Manufacturing
AI engineering for manufacturing.
Production AI for an industry where IT and OT converge, where machine-safety standards apply, and where the failure modes are physical.
Posture
Manufacturing is the vertical where AI's intersection with operational technology is most direct. Predictive maintenance, quality inspection, process optimisation, autonomous robotics, energy-efficiency optimisation, supply-chain integration: most manufacturing AI has a direct or indirect OT integration, and most carries physical-system failure modes alongside software-system ones.
BtSavvy's posture on manufacturing is IT/OT-aware. The firm treats AI workloads that touch operational technology with explicit recognition that machine-safety standards (IEC 61508 functional safety, ISO 13849 safety of machinery, ISO 10218 robotic safety) and OT cybersecurity standards (IEC 62443) overlap with the AI-specific governance frame: and that the overlap is the engineering surface that must be designed in.
Regulatory posture
EU AI Act: manufacturing AI may fall under Annex III high-risk classifications particularly where it is a safety component of products covered by Union harmonisation legislation (Article 6(1)(a)).
EU Machinery Regulation 2023/1230: replaces the Machinery Directive; explicitly addresses AI-enabled machines and AI components used as safety functions, with essential health-and-safety requirements for such systems.
EU CRA (Cyber Resilience Act): cybersecurity requirements for products with digital elements, including AI components in connected manufacturing equipment.
IEC 62443: industrial automation and control system cybersecurity, applicable to AI integrated into ICS/OT environments.
IEC 61508 / ISO 13849 / ISO 10218: functional safety, machinery safety, and robotic safety, where AI is integrated into safety-instrumented or safety-related functions. NIST CSF (OT profile): for US manufacturers and any following NIST as a baseline. ISO/IEC 42001: AI Management System standard.
Which FORGE services carry the most weight
Assurance. The load-bearing service for manufacturing AI in OT environments: composing the conformance evidence the applicable regimes require, where they apply to the workload: EU AI Act, EU Machinery Regulation, CRA, and machine-safety standards.
Run. SRE and managed operations calibrated against physical-system failure modes, with hybrid automation and operator-in-the-loop discipline.
Build. Application engineering: vision-system inspection, predictive maintenance, process optimisation, robotic systems, energy management.
Data Foundation. Manufacturing data is heterogeneous, time-aware, and OT-sourced (sensor streams, MES, ERP, quality, supplier data) brought into agent-ready form under lineage discipline.
How FORGE works in manufacturing
Factory concentrates on accelerators calibrated against manufacturing failure modes: vision-model performance regression by lighting and material conditions, predictive-maintenance signal-quality monitoring, process-optimisation safety-bound enforcement.
Observability carries an explicit IT/OT dimension: AI-system observability correlated with physical-system observability, model output cross-checked against process-variable behaviour, calibration drift detected through both software and physical signals.
Review includes machine-safety review. The signed gate log produces evidence directly consumable by Machinery Regulation conformity assessment, IEC 61508 safety-case maintenance, and CRA cybersecurity evidence.
Governance and Trust treats safety-function classification as policy-as-code. AI components designated as safety functions carry stricter controls than non-safety AI; the boundary is enforced architecturally and at the policy layer.
Expert Intelligence combines frontier-model and traditional-ML competence with manufacturing-domain expertise: vision-model fine-tuning, time-series modelling for process control, retrieval against engineering documentation and quality archives.
Use-case sketches
A vision-system inspection deployment that emits EU AI Act and CRA conformance evidence continuously. A vision-AI quality-inspection system with performance-by-product-class observability, EU AI Act human-oversight integration enforced as a promotion gate, and CRA cybersecurity controls running as policy-as-code from the deployment substrate.
A predictive-maintenance system across an OT estate with IEC 62443 discipline. A predictive-maintenance AI deployed with IEC 62443-aligned cybersecurity controls, named-engineer approval at every model promotion, and traceability from OT sensor through model through maintenance action.
A safety-function-adjacent AI deployment with Machinery Regulation evidence. An AI component near (but outside) a safety-instrumented function, with explicit governance separation from the safety function, conformity-assessment evidence emitted by the delivery system, and the architectural separation enforced at the policy-as-code layer.
Insights cross-links
Evaluate FORGE against a manufacturing workload.
A FORGE Diagnostic takes one manufacturing workload you select and engineers a working slice of it (your standards, your requirements, your security controls and governance, full CI/CD from the first commit), deployed and running, then produces a signed readiness assessment and a proposed next step.
Metadata — Title: BtSavvy — AI engineering for manufacturing · Meta: Production AI for manufacturing with FORGE — EU AI Act, EU Machinery Regulation, CRA, IEC 62443, machine-safety discipline. Signed at every gate. · Canonical /industries/manufacturing · Links — Assurance, Run, Build, Data Foundation, FORGE platform /forge, Book a Diagnostic.