Today, eighty-six million U.S. workers use AI at work, turbocharging a workforce wage base of $7 trillion to $8 trillion. That adoption signals a broader shift toward AI-powered workflows. Increasingly, we’re seeing that when customers need to update a project, approve an invoice, or close a ticket, they will ask Claude, ChatGPT, Gemini, or another agentic interface to handle it – rather than logging into a full platform. This shift makes it essential for customers to be able to access your capabilities without ever having to open up your product.
Why MCP is Necessary
Although AI features within a product are necessary, organizations need to also have an external-agent layer that makes their product the system of action that external agents must call when they want to perform a task. This removes the “toggle tax,” or the productivity cost when employees constantly switch between different apps, tabs, or digital tools.
AI can surface possibilities you didn't know were there. Because AI can explore the data and plan workflows already, it can find connections and answers that would have taken a human hours to uncover – or that simply wouldn't have been obvious to ask for. Comp admins often know what questions they have been asking; MCP opens the door to the ones they didn't know to ask.
Catch issues before they become problems. Because AI can scan across plan configurations, credits, and quota assignments conversationally, it can surface anomalies – misaligned splits, missing assignments, unexpected attainment gaps – that might otherwise go unnoticed until someone raises a dispute.
Security Concerns with MCP
As MCP servers become the action layer for agents, risks begin to surface.
Risk: AI could access more data than intended. Unlike a human who navigates to a specific screen, an AI can explore broadly across whatever it has access to.
How Xactly handles it: The policy system lets admins define exactly what any user, human or AI, is allowed to see at the data level. A rep's AI agent sees only that rep's data. Nothing more.
Risk: AI could make changes, not just read data. There's nothing inherently stopping an AI from writing, updating, or triggering workflows if it has the permissions to do so.
How Xactly handles it: Admins can configure AI access as read-only, explicitly blocking any mutations. However, admins can also allow AI to update the data if that’s the intended behavior.
Risk: It's hard to know what the AI actually did. Autonomous agents acting on your data create accountability questions that don't exist with human users.
How Xactly handles it: Every query made through MCP is captured in Xactly's audit trail, identical to any other user action. The full record of what the AI accessed and when is always available.
MCP is Fueling Xactly Intelligence Studio
Xactly Intelligence Studio lets enterprises build and deploy AI agents on their own terms – and Intelligence Connect is what joins those agents to the rest of the business. Rather than AI that lives in a silo, agents built in XI Studio become accessible across the enterprise tool stack, meaning a seller's unified AI experience can include Xactly's compensation data, plan logic, and benchmarking insights alongside every other tool they use.
Sales reps will be able to:
- Get accurate, real-time answers to commission questions without logging into Xactly
- Easily ask how a commission was calculated and receive a clear, sourced explanation
- Self-serve answers inside the tools they already use – no ticket, no wait
Compensation admins will be able to:
- Interrogate live Xactly data conversationally – without SQL expertise or a dedicated Xactly session
- Validate configurations, and surface anomalies through natural language
- Access compensation intelligence in the context where decisions are actually being made – rather than switching to a specialized system
