Generative AI for Work

What generative AI is and is not, large language models explained without mathematics, and why AI can produce excellent work one moment and confidently invent something the next.

• Chatbots, search tools, research tools and AI assistants
• ChatGPT, Claude, Gemini and other current platforms
• Context, memory and limitations
• When normal search is better, and when a human should stay in control
• What not to put into a public AI tool

Exercise: give several AI systems the same task, compare the outputs, then work backwards to understand why they differ.

The quality of AI output depends heavily on the quality of the job it has been given. This is less about a magic prompt formula and more about explaining work properly.

• Giving context that changes the answer
• Defining the job, the outcome, examples and constraints
• Asking AI to expose its assumptions
• Useful output formats and breaking large tasks into stages
• Using AI as a critic, reviewer and thinking partner
• Building reusable instructions

Exercise: develop a vague everyday prompt into a professional, reusable briefing method.

AI becomes much more useful when it works from real material rather than answering from nowhere.

• Researching unfamiliar subjects and asking better research questions
• Source checking and evidence
• Summarising and comparing PDFs and reports without losing important detail
• Analysing meeting transcripts and notes: actions, risks and unanswered questions
• Turning messy information into a useful structure before a meeting or decision

Exercise: work from a real source pack and produce a useful, source-grounded briefing.

The goal is not to become good at chatting with AI. It is to improve how work gets done.

• Identifying repetitive knowledge work and mapping your current process
• Deciding where AI should enter the process
• Reusable prompts versus reusable workflows; projects and workspaces
• Simple specialist assistants, human approval points and quality checks
• Privacy and security habits, and keeping your system useful when tools change

Final project: map one real task, redesign it with AI, test the new version and document a workflow you can use again after the course.

You learn AI by using it. We deliberately use imperfect examples, because learning how to fix a bad AI result is more valuable than watching an instructor produce a perfect demo. Bring examples from your own work: the closer the exercises are to a real problem, the more useful the course becomes.

AI tools change quickly. We review the teaching stack before each cohort and use the strongest current tools for the job, typically ChatGPT, Claude, Gemini, Perplexity and NotebookLM. The course focuses on methods that stay useful when the software changes.

This is not a machine-learning or AI engineering course. We do not teach neural network mathematics, model training, Python for machine learning, TensorFlow or PyTorch, fine-tuning or enterprise AI architecture.

If your aim is to become an AI engineer, you need a more technical path. If your aim is to become much better at working with AI, you are in the right place.

A laptop and access to at least one major AI platform. A free account is enough for some exercises; a paid account gives access to more models and features. We explain the options before the course, so you do not need to subscribe to everything.