We all know that our GLAM collections hold important data for research in the humanities. An increasing number of institutions are opening up collection data for public use through APIs or downloads. But how do we help non-technical users understand the possibilities of large scale collection data? How do we create pathways that lead them through the code to the questions they want to ask? In my latest attempt to introduce researchers to the wonders of collection data, I've created an GLAM workbench that uses Jupyter notebooks to combine live code with worked examples and tutorials. With minimal setup, and a few presses of Shift+Enter, anyone can harvest bulk data from Trove, or analyse series from the National Archives of Australia -- all within their browser. The workbench is an experiment in itself, as I learn more about the technology and explore different approaches. But I think it offers some exciting possibilities.