Overview
Four cost centre heads ask the same question every month: where does our spend sit against target? The answer sits in Power BI, and someone in FP&A opens the report and types it out. The monthly pack goes the same way, rebuilt by hand every cycle.
Two days of building. Day 1 covers what an agent is, and everybody builds one. Day 2 is Copilot Studio, where an agent reads a published extract of your figures and answers from that alone. You leave with an agent that answers spend questions from your own numbers, and a flow that emails the monthly pack on schedule.
Who should attend
- FP&A, forecasting and budgeting staff who read Power BI reports every day
- Finance managers who sign off what those teams produce
- Participants who have done an introductory Copilot course; building agents is not assumed
What you will be able to do
- Brief Copilot Chat with RCR and the 5W1H checklist, and judge when Copilot can see company data.
- Decide whether a task deserves an agent or a saved prompt.
- Build and share a grounded agent in Agent Builder.
- Publish a Copilot Studio agent that answers only from a budget-versus-actual extract.
- Set up a scheduled flow that emails the monthly pack to named recipients.
- Audit your recurring work on an Effort and Frequency grid and rebuild one FP&A task as an agent.
Course content
Day 1: The foundation and your first agent
- Copilot Chat, Work IQ and the prompting foundations: a fast recap of RCR (Role, Context, Request) with the 5W1H checklist, when Copilot reads your own mail, files and meetings and when it answers from the web, and the privacy rules staff can repeat back.
- Your first Copilot agent: the two-question test that separates an agent from a saved prompt (how many people ask, and how often the answer changes), building an agent in Agent Builder on a spending policy and told to say "I do not know", and handing it to a colleague to find what breaks.
Day 2: Copilot Studio and the rebuild
- Building agents in Copilot Studio: a spending Q&A agent grounded on a published budget-versus-actual Power BI extract, where Power BI fits (agents read a published extract; questions put to the live model are demonstrated), and an agent flow that reads the current extract and emails the monthly pack on a schedule.
- Capstone, rebuilding one FP&A workflow: place your recurring work on the Effort and Frequency grid, put the top task through the two-question test, build the agent for it with the trainer beside you, and show the room the hours saved.
How it runs
- Hands-on building with interactive lectures, live demonstrations and one running case study.
- Bring a laptop signed in to your work account, with a Microsoft 365 Copilot licence and access to a Copilot Studio environment.
- No coding is needed.
- The capstone is presented to the room, with peer review. Delivered in person.
