Description:

Inconsistent behavior was identified in the AI Summary generated for conversations coming from the same flow of the Antioquia Oncology Center (COA) client.

Currently, the agent summary does not always consistently reflect all the fields collected during the conversation, especially those captured through buttons, menus, and structured flow selections.

📌 Observed scenario

In some cases, the AI Summary manages to correctly reflect fields such as:

type of care

utility

option/quote

EPS

However, in other similar scenarios, although the bot does correctly collect information during the conversation, the generated summary omits important operational fields and primarily prioritizes textual/cadastral information.

📌 Example of fields omitted in some cases:

Type of care

Requested service

Study/procedure

Desired time for the appointment

📌 Expected behavior

If the fields were correctly provided and collected during the conversation, they should be consistently reflected in the structured section of the summary generated for the agent.

📌 Impact

This affects the rapid visibility of important operational information for human agents, forcing them to manually review the entire conversation to obtain relevant context of the case.