It support · Production

Untold Studios empowers artists with an AI assistant built on Amazon Bedrock

The problem

Untold Studios needed to harness ML to enhance competitive edge while maintaining strict data security, and faced the challenge of building accessible interfaces for a diverse artist base with varying technical experience and access needs.

Workflow diagram · grounded in source

Artist queries via Slack

trigger

“Users interact with the Untold Assistant through private direct messages or by mentioning it (@-style tagging) in channels for everybody to see”

API Gateway and Lambda acknowledgment

integration

“The incoming event from Slack is sent to an endpoint in API Gateway, and Slack expects a response in less than 3 seconds, otherwise the request fails. The first Lambda function, with reserved capacity, quickly acknowledges the event and …”

Role-based access control

validation

“This system tailors the assistant's capabilities to each user's role and clearance level, making sure that it operates within the bounds of each user's authority while maintaining functionality. The access to data sources is controlled u…”

Claude function calling

ai_action

“Our Untold Assistant uses Amazon Bedrock with Anthropic's Claude 3.5 Sonnet model for natural language processing. We use the model's function calling capabilities, enabling the application to trigger specific tools or actions as needed”

RAG knowledge retrieval

ai_action

“Our RAG setup uses Amazon Bedrock connectors to integrate with Confluence and Salesforce, tapping into our existing knowledge bases. For other data sources without a pre-built connector available, we export content to Amazon S3”

Response delivered via Slack

output

“The implementation uses Slack's event subscription API to process incoming messages and Slack's Web API to send responses. Users interact with the Untold Assistant through private direct messages or by mentioning it (@-style tagging) in …”

Reported outcome

The Untold Assistant handles up to 120 queries per day and cuts information retrieval time from minutes to seconds, delivering a significant reduction in time spent searching for information and reduced load on the support and technology team.

Reported metrics

Queries handled per dayup to 120 queries per day

Queries invoking additional toolsabout 10–20%

Information retrieval timecut down the time from minutes to only a few seconds

Time spent searching for informationSignificant reduction in time spent searching for information

Show all 5 reported metrics

queries handled per dayup to 120 queries per day

queries invoking additional toolsabout 10–20%

information retrieval timecut down the time from minutes to only a few seconds

time spent searching for informationSignificant reduction in time spent searching for information

development timereducing our development time from months to weeks

Reported stack

Amazon BedrockClaude 3.5 SonnetStable Diffusion 3Amazon CloudWatchSlackRAGConfluenceSalesforce

Source

https://aws.amazon.com/blogs/machine-learning/how-untold-studios-empowers-artists-with-an-ai-assistant-built-on-amazon-bedrock?tag=soumet-20

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The Untold Assistant handles up to 120 queries per day and cuts information retrieval time from minutes to seconds, delivering a significant reduction in time spent searching for information and reduced load on the su…

What tools did this team use?

Amazon Bedrock, Claude 3.5 Sonnet, Stable Diffusion 3, Amazon CloudWatch, Slack, RAG, Confluence, Salesforce.

What results were reported?

Queries handled per day: up to 120 queries per day; Queries invoking additional tools: about 10–20%; Information retrieval time: cut down the time from minutes to only a few seconds; Time spent searching for information: Significant reduction in time spent searching for information (source-reported, not independently verified).

How is this it support AI workflow structured?

Artist queries via Slack → API Gateway and Lambda acknowledgment → Role-based access control → Claude function calling → RAG knowledge retrieval → Response delivered via Slack.