Artificial Intelligence: The Art of Prompting

1 x 3 Hour Workshop (online or in person)

Sits between Generative AI Workshop 1 and Generative AI Workshop 2, and works on its own. For staff who have the basics and now want the one skill that decides whether AI is useful to them: how they ask.

Summary:

A follow-along workshop on writing prompts well, and then on writing prompts for agents. It starts from one reliable method, the four essential parts of a prompt, shows how generative AI actually reads what you write, and layers on the techniques professionals use: personas, one-shot and few-shot, question and answer prompting, question and answer grounded in a document, chain of thought and tree of thought, prompt chaining and meta prompting. It closes where a good prompt stops being a prompt and becomes a template, a scheduled job or an agent. Participants write their own prompts throughout, send them into the chat and are coached on improving them live. The course is not tied to one tool, so it runs on whichever chatbot the organisation uses.

Course Breakdown:

Prompt Engineering

• Crafting effective prompts: the four essential parts, goal, context, source and expectations
• Getting started with prompting, worked through on a real task
• How generative AI reads prompts, and what that means for the way you write them
• Prompting techniques: the full menu, from structured and role prompting through to chain of thought and meta prompting

Personas and Consistency

• One-shot prompting: creating a persona
• Persona and style replication for agents
• Use case: advanced persona design for brand and communication consistency
• Output format control
• Why sources and citations are key

Advanced Prompting Techniques

• Setting up a question and answer prompt, so the AI interviews you before it answers
• Assumption handling
• Question and answer from a document
• The “not specified in the document” fallback
• Humans in the loop, and why human verification matters
• Use case: article creation, with the AI proposing directions and pausing for approval

Ways of Thought

• Chain of thought prompting
• Tree of thought
• Use case: AI integration, three adoption strategies compared and one recommended
• Use case: meetings, a transcript, an agenda and a template run through a seven step chain

Advanced Prompting

• Prompt libraries: the move from writing prompts to reusing reliable tools
• Prompt templates
• Meta prompting: using AI to write and improve your prompts
• Scheduled prompting, and the use case of a weekly AI news digest

AI Agents

• Context made powerful: adding role and backstory
• What an AI agent is, and the anatomy of one: role, purpose, main task and target user
• Use case: the AI prompting agent, a low-risk first agent that rewrites your message into a stronger prompt
• Use case: the HR assistant, answering leave and policy questions from the organisation’s own documents
• Tool-calling agents
• Prompt chaining
• Prompt debugging: why prompts fail, and the five questions that fix them
• Decision framework: when to use a prompt, a template, a scheduled prompt or an agent

Outcome:

Participants leave with a repeatable method rather than a list of tricks, able to get a usable answer first time far more often, and able to recognise when a task is repetitive enough to hand to a template, a scheduled prompt or an agent. They take away a hints and tips book with every prompt used in the session written out in full, a plain English glossary of the methods, and a set of ideas to try in the week afterwards.