Hi Emily,
we have only been using Claude for a few days and previously worked mainly with OpenAI’s ChatGPT. However, our experience may still be relevant, as many of these workflows could probably be transferred to Claude.
At the moment, we primarily use AI for quality assurance across our large number of learning courses. More specifically, we have created a custom GPT that analyses SCORM packages exported from Evolve and identifies potential issues, including factual inaccuracies, spelling and grammar mistakes, and readability problems.
Since ChatGPT can also search the web, we configured the custom GPT to use reliable sources and provide both the sources and their publication dates. This allows us to verify the information ourselves. Of course, a human still needs to review every issue identified by the AI, but we have been using this workflow very successfully for several months.
I have also created another version of the custom GPT that extracts the entire Evolve course as text. This means that, in theory, we can also make our learning content available as a text document for further processing.
Setting this up took some time because we first had to teach the AI how Evolve Legacy courses and their SCORM packages are structured. It now works very well. I use a knowledge file that describes the structure of Evolve courses in detail and helps the GPT interpret the content correctly.
Apart from quality assurance, we also use AI to help create learning content.
Unfortunately, our company is currently in the process of switching to a competing product: Articulate. I personally do not support this decision, but I understand the main argument behind it: Articulate already offers integrated AI features.
I am still hoping that Evolve will finally deliver the AI functionality that has been announced, especially as my initial experience with Articulate’s AI has been rather negative. For example, it has generated unsuitable images and incorrect information, and there does not appear to be a reliable way to check the generated content for factual accuracy.
If you can provide AI functionality that is also capable of checking whether content is factually correct and up to date, that would offer real value to everyone. I realise that accessing and evaluating external sources could involve significant token usage and additional costs, but there may be technical or pricing models that could make this feasible.
Another reason many people are interested in tools such as Articulate is their ability to generate complete courses from a few prompts and uploaded source files. It is perhaps unfortunate that some of the creative process of building courses manually may be lost. However, if the generated courses follow sound instructional design and didactic principles, this could still be an important capability for Evolve to consider.
Personally, I have not yet found a workflow in either ChatGPT or Claude that can reliably generate complete courses while meeting our methodological and instructional requirements.
I would be very interested in joining the beta testing, but unfortunately I simply do not have the time at the moment, either professionally or privately. I also suspect that there are more experienced AI users out there who may be better placed to provide detailed testing and feedback.
Best regards
Lars