Agree that costs via Agents or optimized prompt books are improving.  But reality is that the cost of Security Copilot seems to be a lot higher per prompt compared to using another ChatGPT based model directly (especially for larger scale automation use cases - factor of 10x or even 100x).
In these automation use cases some effort is going into building it anyway, so the convenience argument isn't as strong here.   It would be great if there was an alternative model option via Security Copilot for simple automation use cases (e.g. allowing a simpler / cheaper GPT model like mini maybe /AskGPTMini?)