Generative AI Canvas

How Do I Use the Generative AI Canvas?

The Generative AI canvas helps you to generate application ideas for Generative AI in interdisciplinary teams. To do this, you run through the canvas twice: the first time from left to right and the second time from right to left. 

On colored sticky notes, note the input or output data and the functions on the data: green for existing data or available functions, red for missing data or functions, and yellow for planned or in-progress data or functions. 

Connect the input data, functions, and output data with arrows (on transparent notecards). A Generative AI solution can process multiple data sources and execute multiple functions side-by-side or in sequence, so it can also produce different outputs. 

Iteration: Start With the Input

Get an overview of the existing data sets in your company. Think not only of structured data, but especially of unstructured data such as documents (text, code & markup files), media streams (images, audio & video) or technical documents (diagrams, drawings, models, images etc.). Use the Data Landscape canvas as another tool if necessary. An input can also be a user input of commands and parameters for the AI, called a “prompt”.

Then consider what functions are possible and useful from that input data – and what the result of that function is. For example, a Generative AI might convert a textual description of a building into a photograph, a 3D model, or an engineering drawing. Also note the possible outputs. 

Iteration: Start With Output

Start the second round by asking what content the company or customers desire and write down the ideas on the right side under “Output”. Then, consider what functions are necessary to achieve this output and what inputs are required for it.

For example, a customer’s request could be that there is a textual description for every product. A generative AI could generate a summary based on the product photos for this purpose.