The 2024 AWS Generative AI & Data Executive Forum in New York, NY, revealed exciting insights into the adoption of generative AI technology at the enterprise level.

Boston SoftDesign was invited to attend the AWS Generative AI and Data Executive Forum last month in New York, NY. The event drew product and engineering leaders from software companies and digital businesses to learn about and discuss topics related to generative AI and data. 

Here’s our rundown of this compelling industry event, including some macro trends in the software industry, the challenges of genAI implementation at the enterprise level, and planning considerations for AI adoption.

Frank Della Rosa, Research Vice President at IDC, provided an insightful forecast into the field of generative AI-powered business applications. Highlighting a significant shift in software delivery, Della Rosa projected that by 2027, 69% of software will be delivered as SaaS. The spending on SaaS and cloud software is expected to surge to a staggering $1 trillion worldwide. 

Embedded AI is becoming a central value proposition in this space, with Retrieval-Augmented Generation (RAG) emerging as a leading use case. The adoption of applications by business users continues to grow as more companies transition to cloud-based solutions and marketplaces. 

Snowflake is a good example of such solutions based on its advancements in Generative AI and data management. As AI shifts from being model-focused to data-focused, Snowflake’s robust platform supports this transformation by offering advanced edge infrastructure and multi-tenancy capabilities.

Della Rosa emphasized that having a strategic roadmap is essential in this competitive market. According to his forecast, Generative AI is expected to account for 29% of AI spending by 2027, with the total global expenditure on Generative AI projected to hit $151 billion. 

It’s widely recommended to work with trusted suppliers and implementation partners in this environment. Partnering with a company like Boston SoftDesign can provide crucial support in planning and implementing Generative AI, helping businesses keep up with technology and market trends.

Generative AI at Enterprise Scale

Enterprise-level generative AI involves the deployment of AI technologies across multiple business units within a large organization, including dozens of product and engineering teams, and addressing hundreds of different use cases. The scale and complexity of such implementations require infrastructure that can handle diverse requirements and vast amounts of data, ensuring that enterprise AI applications are both effective and efficient.

Amazon AI

Amazon Bedrock offers a compelling solution. It provides a scalable infrastructure that adjusts to varying needs, along with ‘Plug and Play’ foundation models that simplify the deployment of AI applications. Bedrock also supports the easy integration of Retrieval-Augmented Generation (RAG) patterns and ensures that all data remains private and secure, addressing major concerns about data safety and compliance in large-scale operations.

Moreover, Bedrock provides a variety of models, from the native Amazon Titan to the Claude family and Llama 2 by Meta. It also offers flexible consumption models tailored to specific needs and budgets. You can opt for the on-demand model, paying per token used with no long-term commitments—ideal for scenarios with some latency flexibility; or choose provisioned throughput for guaranteed performance in predictable and critical production workloads.

Planning for AI Adoption

It’s critical to focus on scenarios where the return on investment (ROI) in AI adoption can be objectively measured. This involves repetitive and time-consuming tasks where costs add up over time. By shifting these tasks from human labor to AI, businesses can see substantial cost savings. It’s also important to think about how users will interact with the AI from the start, integrating UI/UX feedback mechanisms to ensure the final automated workflows are user-friendly and effective.

The implementation process starts with identifying the right use cases. What are the specific tasks that could benefit most from automation? Next, adjust or design the development process to suit these tasks, choosing the appropriate foundational model that will drive your AI. Since AI technology is always advancing, with new models and updates being released continually, it’s crucial to have a plan for managing these changes. This can include centralizing control of production prompts, automating A/B testing to fine-tune AI performance, and ensuring there are enough resources to handle any arising issues or necessary improvements.

Ultimately, the success of the AI implementation depends on your ability to enhance Generative Al with contextual data. Using an architecture pattern like RAG can significantly boost the AI’s relevance and responsiveness by keeping it up-to-date and aware of the context in which it operates. This approach enhances the user experience and ensures that the AI works as a powerful business tool.