Unlocking AI agents for performance management
AI agents are finally becoming useful for finance teams and performance management. This video is a first look at how we arebuilding and deploying agents for clients at Openvale Group.
MCPs or model context protocols are a major unlock for AIagents. These are essentially the connectors between internal data sources andlarge language models that pass not only data but also context like accessroles to the LLM.
Power BI just released their MCP in November, so thesecapabilities are brand new but already moving at lightspeed. Teams finally havea path to build agents with leading models all within the Microsoft ecosystemwithout relying on out-of-the-box Copilots.
With the Power BI MCP, we can now connect our data andbusiness logic in Power BI with Microsoft Foundry, which allows us to build AIagents based on existing LLMs.
Let’s jump over to Microsoft Foundry to look at a sample AIagent.
This agent is connected to a Power BI semantic model and the GPT 5 Mini model. We’ve provided comprehensive instructions to the agent. We also have the ability to connect other knowledge sources and customize the agent further.
Once connected and configured, these agents can already consistently perform a wide range of actions. Here we see some prompts and responses for summary reporting, BvA reporting, drill-down, drill-through, and trend analysis. AI make insights more accessible than ever for decision-makers alongside decks, spreadsheets, and dashboards.
Here are a few key takeaways on realizing value from AIagents:
· Treat them like team members, not systems: AI agents are trained and instructed, not implemented. Business teams and especially leaders must be responsible forbuilding AI agents. IT should support but not own.
· No shortcuts: like any other reporting, AI agents are reliant on good upstream data and operational processes. There are no shortcuts to AI agents. Garbage in, garbage out still applies 100%.
· Keep it simple: AI agents can easily get overloaded. AI agents must be designed for specific tasks or reporting areas and connected together to be effective. Focus on solving the lowest hanging fruit before expanding scope.
If you’d like to learn more about our agentic performance management approach at OVG, reach out to us today!
