Generative AI Canvas A0 (47" x 33") synthetic paper (EN)

Generative Artificial Intelligence systems (AI for short) extend analytical AI with the ability to not only analyze data, but to generate new data as well as textual and visual content. Well-known examples of generative AI are the text-based chatbot "ChatGPT" and Midjourney for image generation. For companies, there are diverse and promising application possibilities for such AI solutions. The Generative AI canvas helps to identify those use cases. 

The Generative AI canvas maps how an AI-based solution works to generate new content:

• The AI system expects one or more inputs: these can be existing content such as image or video files, text collections or program codes, and vector graphics or presentations. New data, for example, entered by a user to control the AI (referred to as a "prompt"), also flows into the AI-based generation.
• The AI system then performs one or more functions on the input data, for example, in the case of image files: Mask faces, i.e. black them out; detect and highlight certain objects; sharpen out-of-focus elements; generate a description of the image; and / or translate it into another image format.
• Finally, the AI produces one or more outputs: for example, different versions of the generated images, a text description, or a 3D model of the scene being displayed. AI systems that support diverse inputs and outputs are considered multimodal. 

The canvas was developed together with the AI Foundation.

You can use the Generative AI canvas as part of the Data Strategy Design method to design AI applications or applications of Autonomous Analytics (see Analytics Maturity canvas). Or you can use the canvas independently, for example, as part of an AI Use Case Ideation workshop.

1. Start by listing the available data sources under Input. These can be your own or third-party data sources. The Data Landscape canvas helps you explore your data landscape.
2. Then consider what data or content your employees or customers want, or what they currently create manually. Make a note of these "data silos" under output.
3. Finally, under Function, design a function flow for how an AI generates the output(s) from the input(s). An AI solution can use multiple functions and execute them sequentially (from left to right on the canvas) or in parallel (one below the other on the canvas) to produce different outputs.

 For more information visit Datentreiber and Foundation Factory. 

The Generative AI canvas is also available as an online version including step-by-step instructions.