Skip to main content
2024

Condor Develops Generative AI Prototype for IT Support in Weeks by Using Amazon Bedrock

The prototype built with the AWS partner MadeinWeb aimed to provide IT and Helpdesk support with Amazon Bedrock, the AWS managed service for generative AI.

Benefits

Integrated IT service

with the knowledge base of the area and the company's call history

Greater efficiency

in the search and advanced analysis of internal calls

Versatility

for the Charla platform when choosing foundational models, facilitating development and experimentation through APIs

Overview

Condor, a Brazilian company in the personal and household care accessories sector, developed a generative AI prototype to serve its information technology area. The project used AWS services, with support from AWS Partner MadeinWeb, creator of the Charla generative AI solution.

Charla, MadeinWeb's generative platform, is built entirely on AWS and helps companies to easily use Amazon Bedrock in cases such as customer service, process automation, and virtual assistants.

With two factories and more than 1,600 employees, Condor initially considered adopting Charla to expedite helpdesk activity resolution, reducing the burden on the IT team. The company's IT manager, Fabio Fernando Pereira, recalls that building on AWS was important to unlock this initiative. “We have had infrastructure projects with AWS for some time and we saw a lot of potential in AI solutions.”

Building on AWS resulted in an opportunity for the company to test and validate the development of solutions based on generative AI. “AWS has a lot of solutions to explore and opening them helps us experiment new possibilities,” Pereira explains.

Pereira recalls that the project resulted in the prototype of an IT service model integrated with the knowledge base of the area and the company's call history. “Using this chatbot, we had the idea of supplementing service desk activities with the adoption of generative AI,” he says. It's basically the creation of an interface where employees can find answers to their questions based on the company's IT ticket history. Charla also helps direct integration into company tools, such as the calling tool, as it is a service built with APIs.

The prototype built on Amazon Bedrock with the Charla platform provided yet another way to complement the service provided by Condor employees, focused on improving SLAs and optimizing team tasks.

Missing alt text value

About Condor

Founded in 1929, Condor is a Brazilian company that operates in the oral care, beauty, cleaning and painting products sector, and is a reference in the segment of utensils and accessories for personal and home care. With more than 1,500 products in its portfolio, Condor operates in over 100,000 points of sale in Brazil and exports to more than thirty countries.

Building a Solution to Evolve Complex Use Cases

The solution proposed by MadeinWeb, in partnership with AWS, was built based on a robust architecture, starting with the creation of a secure repository in Amazon Simple Storage Service (Amazon S3). This repository provided fast, secure, and cost-effective access to thousands of documents related to Condor calls, knowledge bases, and technical standards.

Native integration with Amazon S3 objects made it easy to implement the Amazon OpenSearch Service to create a fully vectorized database, allowing a more efficient search and advanced internal call analysis while also maintaining the versatility needed to handle various document formats.

“After extracting this data, we used the semantic capacity of a foundational model to understand the calls and structure the solution to the user's demands, which allowed for a very strong assertiveness in complex problems,” explains the director of Cloud, Data & AI at MadeinWeb, Vinicius Machado. Initially, customized models were implemented using Amazon SageMaker, providing substantial computational power. Subsequently, the switchover to Amazon Bedrock brought versatility to the Charla platform in choosing foundational models, facilitating development and experimentation through APIs.

In Machado's point of view, the choice of Anthropic's Claude-2 foundational model was crucial. “Claude-2 has a great advantage because it brings legal concepts and responsibility in its training, facilitating control over hallucinations and meeting the concepts of fairness AI,” he explains. “Hundreds of calls were analyzed and technical documents processed to ensure the accuracy of Charla's answers.”

The architecture was then built using key AWS services, including the Amazon API Gateway and the AWS Lambda function. Pereira explains that the prototype of the solution has been completed and that in 2024 the focus should be expanded to more AI tools from AWS, such as demand forecasting. “With a solution that generates assertive responses, we saw the real potential of AWS AI tools up close,” he reveals.

The other AI front, in its final phase by Condor and MadeinWeb, is the use of demand forecasting on AWS to facilitate Condor's S&OP process. It was fed with historical sales data from the past two years. The results already indicate an accuracy of over 90 percent between what is expected with machine learning and what is achieved in several SKUs.

“We believe that the use of machine learning greatly complements the demand analyses carried out internally, based on the team's human knowledge. In 2024, we hope to expand the use of AI thinking about all of the two thousand items that we sell today”, explains Pereira.  

Missing alt text value
With a solution that generates assertive responses, we saw the real potential of AWS Artificial Intelligence (AI) tools up close.

Fabio Fernando Pereira

IT Manager, Condor

Get Started

Organizations of all sizes across all industries are transforming their businesses and delivering on their missions every day using AWS. Contact our experts and start your own AWS journey today.
Contact Sales