Our product lines
Three product lines
Maintenance · OpSight Maintain
Recognize components, structure distributed technical knowledge
Industrial part identification with computer vision and knowledge graphs. From photo to the correct ERP code in seconds.
Process Engineering · OpSight Execute
Automate repetitive document work, give experts their time back
AI agents for safety dossiers, commissioning plans, warehouse acceptance. The agent does the repetitive work, the expert reviews where it matters.
Warehouse · OpSight Observe
Read SAP movements in real time, anticipate discrepancies
Discrepancies caught during the month instead of at year-end close, structured root cause analysis, alerts prioritised by impact.
The problem
Industry knows what needs fixing. It often doesn't know where to look.
Three knots run through Italian manufacturing: technical knowledge that lives in people's heads and not in systems, document-heavy processes that eat skilled time, and operational signals that only surface when it is too late to act.
The right component at the right moment
Catalogues without photos, codes that differ plant to plant, operators mistaking similar parts. Technical know-how isn't inside systems.
Skilled hours buried in PDFs
Safety documentation, commissioning plans, acceptance dossiers: repetitive content in unique contexts, today consuming expert time.
Signals that emerge at year-end
Thousands of daily SAP movements, warehouse adjustments that only become visible at close. Causes get lost, the process repeats.
Method
Three frontier technologies, one orchestration.
The three product lines grow from the same method. The unity of perception, structured knowledge and agent is what sets OpSight apart from a collection of disconnected tools.
Computer vision
Custom models to recognise what the industrial eye meets every day: a component, a worn label, a document anomaly. Not generalist APIs — vertical datasets and pipelines trained on the domain.
Knowledge graphs & advanced retrieval
Beyond classic RAG. We build living enterprise knowledge graphs — knowledge structured as entities and relations, not as passive archives. It's the evolution of retrieval toward what's beginning to be called a wiki-LLM: a knowledge base that grows with every interaction, queryable in natural language and readable by agents.
AI agents
Not chatbots: agents that execute multi-step flows — observe state, query the knowledge graph, produce artefacts (documents, classifications, alerts) and stop where expert validation is needed.
The heart of OpSight is the orchestration of these three layers inside the industrial context. The method — today the subject of a filed patent application — is precisely the unity between perception, structured knowledge and agent.
Innovative startup →An industrial process that AI could automate?
We work with manufacturing, plant-engineering and construction companies that have repetitive processes, disorganised catalogues or scattered documentation. Let us talk.
Book a free call →