AI is a great tool, if you know how to use it (as with most tools). However the quality can highly differ depending on which model you use. 

I currently use chatgpt: gpt-4o

The difference between gpt 3 and gpt 4 is night and day. 3 gave a lot of unwanted imperfect results. 4 is very accurate and can help you fine tune your prompts to perfection. For more advanced logic I would advise chaining commands as AI tends to make mistakes when the request is too long. So I usually break a request into a bunch of smaller requests. 

What I use AI for: In enrichments I create data for an empty database to get real mock data that actually has references and lookups and is restricted to the data model (domains, constraints etc). That way i can quickly fill a database with some mock data instead of having to insert everything by hand. 

Having developed several Enrichments that make use of AI, I notice its potential for the Software Factory. We use the gpt-4o model and it works quite well. Getting to a prompt that works all the time is a long road and requires a lot of trial and error. I did notice that using prompt-chaining was a good way to keep the LLM focused on the task at hand rather than hallucinating or deviating from what was desired to receive.

For developing the model itself, I don't use it but for writing logic and stored procedures such as Tasks, it is great! Github Copilot is a great extension that helps with a lot of the code that has to be written, such as auto-generating Join clauses or auto-completing variables and functions.

If there are any Generative AI providers that you miss in the Thinkwise platform, feel free to create an Idea here on the Community.