Traditional marketing had a good run, but it has long passed its expiration date. For years, it has centered on the campaign, with marketers identifying the desired customer action, finding or creating an offer, developing content, selecting an audience, and writing the rules that connect them. AI can make many of these steps faster, but when it is applied to individual tasks, it still leaves the basic process intact. Marketers must still go to one system for the offer, another for the content, and a third for the audience and then stitch the pieces together by hand.
Many companies have tried to solve this fragmentation by consolidating their marketing technology and assets into fewer systems. This can simplify the existing process, but it does not fundamentally change it—and consolidation brings limitations of its own, including greater technology complexity and dependence on individual platforms. There is a better option, one that enables AI agents to assemble marketing interactions rather than simply improve the individual steps marketers use to create them. We call it the Composable Shelf. For our clients across retail, travel, banking, and healthcare services, with this approach, BCG has consistently seen three times the ROI and ten times the speed from when the marketing campaign was initiated to implementation.
Consider a bank trying to persuade customers to enable autopay. Today, a marketer might set that objective in a brief, work with another team to create an offer—say, 1,000 points for signing up—find or create the email and image, get the content approved, define the audience, and manually bring the pieces together in the email platform. Increasingly, AI is used for some of these tasks, from generating content to selecting audiences and optimizing decisions. But the underlying workflow remains largely the same.
The Composable Shelf eliminates the shortcomings in this process by adding a new layer to the existing marketing technology. Through it, an AI agent can understand the actions the business wants customers to take, the offers available to encourage these actions, and the content that can communicate them. The marketing assets remain where they are; what the shelf adds is a record of what each piece is for and what it can be combined with based on what rules. Marketers establish the objectives, ingredients, and guardrails. An agentic intelligence layer can then use the shelf to select and assemble the appropriate combination for each customer at the right moment. This requires more than adding a general purpose AI tool to the existing process; the technology and information around it must be organized to support agentic decision-making.
The difference is fundamental. Instead of manually building campaigns, marketers curate the Composable Shelf from which campaigns—or individual interactions—can be created. The work of assembly and deployment shifts to the agent. (See Exhibit 1.)
How the Composable Shelf Differs from Traditional Marketing
The change in the marketers’ role is as much about how marketers work as it is about technology. Curation replaces assembly. Instead of devoting much of their time to finding and stitching together the elements of individual campaigns, marketers focus on stocking and maintaining a shelf from which those interactions can be composed. The campaign is no longer the organizing principle of their work; the shelf is.
In the traditional model, the campaign plan was the deliverable, and assets were inputs to it. With a Composable Shelf, the shelf itself becomes the range of options available to the agent. An agent can’t deploy what it can’t see, and it can’t pursue an objective that has never been recorded. Marketers therefore retain the strategic decisions about what the business is trying to accomplish and the boundaries within which the agent can operate.
Stocking the shelf does not mean moving everything onto it. The Composable Shelf is a virtual reference layer that sits above the asset, content, and offer systems already in place. Automated pipelines pull references and existing tags from these sources; AI reconciles competing taxonomies into a common model and enriches incomplete metadata.
The shelf also adds a marketing campaign resource that typically does not exist today: an action library that records consumer behaviors the business wants. AI can propose relationships among those actions, offers, and content, with marketers confirming them. At delivery, the activation platform follows this reference and retrieves the content from the system where it already resides.
Today’s marketing systems generally record what something is (e.g., a product image) but not what it is for. An asset library may know a file’s format, rights, and expiration date, for example, but not the customer behavior it is intended to encourage. The shelf adds these missing relationships.
These connections are what turn a collection of repositories into a shelf. Three key components of marketing sit on it (see Exhibit 2):
- Actions describe what the business wants the customer to do—for example, make a first purchase, add a second product, or enroll in paperless billing.
- Content supplies the language and visuals used to communicate with the customer.
- Offers provide an incentive—such as a price discount—when one is appropriate.
Every action needs content, but it may or may not need an offer, and different content can be used with the same offer.
These links also enable agentic composition. In our earlier autopay example, the shelf could link the action—activating autopay—to the 1,000-point offer and to the approved emails and images that can be used to promote it. The agent can then select from those ingredients and assemble an interaction tailored to a particular customer, rather than choose from campaigns designed in advance for broader groups.
Consolidation Alone Is Not the Answer
For more than a decade, consolidation has been the response to the fragmentation and inefficiency of traditional marketing: bring assets into a common repository, connect that repository to downstream systems, and establish common metadata, brand, legal, and rights rules. This strategy can drive incremental improvement on individual tasks, but it does not change the underlying model. Marketers still must connect content with offers and desired customer actions and then assemble those elements into campaigns.
A closer look reveals three shortcomings to using consolidation alone to address market fragmentation (see Exhibit 3):
- It can never be finished. Because the systems and content being consolidated continue to change during a lengthy migration, the single clean repository it is trying to create remains a moving target.
- It can never be comprehensive. Even a clean-slate approach that eliminates the legacy problem faces the same design limitation. An asset management system holds creative; it does not inherently hold offers and actions or connect all three. Starting over may produce a cleaner system, but not a comprehensive marketing model.
- It can’t keep pace. Because AI continuously generates content, the volume of material grows too fast to be effectively migrated and organized.
Agentic AI introduces yet another challenge for consolidation. Consolidation was designed to make assets easier for people to manage and retrieve, not to give an agent the purpose and relationships it needs to assemble individualized customer interactions. Even with agentic AI, consolidation tells marketers what information they have, not what it is for. Moving a file into a common repository does not record the customer action it is intended to encourage or its relationship to a particular offer.
Consolidation can also increase the risk of vendor lock-in. The more of the marketing operation that resides in a single platform, the more the company must work within that vendor's roadmap, metadata model, and release schedule. This can make it harder to change suppliers or introduce new technology as needs evolve.
So far, every fix using consolidation alone has started by moving something. A Composable Shelf starts by moving nothing. It leaves content and offers where they reside and adds the actions the business wants customers to take. It gives the agent a view of the available marketing options, what each is for, and how they can be combined to target a customer’s needs. With the Composable Shelf, consolidation remains relevant, but the two concepts are self-reinforcing: consolidation can simplify the existing systems, while the Composable Shelf adds the layer needed for agentic activation.
What Makes the Shelf Composable
The Composable Shelf is not tied to a single marketing platform. It creates a common catalog that different decision and activation engines can use, while the underlying content and offers remain in their existing systems. If one of those systems changes, the shelf can point to its replacement without having to be rebuilt.
This makes the Composable Shelf’s architecture both flexible and reversible. Companies can replace individual platforms without losing the taxonomy, relationships, and records of purpose that have accumulated in the shelf. In that sense, the Composable Shelf is the inverse of vendor lock-in: the software and platforms that hold the content and offers can change, while the marketing intelligence captured in the shelf remains intact.
Composability also does not mean that every interaction has to be broken into its smallest pieces and assembled from scratch. The shelf can hold ingredients at three levels, with all three tagged in the same way and available to the agent:
- Individual reusable building blocks, such as offers, headlines, and images
- Complete touchpoints, such as an approved email
- Designed sequences, such as those used for a product launch, onboarding welcome, or loyalty milestone
Not every interaction benefits from being decomposed. Some interactions may work best as a complete experience or sequence. These can remain intact, while other interactions can be created from individual ingredients and tailored to a particular customer.
This flexibility also reduces the disruption for marketing teams. Content teams can retain their tools, creative processes, release schedules, and approval chains. The difference is in the last-mile decision about which approved content, offer, or experience reaches which customer—a decision the agent can make at a scale that would be difficult to prescribe in advance.
What Using the Composable Shelf Means for Marketing Leaders
Building the technology behind the Composable Shelf is the easier part; the organizational tasks inherent in the Composable Shelf are more challenging. These tasks include agreeing on what the business wants customers to do, writing the rules before anything is assembled, and letting an agent make decisions that marketers have traditionally viewed as their responsibilities. Five moves can help marketing organizations make that transition:
- Run the test, then make it the pilot. Pick one customer behavior the business wants to influence and ask the team to find everything available to support it: the offers that could encourage it, the content that could support and promote those offers in each channel, and the legal and brand constraints that apply. Record how long the exercise takes, how many people are involved, and whether the team can be confident it has found everything relevant. The results will reveal how difficult it is today to assemble a complete picture of the marketing options available—and where the gaps are. Those gaps matter because an agent can only work with what it can find and understand. The company can then turn this test case into its first shelf pilot, with much of the necessary groundwork already done and the gaps clearly identified. Keep any consolidation program already underway. The two can work together, and the shelf doesn’t have to wait for consolidation to finish.
- Give the action library an owner. Offers and content already reside in established systems. The actions the business wants customers to take usually do not. They live in briefs, planning decks, and spreadsheets, making them difficult to consistently identify, govern, and reuse. The action library is the one genuinely new asset the shelf requires, and it needs a senior marketing owner who decides which behaviors matter, how they rank, and what objective each serves. Treat it as a marketing asset, not an IT one.
- Bring legal and brand teams in to write the rules, not review the output. The shelf must tell the AI agent not only which content, offers, and actions belong together, but also the rules governing when and where they can be used. That means marketing, content, legal, and brand teams need to agree up front on those relationships and on constraints involving such things as region, rights, embargoes, and eligibility. A cross-functional group should be given responsibility for making those decisions and resolving disagreements. Once those rules are established, the agent can apply them each time it assembles an interaction rather than requiring the completed interaction to go through the same approval process.
- Decide what the agent may do on its own, and widen the scope deliberately. Once the shelf is stocked, the agent can assemble interactions that have never existed before. That is the point, and it is also the risk leaders need to govern. Start by letting the agent compose freely only from preapproved ingredients for always-on, lower-stakes actions. Leave designed sequences, such as a launch or an onboarding welcome, intact and let the agent choose only when to trigger them. Where regulation or brand exposure requires it, keep human approval in the workflow. Expand the agent’s latitude as the results come in, not before.
- Change the job, then change how success is measured. The shelf changes the marketer’s role from assembling campaigns to determining what the business wants customers to do and giving agents the approved offers, content, and rules they need to pursue those objectives. That shift also requires a different way of measuring marketing performance. Rather than focusing primarily on how many campaigns are launched and how they perform, marketing leaders should also assess whether the shelf contains the range of current, approved content and offers needed to support the business’s priority customer actions—and how effectively those individual ingredients perform when agents use them. Otherwise, marketers may be given a new role while continuing to be measured largely on the work the old model required.
None of these moves requires choosing a new platform first, and that is deliberate. Each builds something the organization keeps, whichever engine or vendor it uses next.
Agentic marketing does not require companies to discard the technology, content, or creative processes they already have. But it does require a different way to make those resources available to AI. The Composable Shelf provides that layer by recording not simply what marketing ingredients exist, but what they are for, how they can work together, and the rules governing their use. An intelligence engine can then use the shelf to select and assemble the appropriate interaction for a particular customer.
That is why the place to start is the shelf rather than the engine. The shelf establishes the marketing options available to the business and the relationships and rules that govern how they can be used. The intelligence engine that uses the shelf may be replaced, and so may the vendor that provides it; the foundation on which those decisions are made does not have to change with them.
