Hi Uzay,
From what I have seen so far, your understanding is correct: today Genesys Knowledge / Knowledge Fabric does not publicly expose the same generative-answer capability that Copilot/Knowledge Assist internally leverage.
The currently available APIs are primarily focused on:
-
knowledge search
-
article retrieval
-
snippets/documents
-
search suggestions
But not a true:
"submit arbitrary text → receive generated contextual answer"
style API endpoint.
So for the exact use case you described:
Inbound Email → send email body → generate contextual answer from SharePoint-indexed knowledge → automatically respond
I do not believe there is currently a native/public Knowledge Fabric API that fully delivers that workflow end-to-end inside Architect/Data Actions alone.
Architect can absolutely orchestrate the process, but the actual generative layer usually requires an external orchestration/RAG service today.
What I have seen working well architecturally is:
Genesys Inbound Email Flow
→ Data Action
→ External AI/RAG service
→ Retrieval against enterprise knowledge
→ LLM-generated response
→ Return generated content back to Architect
→ Automated email response
In many implementations, SharePoint itself still remains the source of truth, while the external AI layer handles:
-
embeddings/vector search
-
retrieval orchestration
-
prompt construction
-
hallucination control
-
answer generation
-
citation handling
-
confidence scoring
One important distinction:
Knowledge Fabric indexing and Copilot experiences are optimized primarily for human-assist and conversational experiences inside the Genesys ecosystem, not necessarily as a fully open generative AI platform API.
That is why many enterprise customers currently end up implementing a hybrid model where:
Genesys handles orchestration/routing/channel lifecycle
while
external AI services handle advanced RAG/generative workloads.
Operationally, this also gives you much more flexibility around:
-
model selection
-
governance
-
observability
-
response validation
-
token/cost control
-
security boundaries
-
multilingual processing
-
prompt engineering
Another important consideration for automated email generation:
I would strongly recommend implementing confidence thresholds and human fallback paths before fully automating outbound responses, especially for regulated environments. Email has a much higher operational/legal persistence compared to chat or messaging.
So in summary:
I do not believe the native generative-answer API you are looking for is publicly exposed today, and for production-grade implementations an external RAG/LLM orchestration layer is currently the more realistic and scalable approach.
------------------------------
Gabriel Garcia
NA
------------------------------