Real Use Cases & Building Your Own Agents
Built around one team's own documents and problems, using their real work rather than examples.
306 words
Carried overfrom a document supplied by the client, taken 7 September 2026 and not rewritten. Source URL not supplied.
1 x Full-Day Bespoke Session
Built around one team’s own documents and problems, using their real work rather than examples.
Summary:
A bespoke day that starts from the questions a team actually has and works through each one properly, ending with them building an agent that lives in a spreadsheet they already own. Every case uses their own material. It suits a team that has done the basics and wants to see what is genuinely possible.
Course Breakdown:
Know your use cases
- Opening on the team’s own questions, and setting up a question and answer prompt that keeps returning to them
Quick wins
- Turning PDF documents into Word
- Pulling a participant list out of meetings and documents
- Validating role and permission conflicts
Comparing spreadsheets and reconciling tickets
- One problem solved three ways
- Option: Copilot in Excel
- Option: Copilot Chat
- Option: a script
- How to choose between them
The tender review agent
- Using AI as a structured evaluation assistant against published criteria
- Where the judgement must stay with a person
Human in the loop
- Article creation with verification built into the process, and why that matters more than the output
Templates that fill themselves
- A stakeholder deck template populated from real material
Persona-based simulator agents
- One-shot and few-shot prompting
- Creating a persona
- Building a simulator to practise against
Agent workflows
- An event creation team of agents
- The different ways agents can support work
- The honest limit of a built agent
The agent that lives in a spreadsheet
- Everything already in Excel
- Data in a table
- A rules sheet the agent reads
- How it works and where it slips
- A prompt library
- The part that rewrites its own rules
Outcome:
Participants leave with several of their own problems already solved, and with a working agent built in a tool they already have, that they can extend without help.