Overview
A hands-on introduction to Microsoft Copilot Chat for everyday work. Attendees learn to write effective prompts, check unreliable answers and use AI safely for practical tasks from their own role. The course covers browser-based Copilot Chat (not Microsoft 365 Copilot integration inside Word, Excel, PowerPoint or Outlook).
What you’ll need
- Laptop, reliable internet connection, headphones and microphone
- Google Chrome or Microsoft Edge
- Work account with Copilot Chat and access to agent features
- Microsoft Excel
- Pen and notebook
Learning outcomes
By the end of the course, attendees will be able to:
- Explain in plain language what an LLM is, what it does well, and where it fails
- Write prompts using a persona / context / constraint / format structure
- Chain two or more prompts to complete a realistic multi-step task
- Detect likely hallucinations and apply at least two validation techniques
- Apply their organisation’s data-classification rules to decide what is safe to paste into Copilot
- Build a personal prompt library of 3–5 prompts tied to their own role
Course topics
AI Foundations
- Machine learning versus generative AI; a non-technical model of large language models
- ChatGPT, Copilot, Gemini and Claude; awareness of image, video and music generation
- Strengths in drafting, summarising, transforming and brainstorming; limitations with numbers, facts and reliability
- Creative versus analytical work: choose the right tool for the task
Prompting Fundamentals with Copilot
- Using organisationally approved Copilot Chat; work-account data protection and organisational policy
- The four-part prompt: Persona / Context / Constraint / Format; rewrite a weak prompt
- Prompt chaining: draft → refine → reformat; turn rough notes into a polished email
- Prompt libraries: repeatable prompts rather than clever one-offs
When Things Go Wrong
- Recognise hallucinations and patterns that suggest an answer may be invented
- Check claims against reliable sources and test outputs against known examples
- Re-ask and request a critique to expose inconsistencies; agreement is not independent verification
Data Safety: Green / Amber / Red
- Apply the organisation’s policy using a traffic-light guide: public/generic, internal/non-sensitive, and restricted information
- Handle identifiers, customer personal data, supplier pricing and contracts appropriately; do not paste prohibited data
- Practical redaction and rewriting; what to do after an inappropriate disclosure
Workshop Part 1: Common Work Tasks
- Tailored workplace exercises with debriefs, using supplied examples where possible
- Summarise an email thread and draft a reply; clarify a confusing stakeholder or supplier request
- Turn meeting notes into actions with owners and dates
- Rewrite a paragraph in formal, friendly and technical tones; choose for the audience
- Describe an Excel formula’s goal, generate a formula and validate it with test data
- Research with a structured prompt and verify three key claims against primary sources
Workshop Part 2: Agents and the Prompt Library
- What is an agent? Use the Prompt Coach agent and build a simple custom agent
- Select a recurring task from the attendee’s role; develop, test and refine a reusable prompt
- Swap prompts with a partner and stress-test the results
- Compile a personal library of 3–5 reusable prompts for work
Wrap and Commitment
- Review the tools and techniques; identify the first workplace task to try
- Questions, next steps and the future of AI and agents