hello,
No magic trick, some calculator on internet, but they just act "how many action, do you do rag on sharepoint" very basic.
First of all : every power automate premium connector you use are not charge on premium power automate licence : the copilot credit cover everything.
since the previous page have many content, needed to be read, most of the case are cover by the table of the page
to understand :
if you query sharepoint with work IQ : 10 credit
Then the agent use AI to generate an answer 2 credit
So common shortcut is : when you ask a question to an agent it's 12 credits.
Letting the agent use AI to decide wich action to do : 5 credits;
-> so the agent send a mail : 5 credits
-> start another : 5 credits
-> start a topic from AI orchestration 5 credits (0 if done by deterministic way)
"classic answer" is just when you write a message previously in a topic and send it : no gen AI just old school chatbot.
So to calculate, it's for each use case in the agent : the sum of credit :
Often it's only a question + answer with genAI so up to 12 credits if on sharepoint
and start few action by itself 5 to 15 credits. (action manualy started from topic are no charge since the auto start of topic is already 5)
The real problem then is : ok you know how much a use case cost. Real life is about how many user, how many trigger system, how many time . if it's start from a trigger easy to calculate, you know how many. if start from human question, the difficulty is here. it depend so much of how the user are use to use AI, the use case etc ; in real life we make some "pilot test" to see how user use it.
Same goes for agent : you have an estimation of the use case cost with the table up of the message, need some real life practice to have more detail. But remember : AI cost is not like the cost of a power apps premium, you don't know exactly how much every month, it's an approx (like apps in azure)
If this helped you to solve you'r problem, it cool be awesome if you check this answer as verify and make it green :) it's important for the community and search engine :)