A new role for AI in manufacturing
In industrial analytics, artificial intelligence can assist early in the analytical process, at the point where analysts face the question: Which method is actually the right one here?
The StatSoft GPT Connector makes exactly this possible. As a node in Statistica Workspaces, it connects tabular process data and a problem description with an LLM, returning a concrete analytical plan tailored to the available data and the specific quality issue at hand.
A practical example
A quality engineer in a chemical production facility is working with a dataset of 40 columns covering raw material properties, process parameters across four production stages, and three final quality characteristics. A specification exceedance has occurred: the colour index of a specialty polymer has repeatedly fallen outside the permitted limits.
She opens a Statistica Workspace, connects the dataset to the GPT Connector, and describes the problem:
“This dataset covers 18 months of batch production data. The columns include raw material measurements (viscosity, moisture), reactor parameters (temperature, pressure, residence time, agitation speed), and final product properties. What analytical approach would you recommend to identify the causes of the colour index increase, and how could this be implemented in Statistica?”
The LLM recommends starting with a correlation and scatter plot matrix to screen all process variables against the colour index. As a next step it suggests a Random Forest variable importance analysis to rank the most influential predictors. It notes that time-dependent effects should be investigated using control charts, and highlights batch-to-batch variability in the raw material moisture as particularly worth examining. For implementation in Statistica it points to the Data Mining module and the General Regression Models node.
Within minutes, the engineer has a structured analytical plan, one that previously would have required either deep personal expertise or an external consulting engagement.
Data privacy and flexibility
One practical advantage of this approach is that it primarily transmits column names and a problem description rather than actual measurement values, allowing the flow of information to the cloud to be carefully managed. For organisations with strict data governance requirements, the GPT Connector also supports local models via Ollama or GPT4All, keeping the entire process within the organisation’s own infrastructure.
Looking ahead
As the reasoning capabilities of modern LLMs continue to improve, the precision of these recommendations will increase further. StatSoft is continuously developing the Connector and the broader ecosystem of AI-enabled nodes in Statistica. The goal is for AI to act not as a retrospective interpreter but as an active participant throughout the entire analytical process, from initial orientation through to final conclusions.
For manufacturing organisations looking to get more analytical value from their existing data and Statistica investments, this is a practical and immediately usable starting point.
++++
The StatSoft GPT Connector is available as part of StatSoft’s AI project offering. Contact StatSoft to discuss how it can be integrated into your analytical workflows.
