...for one automation scheme at a time.

But it makes sense from a software subscription and access control point of view, given the way enterprise apps have been configured and secured. If what you're doing is tapping into Salesforce, Workday, ServiceNow, etc....

Curious to know your thoughts.

So now with Common KG, MCP and agents, users should theoretically be able to do data-layer integration and delegate tasks to agents. But in some ways, it's still user-to-user and point-to-point, because the user associates a selection of data sources to integrate one by one...

My question then was, why do this at the application layer when the real work needs to be done at the data/knowledge layer in a graph?

...you could record a macro of sorts to automate the whole sequence.

Lots of enterprises tried to use RPA during the 2010s, and several startups such as UI Path were known for it. But my impression was that RPA generally was quite finicky, brittle and difficult to get to work....

RPA is a point-to-point means of application-layer process integration for end user task sequence automation. The theory was, if you were doing a repetitive process (such as a data entry process) that involved jumping into one enterprise app, then another, then another....

My initial take on Common KG and Dataverse agentic AI is that is a successor technology to what Microsoft tried to do with what we used to call Robotic Process Automation (RPA). Microsoft orginally called their RPA product Flow, then changed the name to Power Automate.

The video is illuminating--worth watching to see how the approach is designed to fit into existing workflows.

I'm doing more research on how the current MSFT approach fits into overall knowledge graph/Graph RAG implementation space.

"Customers including CSX, ESET, and Encino Energy are using enterprise Knowledge to build (or have plans to build) agents with Copilot Studio to improve business processes. "

More (including a video and recorded demo) at:
lnkd.in/gp6yZzuC

"Pattern 2: Indexes metadata to query. Examples include Azure SQL, ServiceNow, Salesforce, Zendesk, SharePoint List, Snowflake, Databricks, SAP (Hana), and Oracle.
Pattern 3: Read operations on Power Platform connectors. Examples include Azure AI Search and 1.4K+ Power Platform connectors.... "

so makers spend less time on indexing and integration to focus on building agents.

Pattern 1: Ingests data to create vector embeddings. Examples include Files, OneDrive, Dataverse, Salesforce & ServiceNow (unstructured data), Confluence, and Dynamics 365 (F&O v-tables) ...."

"Now it’s easier than ever to unify and operationalize knowledge across your agent ecosystem. The Common Knowledge Graph in Copilot Studio connects your agent to your enterprise systems with new enhanced connectors and simplifies RAG...."

What's unfair about charging 20% more for ind. cost reimbursement is that the star principal investigators who win the Federal grants--get a 20% smaller piece of the pie for actual research as a result. Which is why PIs like John Madey fought against univ. administrations.https://shorturl.at/iBPVH

Agree, but here's the nuance that Polk Award winners like Jeff Gottlieb get that most reporters don't. Big research universities always overcharge for research.

Overreimbursement of indirect costs has been a cash cow for them.

Political theater won't fix that (or any) problem. shorturl.at/AigPB