Better customer engagement starts with better judgment
A lead can be routed, a message sent, and an alert issued within seconds while the buyer receives the wrong response. Speed tells a marketing team how quickly its system acts; relevance depends on whether that action serves the buyer’s current need. Better engagement therefore requires better decisions about when to act immediately, investigate further, wait, or involve a person.
These choices were central to “From journeys to decisions: The next era of customer engagement” at the November 2026 MarTech Conference, moderated by Brianna Miller, director of demand generation at Cohere Health. The discussion examined decision-led engagement: choosing the next action using reliable information and buyer context. Putting that approach into practice starts with questioning the assumptions behind the stages and schedules that guide automated responses.
Campaign schedules and pipeline stages can misrepresent buying
Those assumptions become a problem when enterprise buying committees disagree, change priorities, or revisit earlier questions. A seller’s process may expect an account to advance through a sequence of stages, but people within that account can be evaluating different concerns at the same time. So a preset sequence can produce an unsuitable response when the buyer needs a step outside the seller’s plan.
Mehul Nagrani, CEO of Integrate + CaliberMind, argued that internal pipeline rules must accommodate this behavior before systems can choose useful actions as circumstances change. He leads a marketing technology business with a commercial interest in companies investing in marketing data and automation. In his account, buying committees shift their activity while revenue teams organize work around fixed stages and scheduled transfers between teams. When those rules determine every response, faster execution applies the same assumptions more quickly, so teams need to understand what has changed within the committee before advancing a deal or transferring it to sales.
Campaign calendars impose a similar constraint because they organize the seller’s work around dates the customer may have no reason to follow. “Customers don’t buy in campaigns, yet we market in campaigns,” said Carlos Hidalgo, CEO of Carloshidalgo.co, whose engagement advice concerns work businesses may pay for. He advocated sustained engagement that gives individual stakeholders relevant context throughout their evaluation. That changes the planning question: what does this person need to understand or resolve now?
A project in mind?
Schedule a 30-minute meeting with us.
Senior experts helping you move faster across product, engineering, cloud & AI.
Accurate activity data still needs buyer context
Answering that question requires interpreting activity before assigning it commercial meaning. Hidalgo cautioned that more page views or content downloads alone do not justify an aggressive sales approach. Increased activity gives a team something to investigate, but it does not establish that the buying group is ready for the response an automated scoring rule might trigger.
The stakeholder behind the activity helps explain what to investigate. In Hidalgo’s example, a chief financial officer’s download reflects concerns about protecting the budget and meeting compliance requirements, while an operations lead’s activity points toward workflow problems. Treating both actions as the same buying signal can disrupt a deal because the resulting outreach addresses the wrong concern. A useful response connects the activity to the person’s responsibility within the purchase.
A person’s responsibilities can also span several purchases. Jodie Wilson, freelance lifecycle and marketing technology architect at Jodie Wilson Consulting, described someone researching different solutions for two separate business units. Wilson has a commercial interest in demand for the architecture and consulting work involved in interpreting such activity. Her example shows how a system that assigns every contact to one solution and one journey can misread legitimate work even when its record of the pages viewed is accurate.
Even an accurate activity record may say little about purchase intent. Nagrani warned that generative tools and automation can produce large amounts of digital activity with little underlying intent to buy, so teams need to distinguish meaningful progress within an account from superficial interactions. Data cleaning can establish what happened; understanding stakeholder purpose helps determine an appropriate response.
Reliable decisions depend on usable records
Interpreting stakeholder purpose becomes harder when the records themselves misrepresent an interaction. “Fix your data first before you fix your tech stack,” Wilson urged. Her recommendation sets the order of work: identify and correct information problems before adding technology that will use the same records. Otherwise, the new system inherits the original defect.
Tim Boylan, owner of Grow-Wire Consulting LLC, illustrated that defect with an unnamed rental car company. His consulting business has a commercial interest in organizations seeking help with operational problems of this kind. Customers had agreed to receive SMS text messages, but their opt-in information, the record of their permission, was entered into incorrect fields, so willing subscribers missed important updates. The failure lay between collecting permission and making it available to the messaging system.
Because those customers had already agreed to receive messages, checking their willingness alone would have missed the cause. The investigation needed to trace permission into the fields the delivery system actually used. Adding AI would have left that mismatch unresolved: a more sophisticated system using the same records would still lack a correct account of who had consented. The company intended to send the updates, and customers wanted them, but incorrectly stored information prevented delivery.
Records can also undermine the analysis used to judge performance. Nagrani described experiments in which AI-generated performance dashboards looked polished and credible, but audits of their underlying data undermined that confidence. The appropriate response was to inspect the supporting information before using the presentation to assess results. Where Boylan’s example required checking how permission moved between fields, Nagrani’s required checking the records beneath a convincing output.
Both investigations distinguish consistent execution from a sound basis for action. Automation can repeatedly carry out a configured process using incorrect information, allowing unresolved errors to spread. Correcting the inputs makes a sound decision possible, while understanding the buyer’s circumstances determines whether that decision serves the person receiving the response.
Improve pivotal decisions before expanding automation
Repairing the inputs becomes more manageable when the team starts with a few consequential decisions. Boylan recommended selecting one or two defining moments in the customer lifecycle and cleaning the associated data. This gives the team a specific response to examine, along with the records that determine whether it happens correctly. Work can begin there without waiting for an overhaul of every system.
Once those inputs are usable, the customer’s request helps determine how quickly to respond. Nagrani gave the example of a prospect seeking standard product specifications or documentation. Requiring qualification by a sales development representative, or SDR, adds an unnecessary step before fulfilling a clear request. Immediate delivery serves the prospect because the required action is already known.
A complex relationship situation requires more interpretation, so Nagrani argued that teams should allow time for human context and deliberate thought. Speed alone cannot establish whether a response is appropriate; the decision depends on understanding the relationship before acting. Automation rules need to allow that work rather than requiring immediate action on every interaction.
The same need to change course can arise during an automated conversation. Boylan described a direct-to-consumer chatbot that should recognize when a customer is struggling and connect that person with a human specialist. Continuing the same automated exchange leaves the customer’s need unresolved, so the design needs a way to detect the difficulty and transfer the conversation to someone who can help.
Human control also matters when a model recommends an action before any conversation begins. Wilson described predictive models, which use data to estimate likely useful choices, as tools for suggesting relevant subjects, delivery channels and timing. Marketing architects retain authority over business rules and brand requirements, so a recommendation informs a decision made within those limits. People remain responsible for deciding which recommendations the organization can use and under what conditions.
Before expanding such decisions, Nagrani recommended testing whether the organization can explain one stalled opportunity. The team should record the available data, actions already taken, and assumptions behind those actions, then calculate the hours required across functions and identify the tools needed to investigate why progress stopped. The difficulty of that exercise reveals whether people can reconstruct a decision across teams and systems. It tests their ability to explain existing actions before automating more of them.
Use buyer feedback to revise response rules
Reconstructing a decision internally can explain what the organization did, but understanding the interaction also requires input from the people involved. Hidalgo recommended combining technical audits with direct conversations with sales, customer success, and buyers themselves. Those conversations can test whether the team’s interpretation addresses the business problem behind the recorded activity. They can also reveal why an apparently reasonable response failed to help.
When that feedback changes the explanation of an action, it needs to be able to change the response rule too. A buyer may explain that research concerned a different business unit, or that a download addressed a particular stakeholder’s concern. The team can then reconsider which action should follow that activity and who has authority to approve the change. Buyer feedback becomes operationally useful when it can alter what the system does next.
Key takeaways for leaders
- Adapt engagement to buying committees: Revenue teams need response rules that accommodate stakeholders revisiting questions and changing priorities. Review where fixed pipeline stages and campaign schedules trigger actions that overlook those changes.
- Interpret activity through buyer context: Marketing teams need to connect downloads and page views to each stakeholder’s role, business unit and purpose. Use that context to determine the next response and whether sales involvement is appropriate.
- Verify the records behind automated decisions: Marketing operations teams need to trace critical inputs, including consent, into the fields that delivery systems actually use. Audit the underlying records before relying on AI-generated performance reports.
- Start with pivotal decisions: Lifecycle owners can focus data cleanup on one or two consequential customer moments, then define when automation acts and when a person takes over. Investigating one stalled deal tests whether the organization can reconstruct its decisions before expanding automation.
- Give buyer feedback operational authority: Marketing architects need a process for turning feedback from buyers, sales and customer success into revised response rules. Assign clear ownership for approving changes when that feedback reveals a mistaken assumption.
A project in mind?
Schedule a 30-minute meeting with us.
Senior experts helping you move faster across product, engineering, cloud & AI.


