Volume 1 · Day 1 — Fundamentals of Artificial Intelligence

A leadership team is told to "add AI" to the product. Before a single line of code is written, someone has to say clearly what AI is, what automation is and where analytics ends and learning begins.

Project connection: Identify AI, automation and analytics in ten real-world scenarios.

Reflection: where in your own organisation does this problem already exist today?

• Define intelligence and artificial intelligence in academic terms.
• Distinguish AI as a research field from AI as a deployed product.
• Compare human intelligence and machine intelligence across perception, reasoning, learning, planning and language.
• Differentiate AI, automation, analytics and traditional programming using decision criteria.
• Decompose any AI application into input, processing/model and output.
• Identify at least six industry applications of AI and explain the value each creates.
• Correct five common misconceptions about AI.

Prerequisites: Basic computer literacy: files, applications and web services., Comfort reading simple tables of data., No mathematics or programming background is required for this module.

Skills developed: Intelligence, Artificial Intelligence, Perception and reasoning, Automation vs AI, Analytics vs AI, Input → Model → Output.

Tools used: VS Code, Jupyter, Google Colab.

Intelligence, in the cognitive-science sense, is the capacity of an agent to acquire information from its environment, transform that information into usable internal structure, and select actions that advance a goal. Artificial intelligence keeps that structure and replaces the biological substrate with computation: sensors or datasets provide the information, mathematical models provide the internal structure, and software actions provide the behaviour.

Two very different things are called 'AI' in ordinary conversation. AI as a field is a body of research spanning search, logic, probability, statistics, optimisation, neural computation and linguistics. AI as a product is a specific application — a fraud score, a churn prediction, a chat assistant — built by assembling data, a trained model and an interface. Keeping these separate prevents a great deal of confusion: progress in the field is measured in published methods, while progress in a product is measured in business outcomes and reliability.

So a plain-language definition is: AI is software that produces useful answers for situations nobody explicitly programmed, because it learned the pattern from examples. It is not a mind, it is not aware, and it does not want anything. It is a statistical function that has been fitted to data and then asked to generalise.

Most failed AI initiatives fail at the definition stage, not the modelling stage. A team is asked to 'add AI' to a process that is actually a deterministic rule ('block any transaction above the limit'), and they spend a quarter training a model that a single conditional statement would have solved more reliably and more cheaply. The reverse also happens: a team writes ten thousand hand-tuned rules for a problem — document classification, say — that has too many exceptions to enumerate, and the rule base collapses under its own maintenance cost.

Being able to say precisely whether a problem is a rule problem, a reporting problem or a learning problem is therefore the first professional skill in this programme. It determines the architecture, the cost, the team and the risk profile of everything that follows.