Amazon Bedrock handily outperforms do-it-yourself approaches for common generative AI use cases as platform-as-a-service simplifies enterprise adoption, according to a Constellation Research report by Holger Mueller.

The report landed as Amazon Web Services outlined Bedrock updates with AI orchestration and revamped SageMaker. The storyline for SageMaker and Bedrock is that they are better together. Bedrock is a serverless genAI platform that makes it easy for enterprises to build out from curated models. SageMaker is a platform that's designed for AI, data and machine learning workflows for more advanced and customized deployments.

Mueller's report looks at four use cases for Bedrock including agent creation and operation, retrieval-augmented generation (RAG), guardrails, and AI workflows. Mueller also outlines enterprise AI challenges and offers best practices.

The upshot is that platform-as-a-service offerings for genAI are going to become critical to enterprise AI adoption. Enterprises are struggling with lack of skills, multiple models, a pressure to pick winners and a breakneck innovation cadence. Enterprises are also finding traditional innovation best practices don't hold up with genAI.

Here are a few takeaways that stick out from Mueller's report on Amazon Bedrock benchmarking.

  • Agent creation with Amazon Bedrock takes 11 hours compared to 123 hours minimum.
  • RAG with AWS knowledge bases vs. DIY approaches takes 9 hours to 11 hours on Bedrock compared to a minimum of 84 hours for DIY.
  • "The results of this report are clear for CxOs: Do not go down the DIY path, but instead use PaaS tools such as Amazon Bedrock to achieve the outcomes your enterprise requires to be a winner in the AI era," said Mueller.

More from re:Invent:

 

Business Research Themes

Data to Decisions

Roles

Chief Information Officer

Tags

#AI #artificialintelligence #machinelearning #ML #generativeAI, #AWS, #AWSreinvent