One thing we know about big tech platforms is that if they can, they will try and set the rules for the rest of us. It rarely looks threatening while it happens. The tools get better, the integrations get easier, and everything is simpler when you don’t have to make choices. Then one day the company that runs your tracking also owns your data, your reporting, and the AI making your day-to-day decisions. And then you notice it’s doing things to drive its own growth, not yours.

We have seen this pattern time and again in other sectors. Are we seeing it in partner marketing right now, as Impact.com partners with Rakuten to create a genuine industry giant? It’s too early to tell for sure, but what isn’t in doubt is that the choices that you as a brand manager or agency head make this year will decide how much control you have in the future.

It’s not the platforms’ fault. AI has changed expectations for every partner marketing team. They want more: personalization tuned to the individual customer, live visibility into every partner and every transaction, reporting that rebuilds itself on demand instead of landing a month late, and AI agents that act on all of it without waiting for a developer. The solution providers are scrambling to respond.

This is the lens through which we should analyze the actions of impact.com, Rakuten, Awin and others over the last month. The biggest news, Rakuten partnering with Impact, has sent ripples through the industry: two of the biggest affiliate networks bundling their tracking, attribution, managed services, consumer purchase data, and AI tools into a single stack their clients can join. 

The logic is straightforward: adopt AI deeply themselves, help their clients adopt it through the bundle, and become the ecosystem within which everything else in a brand’s partner marketing workflow operates. It’s what we would traditionally call a walled garden of integrated technology and services. What’s interesting is that it stands in stark contrast to how the independent infrastructure platforms such as Everflow and Tapfiliate are adapting to AI.

Everflow and others are making a bet that clients don’t want to be locked into one-size-fits-all AI tools, because each client’s business is unique and benefits from AI built around its own data, its own customers, and its own goals. Everflow has shipped MCP support in beta. Tapfiliate is operating an MCP server in production. TUNE has been quieter publicly but maintains the structural argument for platform-services separation. Why be locked into a system when AI will build you a personalized solution for every single customer?

Walled garden vs independent infrastructure: both are reasonable approaches. But there are big questions for clients and agencies: are we going to be locked in? Are we ready to go beyond a bolt-on AI chatbot to develop real personalization? And how do we get through the next three years as the AI tidal wave hits?

Around 2012, Salesforce stopped competing on CRM dashboard features and started becoming the ecosystem that every other piece of B2B sales software plugged into. The companies that figured this out and built on top of Salesforce did fine. The companies still arguing about dashboard features did not. Marketing automation went the same way a couple of years later. Productivity tools are in the middle of it right now. The pattern is the same every time: at some point in a category’s maturation, the question stops being who has the best product and starts being who owns the ecosystem that everyone else has to plug into.

Partner marketing just got there. Two of the longest-standing affiliate networks now operate as a single stack with AI tools, consumer data, and managed services rolled together. AppDirect closed its acquisition of PartnerStack two weeks before the Rakuten/Impact news landed. Awin had hit the deadline cycle of its tracking-standards play three weeks before that. Each move is a separate bet on what AI demands of partner marketing platforms. Each is also, structurally, a play for the ecosystem position.

For every brand running a Rakuten program right now, the next twelve months are decision time. Three paths sit on the table. Stay with the bundled model that Impact and Rakuten are building together. Move to an independent agency working on Impact’s platform under a separate contract. Migrate to a different platform entirely. Each comes with costs that compound over the next two renewal cycles, and AI is not waiting for anyone to deliberate. Whichever architecture gets committed to between now and the end of 2026 is, in practice, the architecture that runs partner marketing programs through the rest of the decade.

The case the bundle is making

From Impact and Rakuten’s vantage point, the alliance makes the kind of sense that strategy consultants get paid to articulate. David Yovanno, Impact’s CEO, framed it in mid-May:

“Partnerships are fundamentally reshaping how businesses grow, moving from a supporting channel to a primary driver of performance and customer connection. As this shift accelerates, brands need more than fragmented tools. They need a unified approach that brings together technology, data, and partner relationships.”

That is the consolidation argument as plain as it gets. It says that brands are tired of stitching together a tracking platform, a managed-services agency, an attribution vendor, an AI tool, and four reporting dashboards. The bundle promises to do that work for them. One contract, one accountability point, integrated data flowing across functions, AI deployed at the platform level rather than bolted on at the brand level.

The case is sharper still on AI specifically. Impact has launched “ask impact,” an AI assistant embedded inside its platform. Rakuten launched Mirai on May 6, marketed as the first advanced AI optimization agent for affiliate. Awin rolled out Ava across more than twenty countries earlier this year. Three consolidating operators, three bundled AI tools that live inside the platform, trained on the platform’s data, available to the brand without integration work. For a brand whose marketing operations team is six people and whose AI strategy is “we’d like to have one,” that is a real proposition. The bundle does the integration so the brand does not have to.

The Rakuten Rewards consumer data layer makes the proposition stronger again. Tens of millions of US shoppers, transactions tied to identifiable behavior, the kind of first-party data that competitors can model but cannot easily replicate. Adding that to Impact’s tracking infrastructure gives the alliance an attribution case that very few platforms can match. For a mid-market brand evaluating partner marketing platforms, this is the bundle saying: we will solve the data fragmentation problem and the AI problem in one move, and you can spend your engineering hours somewhere else.

For plenty of buyers, that’s the right answer.

What the alliance means for agencies and OPMs

For agencies and independent OPMs, the calculation looks different.

The structural advantage of the independent agency has not changed. Undiluted focus on the client’s program, no network revenue interest pulling the recommendation off-mission, the experience and relationships that come from running programs across multiple platforms. Greg Hoffman, whose firm Apogee has been managing affiliate programs since the early 2010s, made the point publicly in late April:

“All of the audits we have performed show that network managers favor network-owned or favored properties, such as cash-back or loyalty sites, over the middle tier of content-based partners. Brands pay the cost in partner mix, in funnel coverage, and in long-term program health.”

That is the agency model’s defensible ground. It survives the alliance.

The new pressure on agencies is sharper, though, and it’s not really about competition for management contracts. The concern being said in private agency channels, more directly than in public, is about visibility. With Rakuten’s business development arm sitting structurally adjacent to Impact’s platform, and Rakuten Advertising holding a Titanium Partner tier on that platform, the question every agency principal is now asking is whether Rakuten’s BD team will treat its proximity to Impact’s client base as implicit permission to prospect.

Impact and Rakuten have not addressed this question publicly. The Rakuten Advertising blog post from May 18, the most detailed brand-facing communication so far, states that agency relationships “stay intact” but does not explain how they are protected. That answer is something every agency client should be asking their account team for, in writing, before the next renewal cycle.

There’s also a harder concern worth naming. As bundled AI tools mature inside the platform, some of what an experienced affiliate manager does today (strategic recommendations, partner mix optimization, contextual judgment about which content publishers are worth investing in) becomes addressable by AI agents trained on platform data. In the two- to three-year window, this isn’t yet true. In the five-year window, it’s genuinely an open question. Agencies and OPMs that aren’t already thinking about which of their capabilities are durable against agentic AI will be playing catch-up by 2028.

The first question for brands: what is the layer actually for?

For brands, two questions matter more than the rest. The first is about what the bundle’s intermediary layer actually contributes.

AI gives a brand, for the first time, the genuine prospect of personalizing every customer interaction. The technology exists. The integration patterns are settling. The cost is dropping. A brand that connects its CRM, its affiliate platform, its analytics, and its billing data through MCP or similar protocols can have its own AI agent build personalized partner communications, generate publisher-specific reporting, identify lifetime-value patterns, and execute optimization actions, all without a developer in the loop and increasingly without a managed-services team in the loop either.

Given all that, the brand has to ask what the platform and the agency are actually doing for it that the AI can’t. Some of the answer is obvious and valuable. The platform owns the tracking infrastructure and the partner contracting. The agency brings vertical and relationship expertise that machines don’t replicate well. Some of the answer is less obvious. When the platform’s bundled AI mediates between the brand and its data, the brand has to ask whether “intermediary” is a synonym for value or for additional layer.

The honest answer depends on what the brand is trying to do. If the brand is going to run partner marketing the way most brands run it today (set up the program, trust the network’s recommendations, optimize quarterly), the bundled AI is fine. If the brand is going to run partner marketing the way the brands ahead of the curve are starting to run it (continuous AI-driven personalization, daily cross-channel attribution reporting, agent-driven optimization across publishers), the bundled AI starts to look like a constraint, because it’s the platform’s AI, not the brand’s, and it sees only what the platform sees.

The second question for brands: what does the walled garden contribute over time?

The second brand question follows from the first. If the brand is doing the AI work itself, what does a walled garden contribute in the medium term?

In the short term, the bundled platform contributes a lot. Years of architecture already built. Tracking infrastructure that works. Partner relationships in place. Reporting that doesn’t need re-creating. The brand that buys the bundle gets a running start.

In the longer term, the calculus changes. AI agents that can connect a brand’s data silos through MCP can stitch CRM, billing, ad platform, and affiliate data into a unified performance picture without help from the platform. Once that’s in place, the only real value the bundle has left is the data nobody else can get. Rakuten Rewards’ consumer purchase data is the strongest example: behavioral data on tens of millions of US shoppers that competitors cannot easily replicate. Any brand whose category benefits from that data should weigh it carefully.

Beyond that, the bundle’s contribution narrows. Tracking can be done by neutral platforms. Reporting on attribution can be done by the brand’s own agents. AI optimization can be done by the model the brand chooses. Partner contracting can be done by any platform with a workable contracting layer. The bundle’s case rests on those services staying difficult enough to do independently that the bundling premium is worth paying.

Several platforms are building toward making them easier. Everflow has published llms.txt and shipped MCP support in beta, the two pieces of infrastructure that let external AI agents read its data and call its tools directly. Tapfiliate has an MCP server in production. TUNE has been quieter publicly but maintains the structural argument for platform-services separation. None of these platforms has the scale or the managed-services breadth that Impact, Rakuten, Awin, or CJ Affiliate offer today. The argument for them is architectural: they’re building toward the world in which the brand’s AI does the work, and they aren’t building anything that prevents the brand’s AI from doing it.

CJ Affiliate is the case worth noting because it doesn’t fit either side of the binary cleanly. Structurally CJ is a bundle player: network plus platform plus services. But CJ has kept a more open API posture than its bundle peers. CJ runs a public developer portal with GraphQL access; third-party MCP servers built on CJ’s APIs already exist in production. CJ has also pursued a different strategic direction from the Rakuten and Impact alliance, with no vertical-integration acquisition or alliance, focusing instead on expansion into commerce-media surfaces like CTV, podcasts, and embedded checkout through partnerships with companies like firmly.ai and tvScientific. CJ is the data point that the bundle-versus-open-infrastructure question is not binary, and not the only question a buyer should ask.

Short-term and long-term answer to different people

The short-term and long-term tension splits along predictable lines. The CFO who wants a predictable cost and faster deployment answers walled garden. The CTO who wants the brand’s data infrastructure to stay flexible, and the CMO who wants ownership of the customer relationship, answer open infrastructure. The competing answers map cleanly onto who pays the bill versus who owns the outcome.

This isn’t a tidy split. For a mid-market brand with thin marketing operations and no internal data team, the CFO’s answer is probably the right operational answer. The bundle’s faster deployment matters more than long-term flexibility because the brand can’t execute long-term flexibility anyway. 

For an enterprise brand with multiple agencies, a data team, and an in-flight AI strategy, the calculation flips. The brand has the operational sophistication to keep the layers separate. It has the agency relationships to defend the agency channel. It has the data team to run its own attribution. The right answer here may be to build on open infrastructure, even if the upfront cost is higher and the deployment is slower, because the freedom to switch later compounds over the next five years.

For agencies and OPMs, the structural argument is simpler. A platform that competes with you for managed-services revenue isn’t the platform you want to recommend to your client. The independent platforms that are explicit about platform-services separation are the ones an agency can recommend without conflict. Agencies that don’t make that distinction explicit to their clients now will be having harder conversations two years from now about why they didn’t.