While everyone chases better data and smarter AI, ad tech's actual constraint is that platforms are held together with brittle, hand-built integrations that break constantly.
The ad tech and martech industries have spent years obsessing over data strategy. First-party versus third-party. Identity resolution. Clean rooms. Privacy frameworks. These conversations dominate conference keynotes and executive roadmaps.
But while everyone's been focused on what data they have and how to use it, they've missed the real issue: platforms still can't talk to each other efficiently.
Connectivity is the homely little problem. And AI agents are about to expose just how critical that gap has become.
The Problem Nobody Wants to Talk About
Connectivity isn't exciting. Meanwhile, data strategy always makes the headlines and industry chatter. Identity resolution launches products, but connectivity is the ignorable engineering work that happens in the background. Until the system breaks, at least.
The typical enterprise marketing organization runs 80 to 120 different platforms—customer data platforms, demand-side platforms, analytics tools, measurement systems, creative platforms, and email systems. Each one speaks a different API language. Each one requires custom integration work. Each requires connectivity.
Building just one or two connectors is easy enough to handle. But needing 80-plus platforms to work together involves managing hundreds of platform-to-platform connections at any given time.
This is when things become unwieldy. And if just 10% of platforms push updates each month, you're dedicating an entire engineering team to maintenance.
What's more, that maintenance team isn't building new capabilities to optimize campaigns or develop better measurement. They're down in the basement, fixing broken pipes.
And every organization is doing this independently! The same Meta connector breaks for fifty different companies, and fifty different engineering teams stop what they're doing to fix it. It's inefficient at a scale that defies logic.
Why AI Makes Connectivity Critical
For years, marketers have lived with brittle, hand-built integrations because the alternative—not connecting systems at all—was worse. But AI agents are shifting the situation from a silent struggle to an alarming emergency.
As the agentic era becomes fully embedded in the media buying process, AI systems will continue to expose how fundamental managing connectivity has become.
When a human logs into a marketing platform, they're working through a user interface designed for manual input. They query data, configure campaigns, analyze results, and generate reports.
This interface represents a single connection point with the platform. An AI agent needs to replicate all of that functionality programmatically. It needs real-time, synchronous communication with the platform.
Ideally, the AI agent is having a conversation with the platform in much the same way a human does. This happens through APIs and Model Context Protocols, but it requires reliable, durable connections that don't break every time a platform pushes an update.
Here's where it gets complicated. Agents don't work with only one platform. Just as humans log into multiple systems to complete a task, agents need to orchestrate workflows across platforms.
An audience activation workflow might touch a CDP for segmentation, an identity system for resolution, a DSP for campaign setup, and measurement tools for tracking, all in sequence, all requiring coordination.
If these connections fail, the entire workflow fails. And unlike a human who can troubleshoot and work around a broken integration, an AI agent simply stops functioning.
The industry has been able to tolerate slow, fragile, flimsy integrations when humans were doing the work. But we can't afford that tolerance when AI agents are running the processes. Because agent failures are dead ends.
What the Publicis-LiveRamp Deal Signals
Publicis Groupe's $2.2 billion acquisition of LiveRamp in late May was a clear sign the market has recognized that connectivity infrastructure is strategically critical. You don't write a check that size for something you consider a commodity. You write it to build crucial infrastructure.
LiveRamp built enormous value by solving a massive challenge. The ad tech platform bills itself as helping companies onboard data, resolve identity, and activate audiences across the ecosystem. That centralized model created real efficiency for the market.
But the architectural requirements are shifting rapidly. Brands and agencies increasingly want to leverage best-in-breed technologies within their own cloud and data environments. They want to send and receive data securely across platforms without moving proprietary information to a centralized infrastructure. First-party data doesn't leave enterprise walls anymore.
The challenge isn't LiveRamp's output; that remains valuable. The problem is that the future market requires a different model that is cloud-native—deployed inside customer environments, interoperable, and built for real-time agent coordination rather than batch file transfers.
And there's a neutrality question that can't be ignored.
LiveRamp's value proposition was always built on being a platform-agnostic infrastructure everyone could use. Now it's owned by a holding company that competes directly with every other agency. Will WPP, Omnicom, and independent agencies continue writing checks to their biggest competitor to activate audiences?
This is a big question from a strategic and competitive standpoint, not a technological one.
What Connectivity Actually Requires
Here's what connectivity infrastructure needs to deliver in the agentic era.
- Pre-built, off-the-shelf connectors. Not custom engineering projects that take months to deploy—connections that can be configured and live within minutes, not months.
- Self-healing systems. When a platform pushes an update that breaks a connector, the system should detect it, identify the fix, and implement it before anyone notices. Observability and automated remediation can't be afterthoughts—they're core functionality.
- Zero-copy architecture. Data should stay within customer-controlled environments. What is required, then, is a deployed model where the connectivity infrastructure lives inside the enterprise's virtual private cloud, not a centralized processing model where data moves to external systems for enrichment and activation.
- Bidirectional, real-time communication. Not just batch file transfers. Agents need synchronous connections that allow them to query platforms, configure settings, and coordinate workflows dynamically.
- Governance and security at scale. As agents take on more autonomous decision-making, the connectivity layer needs built-in guardrails, permissions management, and audit trails.
These asks are the baseline for infrastructure that can support intelligent automation, not a "nice to have" list.
On-Demand Connectivity
When we talk to enterprise marketing leaders about connectivity, the response is usually the same: We know it's a problem, but we can live with it.
This was true when humans were managing the workflows. It's not true anymore.
The new standard needs to promise integrations that go live within minutes and stay live without constant engineering intervention. And this isn't some wishful prediction for some time in the future. This standard should be in place now.
Getting there requires treating connectivity as core infrastructure, not as a project. It means investing in systems designed for observability, scalability, and autonomous maintenance. It means understanding that when a platform updates and breaks 50 connections, fixing it once for all 50 is exponentially more efficient than fifty teams fixing it independently.
Most importantly, it means recognizing that the value isn't just in the data you have or the platforms you license. It's about whether those platforms can actually work together when you need them to.
The Uncomfortable Truth
You can have the most sophisticated data strategy in the world. You can have perfect identity resolution and pristine first-party data. You can license the best platforms money can buy. But if those platforms can't reliably share information and coordinate workflows, you're building on sand.
AI agents are going to make that painfully obvious. Soon.
Until your tech stack actually talks to itself, all the AI in the world won't move the needle. And this is the conversation the industry needs to have.
Because the bottleneck was never the data. It was always the plumbing.
More Resources on Martech
Martech Built Empires—Now It Needs Bridges
The Oz Paradigm: Why AI Still Needs a Human Behind the Curtain
