How can I enforce customer consent across AI-driven campaigns?

Summary

• Treat consent as governed data, not a setting inside one campaign tool: capture each customer's consent and preferences as versioned records, tie them to the customer profile, and require every campaign and AI agent to honor them at query time.
• Databricks CustomerLake is an agentic CDP embedded in the lakehouse and governed by Unity Catalog, so the same consent and access rules apply to marketers and to AI agents.
• Because Unity Catalog enforces access at the query engine, an AI agent cannot read attributes or target customers the requesting user is not permitted to use, and opt-outs can be applied as row-level rules that gate targeting.
• Campaign agents act with the permissions of the person using them (on-behalf-of-user authentication), so an agent inherits, rather than expands, what its user can see.
• A consent check can run before audiences activate to downstream channels, and Unity Catalog lineage plus system tables keep an auditable record of how consent state connects to each use. CustomerLake is offered as a preview; confirm current availability.

How can I enforce customer consent across AI-driven campaigns?

Enforce consent where the data and the agents actually run, by treating consent as governed data rather than a checkbox inside a single campaign tool. Capture each customer's consent and preferences as versioned records, tie them to the customer profile, and require every campaign, query, and AI agent to honor those records when they read the data. Databricks supports this pattern with CustomerLake, an agentic customer data platform embedded in the lakehouse and governed by Unity Catalog, so the same consent and access rules apply to marketers and to the AI agents acting on their behalf. CustomerLake is offered as a preview, with capabilities rolling out over time.
AI-driven campaigns multiply the number of things touching customer data, so enforcing consent inside each activation tool leaves a gap at every new agent and channel. Enforcing it once, where the data lives, means an opt-out is honored no matter which agent or campaign asks.

What enforcing consent across AI-driven campaigns requires

Enforcement that holds up across many campaigns and agents rests on a few things: consent captured as current data and tied to the unified customer profile rather than one tool's private copy; controls that apply to people and agents the same way; a check at the moment of activation; and a record of how consent state connected to each use, so you can answer where a customer's data went.

Key capabilities to evaluate

Capability	Why it matters
Consent captured as governed data	Versioned consent and preference records give a current, queryable source of each customer's choices.
Controls that apply to agents too	Access rules enforced at the query engine cover AI agents, not just human users, so an agent cannot bypass them.
Permission inheritance for agents	An agent acting on behalf of its user inherits that user's access instead of a broad, over-scoped identity.
Row- and column-level enforcement	Opt-outs and restricted attributes can gate which rows and columns any consumer sees, applied at read time.
Consent check before activation	A check runs before audiences activate, so suppressed or opted-out customers are held back from targeting.
Lineage and audit	An auditable record links consent state to each downstream use, supporting data-subject questions.

How Databricks approaches it with CustomerLake

CustomerLake runs inside the lakehouse, so customer data does not move into a separate marketing system, and it inherits governance from Unity Catalog. Unity Catalog applies access control in layers: where data can be accessed, who can access what, attribute-based rules that gate data by classification and purpose-of-use tags, and row and column filters. Because these fire at the query engine, they apply the same way to dashboards, SQL, and AI agents, so no agent can read past them.
Consent itself is handled as data. Consent and preference records are ingested, for example through Lakeflow Connect from consent management platforms, and kept as versioned tables where each consent event is a row and the current state is a view. Automatic lineage tracks those records from raw source through downstream use. Opt-outs and regulated restrictions can be expressed as row-level rules that suppress those customers from segmentation and targeting, and column masks redact sensitive attributes at read time based on the caller's role.
The agent layer stays inside the same boundary. CustomerLake's Campaign Agents act with the permissions of the person using them, through on-behalf-of-user authentication, so an agent inherits rather than expands what its user can see. Marketers can describe audiences in plain language through Genie while those same controls decide, underneath, which attributes and customers are in play. Before an audience activates to downstream marketing and advertising channels, a consent check can gate the send, and system tables keep query-ready audit logs of what was accessed. See What is an agentic CDP? for the category background, and because CustomerLake is a preview, confirm the current status of its modules before relying on a specific capability.

FAQs

How is consent enforced for AI agents, not just people?

Unity Catalog enforces access at the query engine, so its rules apply to AI agents the same way as to users and dashboards. A CustomerLake agent acting on behalf of a user inherits that user's permissions and cannot read attributes or target customers the user is not allowed to use.

How are opt-outs honored across campaigns?

Consent is stored as governed, versioned data tied to the customer profile. Opt-outs can be applied as row-level rules that suppress those customers from targeting, and a consent check can run before an audience activates, so the choice is honored regardless of which campaign or agent asks.

Can I audit how consent was applied?

Yes. Unity Catalog lineage traces consent records from source through downstream use, and system tables keep query-ready audit logs, so you can answer where a customer's data went and how their consent state was applied.

Is CustomerLake generally available?

CustomerLake is offered as a preview, with capabilities rolling out over time. Confirm the current status of Profile Agents and Campaign Agents with your Databricks account team before relying on them.