After learning about the SAP Business AI approach, you want to use Large Language Models (LLMs) in your business applications. You want to learn about SAP's methodical approach to embedding LLM capabilities in its solutions.
SAP's approach to embedding LLM capabilities is structured and methodical, aiming to harness the potential of AI while navigating the challenges of implementation and commercialization. The strategy not only seeks to enhance SAP's product offerings but also to establish a framework for integrating cutting-edge AI technologies in a scalable and efficient manner.
This lesson will help you identify the "bigger picture" across the different product development phases. It will identify the roles that must be included for each task, involving the following personas:
- Data Engineers & Software Architects: Focused on the technical aspects of AI integration.
- Product Management & Product Owners: Responsible for the product's road map, vision, and commercial success.
- Solution Management: Engages with the technical and business aspects to ensure that the product meets market needs.
- User Experience: Concentrates on how users receive products, including feedback and usability.
See the video to learn about the key steps involved in integrating LLMs into business applications.
Let's look at each of these steps in more details.
1. Ideation Phase
The Ideation phase is focused on identifying potential use cases for LLMs within SAP's ecosystem. It emphasizes the importance of technology exploration to assess the feasibility and potential benefits of implementing LLMs. We can highlight diverse applications of LLMs, such as language generation, text completion, chatbots, content summarization, and understanding tasks (for example, sentiment analysis, text classification, translation). An iterative approach to experimentation is recommended, starting with simple ideas, and gradually refining them to optimize LLM potential for specific use cases.
During the Ideation phase, the focus is on laying the groundwork for generative AI applications within SAP's offerings. This involves:
- Familiarization: Individuals, particularly Data Engineers and Software Architects, are encouraged to learn about the generative AI technology. This is done through reviewing SAP's AI strategy, accessing learning resources, and experimenting with the technology. For more information, see https://www.sap.com/products/artificial-intelligence.html.
- Customer Collaboration: Engaging with customers to explore potential use cases is key. This step is essential for Product Management & Product Owners who need to understand the customer's needs to ideate effectively. For more information, see Design-Led Development Process.
2. Validation Phase
The validation phase assesses the feasibility, desirability, and viability of implementing LLMs. This involves determining whether the problems at hand are suitable for LLMs, based on their capability to understand or generate natural language. There are clear "exclusion criteria" for tasks not suitable for LLMs, such as calculations, and compares the efficacy of LLMs against traditional machine learning techniques for certain problems. Cost-benefit analysis and customer data access are crucial factors in this phase, with specific emphasis on compliance with data protection policies when using third-party LLMs.
In the Validation phase, the aim is to refine the ideas into viable use cases and assess their feasibility:
- Use Case Iteration: This includes clearly defining the business problem and solution, and iterating on the value proposition through customer feedback sessions. Ensure that you follow ethical and data protections principles for your use case (mandatory for use cases with personal and/or personally identifiable information (PII) data).
- Technical Evaluation: Data Engineers & Software Architects ensure access to data for evaluation, test various models, and follow architectural guidelines defined for their use case.
- Commercialization Evaluation: Product Management & Product Owners should review pricing and commercialization and conduct cost-benefit analysis to understand the financial viability.
3. Realization and Productization Phase
The third phase is where the developed use cases are turned into actual products:
- Use Case Finalization: Based on customer feedback, the use case is refined to ensure it meets business requirements.
- Technical Finalization: Feasibility and security are key focus areas here, ensuring the product can be reliably deployed. Software Architects are primarily responsible for these activities.
- Commercialization Finalization: It is crucial to define the pricing model, finalize business cases, and establish a metering concept for consumption objectives . This is where the commercial aspect of the product is solidified, followed by regular productization steps like executing the commercialization model.
4. Operation, Continuous Improvement Phase
The final phase involves the ongoing management and enhancement of your product:
- Operations: Regularly gathering customer feedback, monitoring user adoption, and assessing model performance are critical tasks. This is where the User Experience team plays a significant role.
- Continuous Improvement: The product is continuously refined based on real-world usage data and customer feedback, ensuring that it remains effective and relevant.
This structured approach ensures that each phase of product development is thorough, collaborative, and geared toward creating robust, market-ready generative AI products that adhere to SAP's standards for software development and operations life cycle.