Hi everyone,
I recently worked on an Agentic Virtual Agent implementation and one of the main challenges we faced was how to measure the AVA's performance beyond traditional conversation-level metrics.
In our project, we found a way to track specific AVA actions and better understand what was happening during the interaction. However, this made me think about how this could be handled more efficiently and natively within Genesys Cloud.
One possible idea could be to have a way to associate AVA actions with outcomes, milestones, or key business steps. This would make it easier to track when an important action is completed, when a business rule is triggered, when a process reaches a specific stage, or when the interaction is successfully resolved or escalated.
Another option we explored was using the Virtual Agent automated wrap-up / classification capability to understand what happened during the interaction. However, one challenge is that this classification considers the overall interaction context, not only the specific behavior or actions performed by the AVA. In scenarios involving multiple flows, agents, systems, or steps, it can be difficult to isolate what was actually handled by the AVA.
I believe it would be very valuable for the platform to provide more efficient and native ways to measure AVA performance, especially at the action, step, or business-process level.
Some examples of metrics that could be helpful, depending on the use case:
-
successful versus unsuccessful AVA actions;
-
number of automated steps completed during the interaction;
-
completion rate of key business processes;
-
AVA contribution to resolution versus escalation;
-
milestone or checkpoint completion;
-
errors, retries, or fallback scenarios;
-
impact on containment, deflection, or customer effort;
-
outcomes by intent, journey stage, customer segment, or business scenario.
This would help teams better understand not only whether a conversation was contained, but also what the AVA actually did, which steps worked well, where it failed, and where the business process could be improved.
I'd love to hear from the community:
How are you currently measuring Agentic Virtual Agent performance today?
Are you using wrap-up codes, participant data, custom attributes, external dashboards, or any other approach?
Also, as a possible platform evolution, do you think it would be valuable to have a native way to track AVA-specific actions and associate them with outcomes, milestones, or business results?
#ConversationalAI(Bots,VirtualAgent,etc.)#Reporting/Analytics------------------------------
Mateus Nunes
CX Manager at Solve4ME
mateus.nunes@solve4me.com.brBrazil
------------------------------