Measurement Joins the Chat

By Lindsay Fordham, SVP, Product, Cint

Marketing measurement has already moved beyond the days of waiting until a campaign ends to understand what worked. In-flight measurement gives marketers the ability to evaluate performance while campaigns are still running and adjust course in response to what they are seeing.

The next shift is about what happens once those signals are available. As AI agents become more deeply connected to the tools and data marketers use every day, measurement can become more directly embedded in the systems where decisions are made. Rather than simply surfacing an insight, agentic systems could help put that signal into context, recommend a response and, within clearly defined parameters, act on it.

Emerging standards such as Model Context Protocol (MCP) are helping make this possible by providing a common way for AI applications to interact with external systems and data. MCP-enabled workflows can make it easier to activate measurement studies, monitor in-flight performance, and bring relevant insights into the tools marketers are already using.

MCP also changes how those connections can be built. Instead of creating a separate custom integration every time an AI application needs to interact with a new system, MCP provides a more standardized way for agents to access available tools, data and context. In practice, that could allow an agent to initiate the appropriate study, retrieve the relevant measurement signal, and use it within the workflow where a decision is being made, without requiring a marketer to move manually between platforms.

Measurement Moves Into the Decision Loop

For measurement, this changes the role of the data itself. When an agent can incorporate a performance signal directly into the decision-making process, measurement becomes more tightly connected to what happens next. That also makes the quality of the underlying measurement more consequential: If an agent is using a signal to inform or recommend a decision, incomplete, misleading, or poorly understood data can quickly lead the system in the wrong direction.

Advertising has never lacked metrics. The harder problem has always been determining which metrics represent genuine business impact. A platform can report a conversion, and an attribution model can associate that conversion with an exposure, but neither can automatically prove that the advertising caused an incremental outcome. As AI becomes more involved in interpreting and responding to performance data, marketers will need to become even more disciplined about distinguishing correlation from causation. Full-funnel measurement connecting brand-lift and sales-lift data will only become more important in an agentic environment.

Faster Decisions Require Better Signals

Agentic measurement will also require systems to communicate uncertainty more effectively. Human marketers understand that an early read, a small sample, or an unusual spike may not justify a major strategy change. An autonomous system needs those distinctions explicitly built into the way it operates. Measurement infrastructure will increasingly need to communicate not only an answer, but the confidence around that answer, including data freshness, sample sufficiency, methodology, and known limitations. Otherwise, real-time optimization can quickly become real-time overreaction.

The industry will also need to avoid treating agentic capability as synonymous with unlimited autonomy. There is an important difference between an AI system observing a change in performance and recommending what to do about it than an AI system actually executing that recommendation on its own. Those functions should not automatically carry the same permissions. A marketer may be comfortable allowing an agent to continuously monitor pacing, brand lift, or conversion performance while still requiring human approval before it changes budgets, shifts investment between channels, or materially alters an audience strategy. As agents move closer to real advertising spend and consequential campaign decisions, clear governance becomes part of measurement infrastructure itself.

Interoperability Does Not Mean Consistency

Interoperability introduces another challenge. Making it easier for an AI agent to move between multiple systems does not resolve the advertising industry’s long-standing problem of those systems defining the same concepts differently. A conversion in one environment may not be equivalent to a conversion somewhere else. Attribution windows, identity frameworks, exposure methodologies, and underlying datasets can vary considerably. Giving an agent access to all of those sources does not automatically reconcile them.

Instead, marketers will need clearer rules about which signals are authoritative for particular decisions and how an agent should respond when sources disagree. The goal should not be to assume that greater connectivity creates a single source of truth, but to make the logic behind each decision more transparent and defensible.

The Dashboard Becomes a Checkpoint

This points toward a larger change in how measurement functions inside the advertising ecosystem. For years, the dashboard has been the destination: the place where data finally becomes visible to a person who can determine what to do with it. In an agentic environment, the dashboard increasingly becomes a checkpoint. Measurement can feed continuously into the systems helping make decisions, while humans establish the objectives, methodologies, confidence thresholds, and boundaries within which those systems operate.

That does not make measurement less important. It makes trustworthy measurement more foundational. The organizations that benefit most from agentic AI will be the ones that are clearest about which evidence an agent should trust, which decisions it can make, and where human judgment is irreplaceable.