Back office ops · Production
How Slack built Slack AI to be secure and private using RAG and AWS escrow VPC
The problem
The generative AI market lacked enterprise-grade security and privacy patterns, and Slack needed a way to leverage top-tier LLMs while guaranteeing customer data would never leave its trust boundary or be used for model training.
Workflow diagram · grounded in source
1
User invokes Slack AI
trigger
“only the user who invokes Slack AI can see the AI-generated output”
2
ACL-gated data retrieval
validation
“Slack AI's search feature, for example, will never surface any results to the user that standard search would not. Summaries will never summarize content that the user could not otherwise see while reading channels. We ensure this by usi…”
3
RAG context assembly
ai_action
“we made the choice to use off-the-shelf models in a stateless way by employing Retrieval Augmented Generation (RAG). With RAG, you include all of the context needed to perform a task within each request, so the model does not retain any …”
4
LLM processing in escrow VPC
ai_action
“By using AWS, we are able to host and deploy closed-source large language models (LLMs) in an escrow VPC, allowing us to control the lifecycle of our customers' data and ensure the model provider has no access to Slack's customers' data”
5
Ephemeral output delivery
output
“Conversation summaries and search answers all generate point-in-time responses that are not stored on disk”
Reported outcome
Slack AI was built using RAG on LLMs hosted in an AWS escrow VPC, keeping all customer data within Slack's trust boundary.
Ninety percent of users who adopted AI reported higher productivity.
Reported metrics
AI adopter productivity gain90%
Reported stack
Retrieval Augmented Generation (RAG)AWSLLMs
Frequently asked questions
What did this team achieve with this AI workflow?
Slack AI was built using RAG on LLMs hosted in an AWS escrow VPC, keeping all customer data within Slack's trust boundary.
What tools did this team use?
Retrieval Augmented Generation (RAG), AWS, LLMs.
What results were reported?
AI adopter productivity gain: 90% (source-reported, not independently verified).
How is this back office ops AI workflow structured?
User invokes Slack AI → ACL-gated data retrieval → RAG context assembly → LLM processing in escrow VPC → Ephemeral output delivery.