In most companies, nobody does. IT owns the infrastructure, marketing owns the richest behavioral data but not the platform it sits on, and no single executive owns the customer end to end. That gap is about to matter more than it ever has, because customer data platforms are moving into agent-driven data lakes governed by teams that do not have a marketing lens.
On this week's Revenue Marketing Raw, Jeff Pedowitz and Dr. Debbie Qaqish worked through what happens to marketing when the customer record permanently leaves the marketing stack.
Revenue Marketing Raw is a weekly unscripted B2B podcast hosted by Jeff Pedowitz, President and CEO of The Pedowitz Group, and Dr. Debbie Qaqish, Partner and Chief Strategy Officer. Debbie coined the term Revenue Marketing in 2010, and The Pedowitz Group formalized the category in 2012. Between them they have written six books, including Rise of the Revenue Marketer, From Backroom to Boardroom, The Revenue Marketer, and AI Agents Made Simple. Episodes run 20 to 30 minutes, publish on Tuesdays, and carry no guests and no sponsors.
Because it was never winning it, and the terms are getting worse.
Debbie framed the stakes directly: wherever data about the customer lives, if it does not live in marketing, marketing is treated as a second-class citizen for access to any of it. For roughly two decades, marketing automation gave marketing teams an unusually rich set of first-party behavioral data. That data is what allowed marketers to understand buyers rather than guess at them.
The question now is what marketing does when that record moves somewhere else. Not just the storage location. The governance, the access process, and the definition of what a customer even is.
The data lake was supposed to end silos. Snowflake, Azure, Salesforce Data Cloud, and now agent-native platforms all sell the same promise: one view of the customer across the enterprise.
In practice it worked partially and created new silos in the process. Jeff put it plainly: he cannot think of a single client where marketing owns the data lake or the warehouse. There is no CMO with a Snowflake administrator on the org chart. It is always IT.
And IT operates the way IT is supposed to operate. It has process. It has operational control. It has a queue.
Which means marketing has to file a request for anything. A segment. A list. A suppression rule. In a client kickoff this month with a financial institution, the answer to "how long does it take you to launch a campaign" started with a request to IT for the segment, and that request takes anywhere from a couple of days to a couple of weeks.
That is 2026. Not 2014.
Only if the governance model changes at the same time. The technology alone will not do it.
The optimistic case is real. If agents can handle enrichment, deduplication, governance, compliance, sourcing, and segmentation, then the quality and freshness of customer data should improve substantially. And logically, if agents make data easier to work with, access for anyone who needs it should improve too.
Debbie asked exactly that question. Jeff's answer was the uncomfortable one: you would think so.
Here is the failure mode. If the data layer gets dramatically more intelligent but marketing still has to follow the old request process, marketing ends up further behind than it was before. All that new capability accrues on one side of the wall while the marketer stands on the other side, filing tickets. The gap widens rather than closes.
The fix is not more technology. It is marketing having its own agents, operating under its own governance, against a shared record.
Which also suggests the lake metaphor has outlived its usefulness. A lake is a place you go. What the model actually needs to be is omnipresent: data that is instantly available wherever the decision is being made, by whoever is making it.
Nobody. And that is the deeper problem this conversation exposed.
Product marketers own a segment because they need to sell a product. Sales owns opportunities. Support owns tickets. Customer success owns renewals. In most large companies, there is no single throat to choke on the customer. Everybody holds a fragment.
This is why buyers still get bombarded by companies that own more customer data than any organization in history has ever held. The data exists. The ownership does not. So nobody is accountable for whether the experience makes sense from the buyer's side.
Nothing in the shift to agentic data platforms answers that question. It moves the data. It does not assign the owner.
Software pricing is running a full loop back to where it started. Enterprise seats and consumption came first, then SaaS made flat per-seat subscriptions the default, and now that nobody logs into anything anymore, AI and data vendors are moving back to token and consumption models.
Consumption pricing is sold as a benefit: you only pay for what you use. The problem is that almost nobody can forecast what they will use, because no predictable model exists yet in this market.
Now add the ownership question on top of it. If IT controls the consumption contract for the customer data layer, IT also controls the cap. A decision to hold monthly consumption to a fixed ceiling is a reasonable budget decision from where IT sits. From where marketing sits, it is a hard stop on the data needed to reach a customer.
That is a new category of budget fight, and marketing has historically lost this kind of fight.
The existing track record deserves attention before anyone assumes agents fix it.
Roughly 64 percent of deployed customer data platforms deliver significant value, and that figure has declined over time. Of the shortfall, 47 percent of martech decision makers point to poor system or data integration as a top hurdle, and 34 percent point to underskilled teams.
Debbie raised the sharper question: is integration even the right word anymore? If agents traverse systems and reconcile records, the old integration vocabulary may no longer describe the actual problem.
But the underlying issue survives the rename. If organizations have an adoption and skills problem now, more architectural complexity does not automatically resolve it. It can just as easily make the gap between capability and use wider than it already is.
The buying committee, and almost everything that happens off the record.
Agents run on the systems and data already present. In B2B, the systems present do not capture how decisions actually get made. There is no martech platform today that accurately tracks what happens across a full buying committee, and a large share of committee activity was never digital to begin with.
The sideline conversation after the meeting. The whiteboard session nobody photographed. The internal Slack thread on a channel no vendor can see. The senior stakeholder who never downloads content, never attends a webinar, and never talks to a large language model, but holds veto power over the purchase.
So when a platform promises that agents will unify customer data, the honest response is that they will unify the data that exists. In B2B, a meaningful part of the decision was never in the dataset. That is a gap in the architecture, not a gap in the effort.
The picture Debbie described is data that is rich, accurate, current to the microsecond, and available to everyone who needs it to do their job and create business impact.
The operational test is simpler than the architecture. Whatever part of the organization a customer touches, they are having one continuous conversation, not four disconnected ones. Marketing, sales, customer success, support, and executive communications all working from the same understanding of that person and that account, at the same moment.
That is the stage four state in the revenue marketing journey model Debbie built in 2010: repeatable, predictable, scalable revenue. The model has not changed. What has changed is that the data layer required to reach it is no longer something marketing controls on its own.
And the only thing that unifies a company holding this much data across this many owners is the customer lens. Applied from the top down, before the architecture decisions get made. Without it, more data and faster agents just produce a more expensive version of the same disconnected experience.
Not strategy decks. Competition.
The pattern has repeated for decades. Smaller, faster competitors take share. Growth stalls. The board replaces the CEO and part of the executive team. New leadership arrives with a mental model built for the current environment rather than the last one.
The expectation Jeff and Debbie landed on is that the next five years bring a significant wave of that turnover, driven by AI moving faster than most organizational structures can absorb. AI is already taking over the execution layer across sales, marketing, and service. What remains for humans is strategy, judgment, critical thinking, and insight, which means roles change, departments change, and the org chart itself changes.
Tomorrow begins today. The companies that treat the customer data question as an operating model decision rather than a technology purchase are the ones that get to the other side of it.
Does marketing need to own the data lake? No. Marketing needs governed access, its own agents, and a seat at the table where the customer record gets defined. Ownership of the infrastructure matters far less than ownership of the definition and the ability to act without a request queue.
Why does it matter which department owns customer data? Because the owning department's lens shapes the data. A technology organization without a marketing perspective will structure, prioritize, and retain customer data differently than marketing needs it, and marketing typically has less internal pull than the other stakeholders competing for that team's attention.
Do agentic CDPs work for B2B? Current agentic customer data platforms are built primarily around person-level identity resolution and high-volume personalization, which reflects B2C requirements. B2B needs account-level resolution and buying group orchestration. No platform has solved committee-level resolution, and much of B2B committee behavior is not captured digitally at all.
What is the risk of consumption-based pricing for marketing data? Consumption pricing places a monthly cost ceiling in the hands of whoever holds the contract. If that is IT, marketing's data access becomes subject to a budget cap it does not control and cannot forecast against.
How do you unify data across departments that all own a piece of the customer? Through a customer lens applied from the top down, with clear accountability for the end-to-end customer journey. Technology can enable a unified view, but without a shared goal, shared incentives, and a named owner, it produces faster silos rather than fewer.
Watch or listen to the full episode at pedowitzgroup.com/revenue-marketing-raw.