A member calls her health insurer for help finding an in-network specialist. The representative listens, sends a list of practices, and earns a positive survey response. A week later, the member still has no appointment and cannot reconcile the information she received with what the practices tell her.
That first answer was honest. It described a helpful call, while the member’s underlying task was still unfinished. A company that treats the response as the final account of the experience loses the next chapter.
This is the opportunity I see in the evolution of surveys. Adaptive interviews can ask a better next question, and continuing conversations can return when circumstances change. The latter requires permission, useful memory, a way to correct the record, and an owner who can act when the customer asks for help.
Our fictional example, Harbor Health Plan, makes the distinction practical. Harbor could retain its existing fixed questions and offer one optional follow-up through its secure member channel. The member could explain what remains difficult, ask for assistance without completing research, and later return with an update instead of reconstructing the whole story.
Leaders should resist buying conversation volume as a substitute for better understanding. More contact can create fatigue and more service work, while people who participate over time may differ from those who decline. The test needs to include customer effort, missing outcomes, review time, cost, and whether the findings change a decision.
Synthetic personas also have a role worth testing. They can help a team prepare competing explanations and rehearse neutral questions before talking with real customers. Keep those simulated answers separate, record what real people contradict or introduce, and remove the synthetic step if it adds no useful learning.
The future extends beyond the company’s own channels. Brands may eventually earn permission to ask research questions inside the assistants people already use, and personal agents may relay authorized facts or decide which questions deserve their owner’s attention. Commercial assistant features do not provide unrestricted access to private conversations, and sensitive health contexts need particular care.
The spending decision deserves equal attention. An AI provider can appear cheaper while the incumbent contract remains unchanged, and released employee hours may create capacity without reducing payroll. Compare the same research workload, identify the bill that actually falls, and include transition, review, and retained costs.
This week’s AI, Not KPI turns those questions into a working method. The full issue includes the Harbor example, a method-selection workbook, a continuing-conversation playbook, a investment readiness model, a pilot workbook, leadership slides, and AI companion files. Founding Members receive an expanded investment model and a staged funding recommendation.
Read the full issue and access the member tools at AI, Not KPI. Before your next platform discussion, take one positive response and ask what your company knew about the customer’s unfinished task a week later. That answer will help determine what needs to change.


