How to use an MCP to act on live GTM data (and the play that shows how)

MCP comes up in every other GTM conversation. Plenty of people can define it. Far fewer have connected one. When 50 senior GTM leaders were asked which AI use cases are working and which are overhyped, agents querying GTM data via MCP split them more than anything else in the survey: 52% were undecided, and only 16% said it was genuinely working.

In this episode of Revenue Architects, Florin Tatulea and John Lloyd define what MCP actually is and where it breaks. Think of it as a USB-C port for AI: one standard interface instead of a brittle custom integration for every platform. An API runs on fixed endpoints set in advance. MCP lets the agent decide what to call and grounds it in live data rather than whatever was last written to the CRM. Which is where the trust problem sits, because the protocol is only as good as the layer underneath it.

Then the play. Churn risk prevention, built as five prompts: pull the book of business filtered to accounts inside their renewal window, layer in pain points from recent call recordings, add competitive intent and review-site activity, weight it into a tiered risk score, then draft outreach to whatever lands in critical. Churn rarely announces itself in one system, so the query has to reach across four.

Florin and John also cover what makes it possible: resolved identity, so a fused risk score isn’t reading one account as four separate records, a grounding test for whether an agent can trace its output back to a real source, and the part nobody assigns, someone owning recalibration.