Condor accelerates IT support with Generative AI and Amazon Bedrock

With two factories and more than 1,600 employees, Condor needed a more efficient way to handle the demands of its information technology area. The central challenge was to reduce the load on the IT team by creating an intelligent interface where employees themselves could quickly find answers to their questions, all based on the ticket history and the company's internal knowledge base.

AWS partner MadeinWeb implemented Charla, its generative platform built entirely in the cloud, to structure Condor's service model. The solution involved:

• Storage and structuring: creation of a secure and cost-effective repository on Amazon S3 for thousands of technical documents and call histories.

• Advanced search: integration with Amazon OpenSearch Service to build a fully vectorized database, ensuring efficient searches and compatibility with various document formats.

• Generative AI and modeling: transition from Amazon SageMaker to Amazon Bedrock, using Anthropic's Claude-2 foundation model. This choice ensured accurate responses, responsibility in training, and strict control over AI hallucinations.

• Systems integration: construction of an API-based architecture, using Amazon API Gateway and AWS Lambda, allowing direct integration with the company's ticketing tool.

The prototype offered a new way to complement the service provided by Condor's employees, focusing on optimizing operational tasks and continuously improving SLAs. The main benefits achieved were:

• Intelligent IT service: support was fully integrated with the area's knowledge base and call history.

• Greater operational efficiency: the search and advanced analysis of internal demands became considerably more agile.

• Versatility for experimentation: the platform gained flexibility to test and adopt different foundation models through APIs in a simplified way.

• Openness to new projects: the success validated the expansion of AI to other areas of the company, such as demand forecasting for the S&OP process, which already projects an accuracy of more than 90% using historical data.

• Amazon Bedrock: a fully managed service that offers a choice of high-performing foundation models through a single API, for building secure and responsible generative AI applications.

• Amazon SageMaker: a platform that centralizes data, analytics, and AI.

• AWS Lambda: an event-driven, serverless computing service for running backend code.

• Amazon API Gateway: a service for creating, maintaining, and securing APIs at any scale.