and integration) | |||
| Agentic OPAIRS (with tool access) | |||
| Automatic OPAIRS (with automation) | |||
| AI models & inference | |||
| Verified base model set | |||
| Model registry / versioning / approval process | |||
| Model routing (multiple models by task) | |||
| Fine-tuning / LoRA (if needed) | |||
| Guardrail OPAIRS (policies, tooling, output checks) | |||
| Workflows & automation | |||
| Standard workflows (ingest, re-index, Q&A) | |||
| Approval flows (human-in-the-loop) | |||
| Integrations with ITSM/CMMS/ticketing | |||
| Custom agents/tools | |||
| Robotics/automation (project-specific) | |||
| Security baseline | |||
| TLS / reverse proxy / network segmentation | |||
| RBAC (roles & permissions) | |||
| Secrets management (no plain text) | |||
| Hardening (server + container) | |||
| Central logging (system + app) | |||
| Audit log (queries, admin actions) | |||
| Prompt injection protection (basic) | |||
| Regular security tests (injection/policies) | |||
| Governance & compliance | |||
| Integration of Apache Ranger & Apache Atlas (in development) | |||
| EU AI Act readiness (transparency & oversight) | |||
| ISO 27001-ready documentation package | |||
| ISO 27701-ready (PII flows, deletion concept) | |||
| NIS2-ready (logging, incident process) | |||
| Traceability (sources, logs, exports) | |||
| Support & delivery | |||
| Standard documentation (install/runbook) | |||
| Extended documentation (security/audit packs) | |||
| Training for admins/power users | |||
| SLA / maintenance windows / incident process | |||
| Project operations / long-term enablement | |||

Phase 1: System setup & productive start
Build a sovereign AI and data foundation in your own infrastructure.

Phase 2: BI & knowledge development
Systematic expansion of analytics, knowledge and decision capability.

Phase 3: The self-learning machine
Developed together with you: models that get more precise with every use case, and workflows that run automatically under control. Learning from approvals and outcomes is our shared goal; the controlled environment for agentic AI is in development.