Updated May 2026
Plain-English definitions for the AI, computer vision, machine learning and automation terms an Australian business actually needs to know. No buzzwords, no consulting-speak.
The application of artificial intelligence and machine learning to industrial environments — production lines, factories, quarries, mining sites, worksites, warehouses, packaging operations. Differs from general-purpose AI by emphasising edge deployment, integration with existing operational technology, and reliability in harsh conditions. Full guide →
The branch of AI that enables systems to interpret and act on visual information from cameras and image data. Used in industrial settings for defect detection, load verification, safety monitoring, object counting and quality control.
A software program powered by an AI model that can take autonomous actions on behalf of a user — reading email, drafting responses, scheduling, retrieving information, executing multi-step workflows. Differs from a chatbot by acting proactively rather than only responding when prompted. Try one →
Running AI models on local hardware on-site — an edge GPU box, a smart camera, an on-premise server — rather than streaming data to cloud services. Reduces latency, keeps sensitive data on-premises, and removes cloud bandwidth and connectivity dependencies. Particularly important for Australian operators in remote or regional locations.
A computer vision model trained on data from a specific environment — your site, your equipment, your products — rather than relying on a generic off-the-shelf model trained on internet data. Required for most industrial applications because generic models don't recognise site-specific objects or events. Learn more →
A widely-used family of real-time object detection models that process an entire image in a single neural network pass. YOLO models are common in industrial computer vision because they're fast enough to run on edge hardware at production line speeds, on cameras already deployed at sites.
Computer vision systems that detect and read vehicle number plates from camera footage. Used at quarry gates for vehicle tracking, at depot entries, and at any site that needs to log or match vehicle movements automatically.
Computer vision systems that automatically detect whether workers are wearing required personal protective equipment — hard hats, high-visibility vests, safety boots, glasses, gloves — in real time on construction sites, factories and worksites. Common starting point for industrial AI on safety-regulated sites.
An open-source AI agent framework that runs on a user's own devices and connects to messaging channels like WhatsApp, Telegram, Slack and email. Used to deploy private AI assistants that handle real business work — enquiries, quoting, document processing, scheduling — without sending data to third-party SaaS. Setup →
An AI model trained on huge amounts of text that can read, write, summarise and reason in natural language. Examples include Claude, GPT-4, Gemini and Llama. The underlying technology behind modern AI agents and chatbots. LLMs are not databases — they generate plausible text, which is why business deployments combine them with RAG (see below).
A technique where an AI model retrieves relevant information from a private database or document store before generating a response. Used in business AI to ground agent responses in the company's own documents, pricing, products and policies — rather than relying on the model's general training data.
Software that takes a multi-step business process — an enquiry comes in, gets classified, a quote is drafted and sent, follow-up is scheduled — and executes the steps automatically. Modern workflow automation combines deterministic rules with AI-driven decisions and natural-language tasks.
Computer vision systems that automatically count loader scoops, track truck movements, and generate a verified record of every load leaving a quarry, mine or aggregate site. Replaces manual paper dockets with an auditable digital record. See Quarry Vision →
Computer vision applied on a production line to identify surface defects, dimensional errors, colour variations, misalignment or contamination on manufactured items — in real time, at line speed, without sampling. Manufacturing →
The first step in any industrial AI deployment — an on-site visit to evaluate existing cameras, network infrastructure, workflows, and the specific events or data the client wants captured. Determines whether existing infrastructure is fit for purpose or where additions are genuinely needed. Industrial AI offers this free.
The act of running a trained AI model on new data to get a prediction or classification — as opposed to training the model. In industrial AI, inference happens on every camera frame as it arrives, on local edge hardware, in milliseconds.
The principle that data should be stored, processed and accessed under the laws and within the geography where it was created. For Australian industrial operators this often means camera footage and operational data must remain on-site or within Australia and cannot be sent to overseas cloud services.
Anthropic's command-line AI coding agent that reads and edits code, runs commands, and completes engineering tasks autonomously. The underlying technology Industrial AI uses to build many custom AI systems and to run OpenClaw deployments.
An open standard from Anthropic for connecting AI models to external tools, data sources and services. Allows AI agents to read from databases, send messages, control browsers, and integrate with business systems without bespoke code per integration.
This glossary covers the terms Industrial AI sees come up most often in real client conversations. If there's a term you want explained, get in touch — we'll add it.
Get in touch