← All stories
● Covered by 1 source · 1 reportMedium impact1 neutral

Amazon Replaces Claude Sonnet 4.6 with Strands Decider for Internal Ticket Labeling

🔄 Updated 11d ago
New to BrevFeed? We gather this story from every outlet covering it into one summary — ranked by real-world impact, not just the latest headline — so you never miss what matters. What is BrevFeed? →

Key points

  • Replaced Claude Sonnet 4.6 with Strands Decider.
  • Achieved 98% cost reduction for ticket labeling.
  • Maintained performance levels.
  • Projected to save millions of dollars annually.

Cost-Saving Model Replacement

Amazon and AWS teams previously used Claude Sonnet 4.6 for labeling tickets within an internal system. This model was chosen for its balance of intelligence and cost-effectiveness in handling a large volume of internal tickets.

Recently, Amazon engineers replaced Sonnet 4.6 with Strands Decider, an internal decision model. This switch resulted in a 98% reduction in token costs while maintaining the same performance level, leading to projected annual savings of millions of dollars.

Transition to a Decision Model

The internal ticketing system required labeling tickets into 20 category types, including customer escalations, capacity requests, and security findings. Sonnet 4.6 processed each ticket with a long system prompt, generating JSON and reasoning against fixed labels.

Strands Decider, a decision model, does not generate text but provides probabilities and confidence scores for a given question and options. This approach was applied to ticket labeling, where the ticket served as the state and the label catalog as the options.

Model Selection and Implementation

The initial implementation used Strands Decider 2B, but testing showed that the E4B Gemma 4 build yielded better results and was subsequently adopted. The model's confidence threshold for applying labels was also fine-tuned.

The E4B model runs on SageMaker ml.g5.xlarge instances, utilizing about 19 GiB of memory per instance. Each request takes approximately 640 ms. The endpoint auto-scales with traffic and operates within an isolated network.

Standard AWS Services Utilized

The architecture for this solution uses standard AWS services. Ticket events are sent to an SQS queue, which triggers a Lambda function. This function reads ticket data and calls the SageMaker AI Endpoint hosting Strands Decider.

AppConfig manages the switch between models, and CloudWatch handles logs, metrics, and alarms. The entire stack is defined in CDK, with deployment stages for Beta, Gamma, and Production environments.

✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →

The daily brief

One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.

One email a day. Unsubscribe in one click, any time.

Today's brief

Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.

~4 min · 3 stories · Nov 21

▶ Play today's brief Listen on Spotify

New every morning, and the back catalogue is archived by date.

Reporting from

Amazon and AWS teams replaced Claude Sonnet 4.6 with their internal Strands Decider model for labeling tickets in an internal system. This change reduced token costs by 98% while maintaining performance, projected to save millions of dollars annually.