Is your data AI-ready? Confluent panel with AWS, MongoDB, and Deloitte (AIM208)

Summary of Key Takeaways

Introduction

• The panel discussion focused on the challenges and best practices for getting Generative AI applications into production, with a focus on the importance of data readiness.
• The panel consisted of representatives from AWS, MongoDB, and Deloitte Consulting, providing diverse perspectives on the topic.

Emerging Trends and Opportunities in Generative AI

• The panelists highlighted the growing trend of "Agentic AI" - AI applications that can perform multi-step tasks in an automated and seamless fashion, going beyond simple question-answering.
• There is a shift from proof-of-concept (POC) stage to true production workloads for Generative AI applications, with over 50% of AWS customers now in production.
• However, building and deploying production-ready Generative AI applications is more than just the model itself - it requires additional technologies and capabilities around data contextualization, search and relevance, and data engineering.

Barriers to Success and the Role of Data Readiness

• The panelists identified several key barriers to getting Generative AI applications into production:
  • Demonstrating real-world use cases that can drive business value and reduce operational costs
  • Challenges around data modeling and bringing disparate data sources together in a usable format
  • Establishing compliance and data governance across the AI data pipeline
  • Building trust in the model outputs, especially in highly regulated industries
• Data readiness plays a critical role in overcoming these barriers, as "garbage in, garbage out" applies to these sophisticated AI systems.

Overcoming the Challenges: Emerging Best Practices

• The panelists suggested several emerging best practices to help organizations overcome the challenges:
  • Start small, focus on specific business processes, and demonstrate value before scaling
  • Simplify the data pipeline as much as possible, leveraging technologies like Confluent for real-time data integration
  • Establish a strong data foundation and work to ensure data quality, compliance, and governance
  • Leverage the capabilities of different platform providers (AWS, MongoDB, Deloitte) in a collaborative, "team sport" approach
  • Customize models and data to the organization's specific domain and requirements

Successful Real-World Examples

• The panelists shared several success stories where the data curation and delivery aspects were critical to unlocking the value of Generative AI:
  • Cisco's virtual assistant for customer support cases, where MongoDB's data compliance and security capabilities were key
  • Healthcare providers using Generative AI to transcribe patient conversations in real-time, saving time and improving patient experiences
  • Retailers and construction companies leveraging Generative AI to provide step-by-step guidance and recommendations to frontline workers

Advice for Getting Started

• The panelists provided the following key advice for organizations looking to get started with Generative AI:
  • Focus on data quality and data readiness as a foundation
  • Define specific use cases and requirements, rather than trying to "boil the ocean"
  • Leverage existing data architectures and customize models/data to your domain
  • Educate and upskill your teams on AI and data engineering best practices
  • Start small, prove value, and then scale iteratively

Overall, the discussion highlighted the importance of a holistic, collaborative approach to successfully deploying Generative AI applications, with data readiness and sound data engineering practices being critical to unlocking the true potential of these technologies.