The Glass Box Living Network
How might we use a cognitive intelligence system built on a neuron-style architecture, that combines known asset data with real-time operational signals, weather events, field reports, regulatory updates and external intelligence to normalise real-world risk through transparent, evidenced, and explainable intelligence.
Problem / Challenge Statement
Asset intensive industries need to base investment decisions on whole-life cost and benefits. For high criticality assets, like pipes and pumps, well-established deterioration models as in use. Many other asset classes lack these models. Each is unique in construction, failures are infrequent, and there simply isn’t enough data to build traditional survival curves. Asset Planners are left relying on limited asset data, perception and experience to predict how these assets will behave over time, leading to inconsistent risk assessments and difficulty making like-for-like value comparisons across the full asset portfolio.
This sprint will seek to help Asset Planners generate risk profiles for these asset classes.
Rather than writing code & building static relationships, we will explore how new & emerging technologies can gather expert knowledge, cross-reference it against maintenance history and similar data sources, and producing a quantified risk-over-time profile that can feeds directly into decision making. The sprint will focus on select asset class as a proof of concept, with a clear path to scaling across other asset types.
Participants will collaborate through a mix of workshops, structured design sessions and real-world testing with Asset Planners to:
- Deep-dive into the problem with Planners and map data sources (CMMS, inspections, asset registers)
- Co-create the conversational logic and knowledge base content
- Test a prototype with real Planners on real assets, iterating the logic based on outcomes & feedback
- Demo a working prototype and explore a roadmap for scaling across asset types
A functional prototype, configured entirely through instructional text and a knowledge base within Copperleaf’s agentic framework, that can guide a Planner through a structured risk assessment for an asset class and produce a quantified risk-over-time profile ready for use in decision making.
A clear roadmap for extending the approach to other asset types, enabling more consistent, auditable, value-based decisions across the full asset portfolio.
Anyone working in asset intensive industries.
While the specific use case may be based around water sector data, the same problem persists in Electricity & Gas sectors in the UK & further afield. How do we make objective decisions with limited to no data?
Wider to this - data scientists and AI practitioners who are interested in practical applications of LLMs & conversational AI to engineering problems.
No technical expertise is required, as much as we need domain experts & data specialists & fresh perspectives are invaluable.
Any questions: Chris Errington (Chris.Errington@nwl.co.uk) and Danielle Ruddick (Danielle.Ruddick@nwl.co.uk)