Process Operations - AI Enablement Technology Study
This study report documents Stage 1 of the three-stage approach: the desktop evaluation and stakeholder engagement. Its purpose is to understand the current OT/IT data landscape, identify the key data sources (PCS, historians, SCADA, MES and enterprise systems), document the data flows and integration points, and define the potential target use case(s) — then evaluate the shortlisted platforms against that need and close out with a recommendation that sets up the Stage 2 live trial.
- Identify (covered by this project)
- Select (covered by this project)
- Define
- Execute
- Operate
Challenge / Problem: PLS's Pilgangoora operation is a lithium (spodumene) processing site with an established OT/IT estate spanning process control, SCADA, historian and enterprise data systems. Through the IT–OT Workshop, PLS set a strategic direction toward AI-assisted operations, predictive maintenance, and integrated operational data — but the value of its OT data to process optimisation and advanced control remains unrealised without the right platform to contextualise it.
PLS engaged Pivotal Engineering to evaluate technology options against that need, with the objective of providing an evidence-based path toward AI-enabled operations at Pilgangoora.
Solution: The study followed a three-stage approach. Stage 1 (this study) was a desktop evaluation with a demonstrable simulated sandbox for stakeholder engagement, closing out with a study report and recommendation. Stage 2 is a live trial of the shortlisted technology against real plant data and production systems to provide a substantial evaluation of use cases, ROI, and the role of AI in process control. Stage 3, informed by the learnings and recommendations of Stage 2 and subject to approval, will be a full-scale implementation across P2000 and the existing P1000 plant.
The Pilgangoora processing operation generates a large volume of operational data, yet only a subset reaches higher-level systems, and what does arrive traverse numerous hops with little engineering context attached. No platform within the current business application estate is suited to realising PLS's vision of AI-enabled processing operations and predictive maintenance. This study evaluated two platforms against the common crushing-circuit dataset:
- Option A: (FactoryTalk DataMosaix / Cognite Data Fusion) — a mature, feature-rich platform, but heavier, costlier, and more time-consuming to configure for data ingestion;
- Option B: (Ignition + ThredCloud + Claude) — a lighter, lower-cost stack that contextualises data at the source, supports plain-language queries, and builds directly on Ignition, PLS's new MES platform and likely selection for P2000 SCADA.
Outcome / Results: This study successfully recommended a technology stack for the Stage 2 trial and captured some key potential use cases to evaluate the technology against. Through these findings a business case was put together that outlined the possible benefits and return on investment.
It further recommended a system architecture which adheres to the "one source, full context, AI-ready" principle. The proposed architecture reduces system complexity and delivers cost savings by consolidating multiple licensed layers into fewer integration points, and materially improves maintainability:
- contextualisation is performed once, at the source, and carried through the pipeline rather than reconstructed at each layer.
Pivotal Engineering is now performing the Stage 2 Live Trials with PLS.
Platforms Evaluated: Ignition 8.3, Thred Cloud, Claude AI, FactoryTalk Datamosaix (Cognite Data Fusion)