Everyone's Asking About AI. Few Are Asking About the Right AI.

We've had a few enquiries recently asking what AI functionality is built into the equipment and systems we supply. It's a fair question, and one that would rarely have come up two years ago.

Most of that shift traces back to large language models - ChatGPT and similar tools putting a conversational idea of AI in front of everyone. That's useful for a lot of things. But it's also narrowed what "AI" means to most people down to one shape: a chat interface that answers questions.

That's not the AI with real results in research and applied engineering over the past decade. The methods with an actual track record in industrial settings are narrower and built for a specific job:

LSTM networks: recurrent architectures for time-series data, used for forecasting demand, load, remaining equipment life, or process drift.

CNNs: convolutional networks for image and pattern recognition, used in visual inspection and defect detection on production lines.

PINNs: physics-informed neural networks, which build known physical constraints into the model so predictions for closed systems stay grounded in real system behaviour rather than pattern-matching alone.

None of this is new. These are established, published methods with years of results behind them. What's changed is that customers are finally asking about AI - they just don't have the vocabulary yet to ask precisely. "Does this have AI in it?" usually means "can this predict, detect, or optimise something that used to depend on someone noticing it first?"

That's the better question to start with. The right answer depends on the problem you're trying to solve, not on whether "AI" appears on a spec sheet.