Carl Yeh

One thing you could do is just build a data lake.

You can make it super simple, use Google BigQuery, and have it pull the data from...

all these different places.

And as long as they have an API, you can do that.

And then, then you put your MCP on top of that data lake.

And that's the key.

Cause I think like one of our clients, we've already done that where they have so many different tools.

You put a data lake in there and they can use their chat GPT to be like, Hey, what did this in, in Sage, we do this, this, we could do this.

And you're not doing any work on the,

the SAS tool itself you don't have to mess around with the files in those places so that you operate as usual but now your AI sits on top of a lake that is pulling in some cases real time and that just makes it until whatever happens with the SAS tool happens but at the very least and it's a much reasonable investment in time and resources than trying to do the other way around

The question is, how does Cloud Cowork fit into it?

Oh, perfect.

Claude Cowork has MCP, right?

So all you need to do, if you have a data lake, and Google has made this even easier because they have a native MCP for BigQuery.

So you just tap into that MCP into your instance of BigQuery with your data pulling from various places.

And you just go to Claude Cowork and it's part of the task.

Hey, I want you to go pull...

from whatever their name of the data lake however you partitioned it hey i want you to go to sage do this go into bid to win do this now i also want you to go into my google drive and do this like you can list out all these tasks and it's like a true agent it's like this this this this this this this execute and if you're really you want to throw a fun into that and

add like a Ralph Wiggum loop into there and be like, Hey, I'm going to go to sleep and, or I'm going to take a nap.

And when I'm done, you brute force your way into whatever you have.