TaskUs Enhances Customer Interactions with Modular TaskGPT Platform and AWS Integration

For more than a decade, TaskUs has worked to annotate the data behind everything from voice assistants to self-driving cars. Its secure, proprietary AI customer service tools integrate into clients’ businesses and help customer service teams be more accurate and efficient.

In 2023, TaskUs launched TaskGPT, a generative AI platform and suite of tools, providing centralized product entitlement, multiple large language model (LLM) support, personally identifiable information (PII) masking and replacement, cost tracking, and observability. Using AI to enhance the efficiency of those in customer service roles, TaskUs is facilitating consumer, patient, and product end user interactions and helping agents provide the right answers, quickly. With nearly 50,000 employees worldwide, TaskUs capitalizes on their innovative technology, and their talented customer service teammates, to support their clients’ growth.

Initially, TaskUs’ engineering teams managed large-scale datasets with embeddings stored in flat files, using Python for cosine similarity and similar searches. As demand for TaskGPT’s offerings grew, the team needed a more robust vector database technology to meet increased demand with minimal latency.

Their goal was not only to help their global teammates continuously improve—reducing Average Handling Time (AHT), improving Customer Satisfaction (CSAT), and other metrics—but to keep pace with the constant evolution of generative AI capabilities.

The TaskGPT platform is designed with a modular architecture that supports extensive customization and scalability, making it adaptable for a wide range of client needs. This flexibility allows clients to deploy solutions that enhance customer interactions through secure AI assistance and intelligent chat support. By leveraging the comprehensive suite of AI and Machine learning (ML) tools on AWS—like Amazon Bedrock, Amazon Connect, Amazon ML Lens, Amazon Lex, Amazon Transcribe, Amazon Translate, and Amazon SageMaker—TaskUs has deepened its generative AI capabilities, benefiting from the infrastructure and services of AWS tailored for high-intensity workloads and AI development. This strategic integration enables TaskUs to innovate continuously, enhancing both customer experiences and operational efficiencies across various industries.

Amazon Bedrock serves as a pivotal component within TaskGPT, facilitating efficient access to a variety of LLMs and foundational models. This unified Application Programming Interface (API) approach across all first-party and third-party models allows TaskUs to deliver tailored solutions efficiently. Integrating Pinecone as its vector database solution powers semantic searches and Retrieval Augmented Generation (RAG) for AssistAI, and client-based recommendations and suggestions via Prompt AI. Prompt AI utilizes Natural Language Processing (NLP) and intent mapping stored in Pinecone to provide personalized recommendations via chat, enhancing the effectiveness of customer support interactions.

TaskUs utilizes AWS services for developing tailored generative AI solutions that deliver results at the speed, scale, security level, and cost required by their clients. Their strategy includes utilizing the compute capabilities and data management tools on AWS like Amazon Relational Databases (RDS), Amazon Redshift, and Amazon OpenSearch Service, ensuring robust performance and scalability.

TaskGPT is changing the landscape of customer service interactions, increasing CSAT scores, and teammate efficiency, with results such as reducing AHT by 20% for voice calls, and a 10% reduction for chat interactions.

Other results include:

• Delivering generative AI-fueled responses securely, 24/7, in multiple languages—freeing teammates from mundane tasks and time-consuming research.
• Empowering teammates to provide faster and more precise responses to customer queries, and customized responses according to ongoing customer interactions.
• Unlocking teammate potential and providing insights for improvement from voice and digital interactions through sentiment analysis, named entity recognition, classification, and natural language to structured data querying.
• Increasing CSAT scores by 5%, reflecting improved service quality and responsiveness achieved* with the TaskGPT platform.

TaskUs has worked with AWS since 2015, maximizing ML and generative AI capabilities to enhance its suite of digital services. The collaboration has been instrumental in optimizing security and enhancing DevOps automation processes. The AWS cloud infrastructure has been instrumental in scaling operations, optimizing security, ensuring seamless and efficient service delivery across diverse client needs. Amazon Bedrock’s unified API enables TaskGPT
to access Amazon and third-party generative AI models seamlessly, aligning with TaskUs’ adaptable architecture.

By capitalizing upon a wide range of models available on Amazon Bedrock—Meta Llama 3, Amazon Titan, among others—TaskUs has enhanced its ability to deploy sophisticated AI solutions, leveraging embeddings and semantic search functionalities to drive continuous innovation and meet increasing client demands. Looking forward, TaskUs plans to expand into new areas such as image recognition, anomaly detection, and advanced natural language processing capabilities. 

*Results shown are based on specific use case for a financial services client