If the Platform Now Has the Agents, What Are Clients Paying Us For?

May 1

This post builds on the analysis in McKinsey's April 2026 article "Reinventing marketing workflows with agentic AI" by Dianne Esber, Eli Stein, Julien Boudet, Kelsey Robinson, and Nilay Shah, with my own perspective on what it means for the marketing technology partner ecosystem.

Every CMO has heard the same uncomfortable statistic this year: nearly 90% of marketing leaders are running AI pilots, and fewer than 10% can show end-to-end value on the P&L. That gap — the "gen AI paradox," as McKinsey called it in their April 2026 piece on agentic marketing workflows — is the entire commercial opportunity for the next three years. It is also a problem most marketing technology partners are not currently structured to solve.

I have spent my career implementing marketing workflow solutions — Aprimo, Workfront, and the constellation of tools that orbit them. The "workflow" in marketing workflow software has, until very recently, meant deterministic routing: briefs moving from requestor to creative to legal to publisher along a defined path, with humans owning every decision and every handoff in between. That paradigm built a healthy services economy around it — discovery, configuration, approval-matrix design, taxonomy work, integrations into the DAM, the CMS, and the CRM. Agentic AI is now dismantling that paradigm, and it will take a chunk of the partner ecosystem with it unless we adapt.

For consulting and systems integration firms whose business model has been built on Adobe and Salesforce implementations, this is a moment of genuine reinvention. The platforms have moved. Adobe rolled out CX Enterprise and the Coworker offering in April. Salesforce has been pushing Agentforce and Einstein agents deep into Marketing Cloud and Data Cloud. The agents are no longer something a partner builds for a client — they ship in the box. The title above, in other words, is the uncomfortable question every Adobe and Salesforce partner now has to answer.

The honest answer is the part of the work that the platforms can't ship: the redesign of the workflow that the agents plug into.

The Pilot Trap, and Why It Persists

Most marketing AI work has stalled at the pilot stage not because the models are weak, but because they have been bolted onto workflows designed for a pre-AI era — siloed content management systems, disconnected digital asset libraries, CRMs with their own data model, analytics living somewhere else entirely. Drop a copy-generation agent into that environment and you get more drafts. Drop a personalization agent in and you get more variants. Activity goes up; the P&L doesn't move, because the workflow around the agent still has the same handoffs, approvals, and rework loops that consumed the time in the first place.

Marketing teams know this. McKinsey's recent CMO survey makes the point that brand and legal governance, human capability gaps, and data bottlenecks are what executives actually worry about — not the cleverness of the underlying model. Those are workflow and operating-model problems, not technology problems. And solving them is squarely in the wheelhouse of a good services firm.

The New Partner Mandate

The firms that pull ahead in this cycle will lead with workflow reinvention and treat the platform agents as components inside that larger redesign — not as the product they're selling. Concretely, that changes how engagements get scoped and staffed.

Discovery has to get more granular. The current-state mapping for an agentic redesign goes deeper than the typical implementation discovery. It is not enough to document "the campaign approval process" — you need the underlying micro-tasks, the systems they touch, the data each step depends on, and the human judgment calls that should never be automated away. The leading client examples in the McKinsey work involve breaking single workflows into hundreds of micro-tasks before deciding which should become agent work. That is a different kind of analyst output, and partners that haven't built reusable taxonomies and archetype libraries will end up reinventing them on every engagement.

Integration architecture moves from second chair to first. The most under-discussed point in the McKinsey piece: interoperability, not model capability, is what is actually limiting agentic deployments. Agents need to read from and write to identity systems, content repositories, journey orchestration tools, and analytics layers in real time. In the Adobe and Salesforce ecosystems, this means the unsexy work of unified identity, content metadata standards, shared data models, and API governance is the gating factor for everything else. Partners who treat this as the foundation will deliver. Partners who treat it as plumbing to be sorted out later will produce demos that fall over the moment they hit production volume.

Change management stops being an afterthought attached to the statement of work. Marketers will need to learn to prompt, supervise, and steer agents; to spot quality and compliance drift; to refine AI output with the kind of taste and cultural intuition that doesn't automate. None of that happens by handing out a one-page job aid at go-live. Partners with a real upskilling practice — curricula, role redesign, performance frameworks — will see their engagements stick. Partners without one will watch carefully built systems sit unused six months after launch.

Governance becomes a service line in its own right, not a checkbox. Marketing is consumer-facing, which makes the risk profile different from automating an accounts-payable workflow. Brand voice guardrails, legal review thresholds, confidence floors for AI-generated insights, escalation paths when an agent goes off-script — these need to be designed in, documented, and maintained. CMOs are visibly worried about brand and legal exposure, and they will pay for partners who can de-risk the rollout rather than waving at it.

How the Pitch Should Sound

A useful test: if a partner's pitch deck still leads with "we will deploy your Adobe Experience Platform" or "we will stand up Agentforce for marketing," it is a pitch from the last cycle. The pitch from this cycle leads with the operating-model redesign — what the marketing team will look like, how decisions will flow, where humans will sit in the loop, how the platform agents fit in — and treats the technology configuration as a means to that end.

The numbers are there to support the case. McKinsey's estimates put revenue uplift from hyperpersonalization in the 10-to-30% range, with campaign cycle times collapsing by an order of magnitude when agentic workflows are properly redesigned. Those are transformation-scale outcomes, not implementation-scale outcomes, and they justify a different kind of conversation with the CMO and CFO than the one a partner has been having for the last decade.

Don't Forget ServiceNow

There is a bigger orchestration question an Adobe-or-Salesforce-only frame misses entirely. ServiceNow has spent the last year aggressively positioning itself as the control plane for the agentic enterprise — what CEO Bill McDermott has been pitching, with characteristic restraint, as "while AI thinks, workflow acts." Their AI Agent Orchestrator, AI Agent Studio, AI Agent Fabric, and AI Control Tower are explicitly designed to manage and govern agents across the enterprise, including agents that ship inside other vendors' platforms. The recently announced Microsoft Agent 365 integration is a tell: ServiceNow wants to be the layer that monitors, governs, and orchestrates whatever agents an enterprise ends up running, regardless of where they came from.

If that vision lands even halfway, the Adobe agents and the Salesforce agents will not be operating in their own clean little marketing sandbox. They will be coordinating with HR agents, IT agents, customer service agents, and procurement agents, all under some kind of cross-platform orchestration and governance layer. For marketing partners who have spent a decade going deep on a single platform stack, that is uncomfortable. The most valuable expertise in the next cycle may turn out to be the ability to orchestrate across platforms, not within one — and the partners who recognize that early will be in a different conversation than the ones who keep treating ServiceNow as somebody else's problem.

This reinforces the central point. If the orchestration layer is moving above any single application, "implement the platform" engagements are even less defensible than they were a year ago. The work that retains value is workflow design, integration glue, governance, and operating-model change — exactly the muscles that anyone who has spent serious time inside Aprimo or Workfront has been building all along. The opportunity is to take that workflow-first instinct, which marketing operations veterans already have, and apply it across an agent ecosystem that no longer respects platform boundaries.

The Bigger Shift

Step back and what is happening to the Adobe and Salesforce partner ecosystem is a familiar pattern. When a platform absorbs the work that used to be a partner's bread and butter, the partner either moves up the stack or gets squeezed out. The agents in the box are the squeeze. Workflow reinvention, integration architecture, operating-model change, and governance are the move up.

The firms that get this right in the next 12 to 18 months will look less like systems integrators with an AI practice attached, and more like operating-model consultancies with deep platform fluency. That is a meaningful repositioning, and the firms that start it now will be the ones their clients call first when the next wave of agents ships — which, given the pace of the last six months, will probably be sometime around lunch.