By Michael Fuchs, Head of Strategy & Planning, Data & Technology  

Why Marketing and Advertising Need New Models for Operational Intelligence Ownership

The AI conversation in marketing has a blind spot.

Organizations are moving fast to build agentic systems that accelerate planning, optimize media, automate execution, and compress timelines. That focus is understandable, as the industry is still in the early phases of operationalizing AI at scale.

But beneath those conversations, a more important strategic question is beginning to emerge:

Where does learning actually live? Where learning accumulates determines who holds the operational advantage in the next campaign, the next creative, and the next audience. It shapes who can optimize faster, and ultimately who controls the intelligence driving growth decisions.

Where Learning Actually Lives

Historically, most discussions around AI learning have focused primarily on the models themselves — training, refinement, retrieval, and improvement. In many ways, the industry has treated intelligence as something largely contained within the model layer.

But as agentic systems move deeper into operational marketing environments, forms of learning increasingly begin accumulating outside the models themselves: within memory systems, workflow histories, optimization patterns, approval structures, orchestration layers, campaign learnings, and persistent organizational context.

Over time, these systems begin accumulating forms of institutional and operational learning shaped by how organizations plan, optimize, approve, measure, and adapt decisions. That accumulated learning is not the same thing as data, workflows, or models themselves.

Data may be portable. Models may be replaceable. Workflows may be interoperable.

But organizational learning is considerably harder to disentangle.

It’s worth noting that the ability to meaningfully, disentangle, or transfer accumulated organizational learning across AI systems remains technically immature, but the strategic stakes are forming faster than the solutions.

The Governance Gap

The marketing and advertising industry already has relatively mature frameworks around data ownership, media transparency, identity governance, privacy, and interoperability.

It also has long-established norms around how to think about employee experience, agency know-how, strategic expertise, institutional memory, and the movement of talent between companies.

The industry already knows how to govern data and assets, but it doesn’t know how to govern learning. As AI systems move deeper into the operational core of marketing organizations, that gap will become increasingly important, partly because modern marketing systems are no longer isolated execution environments. They increasingly influence interconnected decisions across the entire marketing operation from audience strategy to creative iteration and media allocation.

The issue is no longer simply:
“Are you training your model with my data?”

Increasingly, organizations need to ask:
“What forms of operational learning are developing around these systems, who governs them, and how portable are they?”

This distinction matters particularly in marketing because the industry operates through deeply interconnected operational ecosystems. Agencies, advertisers, cloud providers, media platforms, commerce systems, identity environments, and measurement frameworks increasingly operate across shared workflows and decision systems.

That doesn’t necessarily imply malicious intent or strategic capture. In many cases, shared operational learning creates enormous value. Agencies need persistent optimization systems, accumulated workflow intelligence, and orchestration capabilities in order to compete effectively in more compressed AI-mediated environments.

But advertisers may simultaneously become more sensitive to a different set of questions:

• What intelligence remains sovereign?
• What becomes shared?
• What remains portable?
• What operational learning persists after a relationship ends?
• What forms of accumulated optimization intelligence belong to whom?

The Categories of Learning

The challenge is that not all accumulated learning behaves the same way. Some forms are deeply proprietary and organization-specific, other forms are created collaboratively across operational relationships. Still others derive value precisely because they are generalized across broader ecosystems.

The industry likely needs more nuanced frameworks for distinguishing between these different categories. At a high level, four are beginning to take shape.

1. Sovereign Intelligence

Sovereign intelligence is accumulated learning that belongs primarily to a single organization such as a brand and reflects their proprietary institutional context.

This includes things such as:

• Brand audience methodologies
• Their internal planning heuristics
• Their budget allocation logic
• Organizational approval structures
• Optimization approaches unique to the client
• A brand’s historical performance
• A brand’s internal workflow patterns

As AI systems become more embedded inside marketing operations, these forms of accumulated learning may increasingly become part of a brand’s long-term strategic differentiation.

2. Shared Intelligence

Shared intelligence refers to operational learning created collaboratively across organizational boundaries between the Client and the Agency.

In marketing and advertising, this may include:

• Advertiser-agency workflow coordination
• Collaborative campaign optimization
• Shared orchestration systems
• Integrated measurement and reporting processes

Many of the benefits promised by agentic systems depend on intelligence being able to accumulate across interconnected environments rather than remaining isolated within organizational silos.

But this also creates governance ambiguity.

As learning becomes embedded inside shared operational systems, the boundaries surrounding ownership, portability, and long-term control will become increasingly difficult to define cleanly.

3. Ecosystem Intelligence

Ecosystem intelligence refers to accumulated operational learning that emerges through participation in a broader ecosystem rather than from any single organization or relationship.

This can show up as:

• Category-level optimization patterns
• Generalized workflow heuristics
• Activation best practices
• Cross-client operational insights
• Measurement methodologies
• Organizational operating patterns
• Ecosystem-wide performance trends

This category is particularly important in marketing and advertising because it represents a form of institutional learning that has historically been one of the industry's most valuable assets.

Agencies have always accumulated knowledge through repeated participation across hundreds or thousands of engagements. Over time, they develop pattern recognition around what works, what fails, how organizations successfully operationalize change, which optimization approaches consistently outperform alternatives, and how emerging channels and platforms evolve over time.

Historically, this learning largely resided in people, playbooks, methodologies, and institutional experience.

In persistent AI systems, that same learning becomes embedded in the system rather than the people. An important distinction of ecosystem intelligence is that it doesn’t belong exclusively to a single advertiser and isn’t necessarily created through a specific collaborative relationship. Instead, it derives its value from aggregate experience, broad pattern recognition, and accumulated operational learning across many environments. While sovereign intelligence belongs to the advertiser and shared intelligence may be jointly created, ecosystem intelligence helps improve performance but isn’t “owned”.

That dynamic is not fundamentally new. Marketing and advertising have always benefited from aggregated institutional learning. What changes in persistent AI environments is the scale, persistence, speed, and operational embedding through which that learning accumulates.

4. Portable Intelligence

Portable intelligence refers not simply to moving data or workflows between systems, but preserving the continuity, governance, and operational utility of accumulated organizational learning across environments over time.

That is considerably more difficult.

Portable workflows are relatively straightforward. Portable accumulated learning may be much harder because institutional learning becomes deeply contextual over time, shaped by persistent interaction across workflows, approvals, optimization histories, organizational structures, and operational environments. As those systems mature, disentangling what belongs to who is becoming increasingly complex.

The industry already understands how to port data, content, and even infrastructure. It does not yet have equivalent frameworks for porting accumulated organizational learning.

As AI systems become more operationally embedded across marketing ecosystems, organizations will increasingly require architectures capable of distinguishing between these different types of intelligence to shape how decisions, optimizations, workflows, and institutional knowledge evolve over time. The organizations that understand how to govern, partition, preserve, and coordinate that learning will ultimately shape the operational architecture of modern marketing more than the models themselves.