Large Language Models for Business: Prompting, Grounding, and Evaluation
Get reliable, trustworthy results from LLM tools by prompting well, grounding answers in your own content, and checking what comes back.
Course overview
Move beyond casual chat and get dependable business results from large language models in this two-day introductory instructor-led course. Learn how LLMs such as those behind Microsoft 365 Copilot, ChatGPT, Gemini, and Claude produce answers, write structured prompts that return consistent results, ground responses in your own documents and company data so answers are accurate and sourced, and evaluate AI output for accuracy, risk, and business value before you rely on it. Every module uses realistic workplace scenarios and hands-on exercises, with no technical background required.
Course coverage
Progress from how LLMs answer questions to prompting, grounding, and evaluating results you can trust at work.
- 01 How LLMs work: predictions, context, knowledge limits, and hallucinations
- 02 Prompt structure: role, task, context, examples, and format with RICECAR
- 03 Advanced prompting: iteration, step-by-step requests, and prompt templates
- 04 Grounding basics: why LLMs need your content and trusted sources
- 05 Grounding in practice: files, SharePoint, Copilot, notebooks, and custom assistants
- 06 Evaluating output: accuracy, completeness, sources, tone, and bias
- 07 Comparing tools and measuring value: quality, cost, and time saved
- 08 Responsible use: data protection, policy, review steps, and accountability
Participants will learn to
- Explain in plain language how LLMs produce answers and why they make mistakes.
- Write structured prompts that return consistent, usable results.
- Build reusable prompt templates for recurring business tasks.
- Ground answers in company documents and ask for cited sources.
- Set up a simple custom assistant with instructions and reference files.
- Check AI output for accuracy, missing information, and bias.
- Compare tools and measure quality and time saved on real tasks.
- Apply data-protection rules and human review before acting on AI output.
Two-day course outline
Day 1 - How LLMs work and prompting for reliable results: Learn in plain language how large language models generate answers by predicting text, what a context window is, why knowledge has a cut-off date, and why models hallucinate, show bias, and give different answers to the same question. Compare LLM tools including Microsoft 365 Copilot, ChatGPT, Google Gemini, and Claude, and understand the difference between consumer and enterprise versions. Structure prompts with role, task, context, examples, constraints, and output format using the RICECAR prompt framework. Practise iteration, asking for step-by-step reasoning, breaking large tasks into steps, and asking the model to check its own work. Build a library of reusable prompt templates for emails, reports, summaries, and analysis. Labs: improve weak prompts and create prompt templates for your team.
Day 2 - Grounding in your content, evaluating output, and responsible use: Understand grounding and why answers based on trusted sources are more accurate than answers from model memory alone. Ground LLM answers by attaching files, referencing SharePoint and OneDrive content with Microsoft 365 Copilot, using notebook-style tools such as NotebookLM, and building simple custom assistants such as custom GPTs, Gemini Gems, or Copilot agents with instructions and reference documents. Ask for citations and verify them. Evaluate output for accuracy, completeness, relevance, tone, and bias using checklists and scoring rubrics, and compare tools and prompts side by side. Measure business value with time saved and quality scores in a small pilot. Apply responsible AI practices: confidential and personal data, copyright, company AI policy, human review steps, and accountability. Final lab: build, test, and score a grounded document assistant and present recommendations.
Format and audience
For: Managers, business analysts, project managers, and professionals in HR, finance, marketing, sales, operations, and administration who want dependable results from LLM tools. Developers and IT staff should consider Large Language Models for Technical Teams.
Prerequisites: Comfort with email, Word, Excel, and PowerPoint. Some experience with a tool such as ChatGPT or Copilot is helpful but not required. Participants need a computer with a modern web browser and access to an approved LLM tool such as Microsoft 365 Copilot, ChatGPT, or Google Gemini.
How it works: Two days of instructor-led explanations, demonstrations, and hands-on workplace exercises. Contact us to discuss dates and virtual or onsite delivery.
Explore the course
Request a quote for the two-day Large Language Models for Business course and discuss your team's training needs.