Enhancements and solutions at Desk Manager with AWS Generative AI
Medium
ITSM and ESM, optimizing IT and business services
Brazil
Medium
Brazil
ITSM and ESM, optimizing IT and business services
Desk Manager is one of the leaders in the development and provision of ITSM (IT Service Management) and ESM (Enterprise Service Management) solutions in Brazil. The company offers corporate service management platforms to more than 500 clients in Brazil and abroad, covering various areas such as IT, human resources, marketing, administration, and building maintenance. With a young and engaged team, Desk Manager is recognized for its excellence in customer service, always striving to integrate the most advanced technologies to provide a differentiated and highly efficient experience.
The Challenge
Desk Manager faced a challenge when attempting to efficiently integrate generative artificial intelligence into its platform without adding extra steps for users. Since the launch of ChatGPT in 2022, the company had been exploring ways to incorporate this technology but encountered difficulties in implementing it without requiring clients to create separate accounts. Although adopting ChatGPT yielded positive results, obstacles arose during the process, such as the need to adapt the company’s databases for generative AI and to achieve deeper, seamless integration of the technology into the platform.
The Solution
In search of a more robust and efficient solution, Desk Manager partnered with CloudDog, an Advanced Tier Services partner of AWS. With CloudDog's consultancy, Desk Manager adopted a new architectural approach, expanding its use of AWS services to integrate Amazon API Gateway, AWS Lambda, and Amazon Bedrock. This serverless architecture was designed to optimize data processing and integrate the generative AI agent directly into the platform.
The integration with Amazon Bedrock, using the large language model Anthropic Claude 3, enabled more efficient, faster, and cost-effective content analysis. The implementation was completed in April 2024, and the new AI agent fully replaced the use of ChatGPT. Desk Manager’s team also began restructuring its knowledge bases to enhance the AI's comprehension capabilities.
Architecture:
Amazon S3: Serves as a central repository for storing data such as backups, multimedia files, and documents. It offers security, durability, and high scalability.
Amazon SQS: Acts as a message queue to process events asynchronously. It ensures organization and scalability in managing messages in complex data flows.
AWS Lambda (Initial Processing): Executes functions on demand to perform validations, transformations, and other logic on files before sending them to analysis or processing services.
Amazon Transcribe: Automates the conversion of audio into accurate text. Ideal for transcribing meetings, recordings, or any audio stored in S3, integrating seamlessly with other services.
Amazon DynamoDB: Stores information in a NoSQL database structure, ensuring fast and efficient access. It is ideal for handling large volumes of data and offers high availability.
Amazon DynamoDB Streams: Allows real-time monitoring of changes made to DynamoDB tables. It integrates with Lambda to automatically trigger events after any modification.
AWS Lambda (Stream Processing): Manages data coming from DynamoDB Streams, activating notifications or integrating with other systems, ensuring continuity in processed data flows.
Amazon SNS: Sends automated notifications to systems or end-users about the completion of processes, using channels such as email, SMS, or webhooks for integration.
Amazon Bedrock: Provides advanced machine learning models to generate predictive analyses or semantic insights, complementing data flows with artificial intelligence.
AWS Lambda (Advanced Analytics): Performs complex data analysis and processing before sending it to Bedrock or other solutions, applying specific logic as needed.
Amazon CloudWatch: Monitors metrics and logs for all integrated AWS services, sending alerts in case of issues. It ensures visibility into operations and performance.
AWS CloudTrail: Records all activities performed on AWS services, providing a complete audit trail and tracking administrative actions and automations within the environment.
AWS Config: Assesses and monitors the infrastructure’s compliance with defined policies, detecting unauthorized changes and generating real-time alerts.
AWS Budgets: Monitors costs associated with AWS services, alerting stakeholders when predefined financial limits are reached or exceeded.
This modular and integrated architecture combines the robustness of AWS with services like SQS, DynamoDB, Bedrock, and Lambda, offering scalable processing, security, and intelligence. Tools like CloudWatch, CloudTrail, and Config ensure observability and compliance, efficiently addressing complex demands and promoting continuous system evolution.
Results:
The implementation of this architecture by CloudDog for Desk Manager delivered several notable improvements:
Reduced Customer Support Costs: By adopting generative AI and leveraging Amazon Bedrock, Desk Manager achieved a 40% reduction in support costs, enabling more efficient resource allocation.
High Accuracy in Audio Recordings: The new solution demonstrated a 99% accuracy rate in transcriptions of Portuguese audio, enhancing the quality of customer support.
Customer Support Automation: Between April and June 2024, 13% of customer interactions were resolved automatically by AI, freeing the team to focus on more complex and strategic issues.
Compliance with LGPD: The integration of AI with Amazon Bedrock allowed Desk Manager to enhance compliance with LGPD, removing personal data from automated responses and improving information security.
Alexa Integration: Desk Manager now offers report generation via Alexa, providing users an additional channel to track the status of their operations.
Processing Longer Texts: The use of Amazon Bedrock enabled the processing of longer texts, surpassing previous token limitations and improving support capabilities.
Continuous Restructuring of Knowledge Bases: Desk Manager continuously updates its knowledge bases to enhance the accuracy and relevance of the AI’s automated responses.
These advancements have resulted in greater operational efficiency, improved customer satisfaction, regulatory compliance, and cost savings for Desk Manager.
