Hi Haridass,
I was looking into this as well, and I think there is an important distinction between Architect User Prompts and AI Guide variables/messages.
Based on the current documentation, the Say command expects the message itself to be enclosed in quotes, although it can include simple Guide variables. I could not find a documented syntax that allows an AI Guide to directly dereference an Architect User Prompt such as Prompt(InitialGreeting).
An Architect User Prompt is not simply a string. It is a prompt resource/container that can contain TTS or recorded audio resources per language. Architect can resolve and play that prompt, but the Guide does not appear to expose a method for retrieving the underlying TTS text from that prompt.
If the main objective is to keep the greeting independently maintainable, I wonder whether the cleaner architecture would actually be:
Architect Bot Flow → play User Prompt InitialGreeting → Call Guide
This keeps the corporate/static TTS content in Architect and the conversational logic in the Guide. It is also worth noting that Genesys documentation states that bot flows use the latest version of a User Prompt rather than a snapshot.
Another option would be to create a String input variable in the Guide and pass the greeting text from Architect. AI Guides support input variables from the Virtual Agent flow and those values can be referenced in Say instructions. However, that would still be a String passed to the Guide - not a direct reference to an Architect User Prompt.
For the newer Agentic Virtual Agent, I see a similar separation. Architect can pass primitive values through Start Context, while AVA configuration provides specific exit and escalation responses, but I do not see a documented mechanism to reference an Architect User Prompt directly from the agent.
So I think the key question for Genesys would be:
Is there any supported mechanism for an AI Guide or Agentic Virtual Agent to dereference an Architect User Prompt and use its underlying TTS resource, or should User Prompts always be played by Architect before/after invoking the Guide or Agent?
If the latter is the intended design, then keeping reusable corporate prompts in Architect and the agentic conversational behavior in AI Studio seems like the cleanest approach.
------------------------------
Fernando Sotto dos Santos
Consultor de Atendimento Senior Grupo Casas Bahia
------------------------------