Answer capsule
Before an AI system writes or selects a lifecycle message, the CMO should classify its purpose, content, and context so a service label cannot hide a promotional decision.
What the source establishes
- The ICO says direct marketing includes activities that lead up to, enable, or support sending direct marketing, including targeting and profiling.
- The guidance says a service message with direct-marketing elements counts as direct marketing even when promotion is not the message's main purpose.
- The ICO advises organizations to examine why they use information, why they communicate, whether they seek to influence behavior, and whether the content is promotional.
- The guidance says neutral tone alone does not prevent a message from being direct marketing because the context and promotional purpose also matter.
Classify purpose before the generator writes
The direct CMO answer is to classify a lifecycle message upstream of audience selection, copy generation, and next-action choice. A label such as service, support, onboarding, renewal, or account notice does not settle what the communication is doing. The ICO says organizations should examine why they are using information, why they want to communicate, whether they are trying to influence behavior, and whether the content is promotional. That analysis belongs in the journey brief, not in a legal review after thousands of variants exist.
For each trigger, record the necessary service fact, intended recipient, customer need, business purpose, permitted action, channel, jurisdiction, and owner. Separate what the person must know to manage the relationship from an objective to sell, upgrade, cross-sell, renew, retain, donate, or support a campaign. Give the accountable owner a small set of explicit classifications and an unresolved state. The model may help apply an approved pattern, but it should not decide whether a commercial objective is administrative simply because the message sounds helpful.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Keep necessary service content separate from promotion
The ICO distinguishes non-promotional administrative or customer-service information from messages that also promote an offer or service. Its guidance says a service message with direct-marketing elements counts as direct marketing even when promotion is not the main purpose. That makes modular message design a useful governance control. Keep the required notice, approved explanation, optional recommendation, promotional module, call to action, and destination separately identifiable so the journey can send the necessary communication without silently adding a sales objective.
Do not rely on a neutral voice, short message, customer relationship, or familiar channel as a substitute for classification. A generator can turn a balance alert into an upgrade suggestion, a delivery update into a cross-sell, or a contract notice into a retention offer while preserving an administrative tone. Review the actual objective and effect of the added material. If the promotional purpose cannot be supported for that person and channel, omit the module rather than asking the model to disguise it as service.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Make classification an upstream system input
The guidance treats direct marketing as wider than the message itself. It includes preparatory use of people's information that leads up to, enables, or supports direct marketing, with targeting and profiling among the examples. The message classification therefore needs to travel upstream into audience eligibility, data access, lawful-basis review, preference and suppression checks, frequency rules, channel permissions, and any model context used to rank or personalize content. A final-send filter cannot correct an audience assembled for a purpose the organization never classified.
Enforce those controls outside the generator and preserve the decision record the generator receives. The record should identify the approved purpose, eligible audience, prohibited inferences, permitted content modules, jurisdiction and channel conditions, current preference state, evidence owner, and expiry or review event. When the classification is unresolved, the system should select a non-promotional service template, route to review, or stop. It should not infer permission from engagement likelihood, past purchase, account status, or the availability of customer data.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Review the served message and reopen on change
Test the complete communication in context before broad activation: trigger, recipient, subject line or notification text, generated body, recommendation, call to action, destination, timing, frequency, and adjacent journey messages. Sample ordinary cases and edge cases in which the service fact could invite an offer, such as a failed payment, expiring contract, low balance, delayed order, depleted allowance, or support interaction. Preserve the exact rendered variant, inputs, classification, reviewer, date, and outcome. A component library passing review does not prove that every generated combination preserves the approved purpose.
Monitor for promotional drift, preference failures, complaints, confusing service communications, and changes in message goals or model behavior. Reopen review when the trigger, audience, data, offer, channel, jurisdiction, generator, prompt, retrieval source, template, destination, or journey objective changes. Give service teams a route to identify a disputed communication and marketing an immediate way to suppress the promotional module or pause the journey. The ICO source is UK regulatory guidance, not a universal classification decision; qualified review must apply the relevant law and facts to the actual communication.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Decision test
Ask whether the source changes the decision itself, the evidence required, the implementation sequence, or only the language used to describe an existing capability. Record which claims are directly supported, which are provider statements, which require an independent test, and which remain unknown. A source-linked review should make uncertainty easier to see, not bury it inside a blended score.
Questions to take into review
- What customer data and lawful basis support the decision?
- Which offers or messages are prohibited?
- Which repository owns approved content?
- How are market and channel variations controlled?
- Which people and channels are represented?
- Can each insight be traced to evidence?
- What is the optimization target?
- Which placements and audiences can be excluded?
The publication supports research and executive decision preparation. It does not provide legal, financial, accounting, employment, clinical, cybersecurity, investment, procurement, or implementation advice.