The dashboard may be becoming the wrong place to start when working with enterprise software. Capacity's recent AI Analytics Assistant lets CX leaders query operational data by asking a question rather than opening a report. Salesforce has pushed the idea further with what it calls a headless enterprise strategy: agents that act directly on enterprise capabilities, without ever opening the application. Tealium's latest announcement takes that further still. This time, the agent isn't just querying the customer data platform. It's helping configure it.
Configuration MCP: agents become platform operators
The centrepiece is Tealium Configuration MCP. It connects AI clients to a server-side profile, allowing users to explore configuration and propose changes using natural language rather than the Tealium UI.
In practice, a user no longer has to open Tealium and work through its configuration screens. They can ask an AI agent to explain how a profile is set up, build a new audience, or flag configuration that's no longer being used. According to Tealium's documentation, the agent can also trace dependencies between attributes, audiences and other entities, so a user can ask what else relies on something before changing it.
This changes the role of the agent. It's no longer just answering questions about the platform. It's helping operate the platform itself. Tealium isn't handing the keys to the platform to an autonomous agent, however. Changes stay in draft, and a human has to review and publish them through the Tealium UI before anything goes live.
Tealium says the approach can work with Claude, ChatGPT, Gemini and other MCP-compatible systems, although current Configuration MCP documentation lists Claude as the supported AI client, via claude.ai or Claude Desktop, on a paid plan. That distinction is worth noting as the company expands the technology.
From customer context to customer-data infrastructure
The new capability builds on Tealium Context API, launched in June. This gave AI agents a managed MCP server for securely querying live customer context, including profile state, audience details and consent, in real time. Context API lets agents use customer context. Configuration MCP lets them work with the platform configurations that organise and activate that data.
The interesting part isn't that Tealium has added another AI interface. It's that the interface is starting to sit between the user and the platform itself, which increases the importance of the data, the APIs and the permissions sitting beneath it.
Other announcements broaden the strategy
Tealium paired Configuration MCP with three related releases. Tealium Studio adds a natural-language interface for explaining data flows, visualising audiences and proposing configuration changes. New Platform APIs, released through a self-service Developer Portal, give developers versioned and validated ways to operate Tealium programmatically and reverse changes when needed. And AI Recommended Audiences surfaces segments automatically from data already unified in Tealium, letting teams move from insight to activation without SQL.
Why it matters for CX
CX platforms were built around humans navigating applications. This announcement, alongside comparable moves from Salesforce and Capacity, suggests a shift from dashboards to conversational interfaces, agent-operated platforms and, increasingly, agent-configured infrastructure. The question for enterprise buyers is no longer simply whether AI can access customer data. It's how much of the systems that decide how that data can be used, and where it can go, they're willing to let AI operate.
For now, Tealium's answer is to let the agent do the work, but make a human approve the result before anything goes live.