Talk to us

Free Claude course for marketers

Beyond the chat box

You already use AI every day. This is how to get it doing the work, not just answering questions. Here are the slides, the files and the exact prompts from the session, so you can run it yourself.

Watch the walkthrough

The slides

Your browser can’t show the slides inline.

Try it yourself

Rebuild the two demos from the session in about 15 minutes. Brightfield is a fictional payroll software company we made up, so the files are safe to upload to any AI tool.

Step 1. Download the files

  • Brand voice guideHow Brightfield writes.Download Brand voice guide
  • KPI definitions and targetsWhat each number means and what good looks like.Download KPI definitions and targets
  • Monthly report templateThe structure every report follows.Download Monthly report template
  • Marketing campaign dataA year of made-up campaign data.Download Marketing campaign data

Download all files (.zip)

Step 2. Build a reporting Project

Create a Project in Claude, add the three guides as files, and paste this into the Project instructions.

You're a marketing analyst for Brightfield Software, a UK payroll software company. Every report you write follows the monthly report template and uses the KPI definitions and targets in the project files. Write in the Brightfield brand voice: British English, specific, no emoji, no hype words. Always compare results with the targets. Flag any channel more than 20% off target. If the data doesn't support a conclusion, say so rather than guessing.

Then start a chat in the Project, upload the campaign data and ask:

Write the monthly marketing report for September 2026.

Next month, upload new data and ask again. Same structure, every time. That’s the point of a Project.

Step 3. Turn the data into a dashboard

In a new chat, upload the same spreadsheet and ask:

Turn this into an interactive dashboard for our marketing team: monthly spend, leads and cost per lead by channel.

Then keep asking. Add date and channel filters. Ask what you spent on LinkedIn. Works on the free plan.

Before you use your own data

  • Strip it. Remove names, emails and anything under NDA before you upload.
  • Turn off training. In your settings, choose not to let your chats train the model.
  • Check the answer. Spot-check one number yourself. If it’s right, trust the rest more. If it’s wrong, start again.

Want this working inside your team?

We run hands-on Claude workshops for marketing teams, built around your own data and the reports you already produce.