Customer-data work: what is changing is who does it
The CDP’s work is being divided among more suppliers, but accountability for the result hasn’t followed.
Four of the five large data platforms leave the CDP to partners or keep it apart
Positions as of June 30, 2026 (Table 24).
All three major analyst firms describe pressure on the standalone CDP as data platforms take on customer-data work.
The work is being divided among more suppliers. No large data platform yet supplies it all, and responsibility for the result has not followed.
The five large data platforms take three positions.
As of June 30, 2026, three route the category to partner CDPs, one sells a data platform and a CDP and keeps them apart, and one embeds a CDP in its platform, not yet generally available.
No large data platform supplies what agents depend on.
None of the five supplies consent and data rights, first-party behavioral capture and identity resolution together as generally available first-party products. The one capability all five document, the clean room, depends on all three. Even in this case the work a CDP does is not displaced; what is contested is who packages which parts of it.
Architectures divide the work differently.
Hybrid and warehouse-native vendors claim next-best-action, agent context, interoperability and agent building more often than own-store vendors, and channel execution less often. Profile unification is claimed at a similar rate across the architectures, and every CDP type includes vendors of more than one architecture.
Members already obtain AI through several routes.
Two-thirds of members (66.7%) use general AI providers directly and 62.0% use AI built into software they already own. Each route needs access to the same customer record.
Accountability has not followed.
Buyers still assemble it, from a CDP or from several suppliers acting as one. Responsibility for a failure is getting harder to place.
Vendor documentation shows who supplies each function, not who is accountable when an agent fails, and no published work or this edition links architecture to deployment outcomes (§15).

