Salesforce argues that marketers should move beyond fixed campaigns and respond continuously to customer signals. Instead of waiting for the next scheduled send, an intelligent system can react to browsing, abandonment, purchase cadence, or declining engagement through email, SMS, and WhatsApp. The article correctly identifies a real limitation of campaign calendars: customers do not arrange their needs around a marketer's planning cycle.
But the alternative is framed almost entirely from the sender's perspective. A third product view becomes evidence of peak intent. An abandoned cart becomes a trigger. Cooling engagement becomes a condition requiring intervention. Each inference may be plausible, yet none necessarily expresses a customer's wish to hear from the brand. Relevance is not the same as permission, and accurate timing does not make an unwanted message welcome.
Calling this architecture a relationship makes the omission more visible. Relationships contain pauses, ambiguity, and the right not to respond. An automated system is rewarded for finding the next actionable signal. Unless restraint is designed as an outcome, it will interpret silence as a problem to solve. That can produce an always-on loop of observation and persuasion in which ordinary behavior is continually converted into commercial intent.
The article warns that always-on should not mean sending more messages, but its proposed examples and competitive framing create exactly that pressure. If every interaction improves the model and the organization that starts first gains a compounding advantage, teams are encouraged to collect more signals, build more triggers, and shorten the time between observation and action. Frequency caps applied channel by channel will not solve the problem when email, SMS, app notifications, and messaging platforms all share the same profile.
A customer-centered design would begin with different questions. Which signals are appropriate to use? How uncertain is the inferred intent? What is the cumulative contact burden across channels? Which moments should deliberately produce no message? Customers need understandable preferences that control purposes and frequency, not only a list of channel opt-outs. Sensitive inferences should face higher thresholds, and systems should measure complaints, disengagement, and regret alongside conversion.
There is also a business risk. Real-time optimization favors outcomes that appear quickly, while irritation and brand erosion emerge slowly and are harder to attribute. A perfectly timed discount may train customers to abandon carts. A re-engagement flow may inflate a short-term metric while teaching subscribers that every pause will be pursued. Continuous learning is only as wise as the objective it is given.
The addendum is simple: the mature form of always-on marketing is not constant responsiveness. It is continuous judgment about when not to act. Brands should build suppression logic, uncertainty thresholds, quiet periods, and cross-channel contact budgets before celebrating an architecture that can reach people at any moment. The meaningful competitive advantage is not being perpetually present. It is being trusted to recognize when absence is more relevant.