top of page

Architecting and Delivering Value Using Generative AI - an Example

Delivering Value Using Generative AI – an Example at Boston Scientific.png

by Daniel Lambert (book a 30-minute meeting)

The job of business and enterprise architects is changing significantly right now. They are still designing and aligning an organization's business strategies to business capabilities, information, and IT infrastructure, ensuring that their technology and processes support their overarching goals, but generative AI is changing how their output is made. They are generated at a much rapid pace. Planning in the organization can now handle many more scenarios and iterations in a shorter time than a year ago because of generative AI. With the willingness of top management, it has become simpler to plan and architect from strategy to execution in an agile setting. In this article, I intend to show you what is now possible to do for a business and enterprise architecture team at Boston Scientific, for example, with generative AI using strictly publicly available information[i].

Business Context

To enhance the quality and relevance of outputs generated using AI, establishing a clear business context is essential. Figure 1 below presents the business context for Boston Scientific, compiled from publicly available information. This context includes the company's mission, vision, primary strategies, customer segments, and product lines. This draft took only a few minutes to create, offering a significant time-saving advantage over starting from scratch.


The Current Use of Enterprise Architecture in SAFe®


According to the Scaled Agile Framework (SAFe), the Enterprise Architect plays a vital role in defining the portfolio’s technology vision, strategy, and roadmap. Enterprise Architects (EAs) are responsible for shaping the technical direction of the enterprise by developing and evolving the technical architecture and creating portfolio-level roadmaps that embrace cutting-edge technologies, as shown in Figure 1 below.

Figure 1 - Boston Scientific Business Context.png

This generative AI output can be created using platforms like OpenAI, Co-Pilot, Gemini, Claude, Le Chat Mistral, and others. While these platforms allow you to set a business context for prompts related to Boston Scientific, you may need to re-establish this context frequently. For this reason, using a more specialized generative AI tool tailored to your specific workspace or project—such as ‘Boston Scientific’ in this case—can be beneficial.

The Importance of Prompts in Generative AI

Prompts are crucial in generative AI as they guide the model's response, directly impacting the quality, relevance, and accuracy of the output. Every word matters in a well-crafted prompt. A good prompt can help narrow focus, align responses with specific contexts, and produce insightful results tailored to the architect’s needs. Prompts can also generate significantly better results using confidential information and documents. For example, in complex projects like business architecture or strategic planning, providing context through preliminary prompts allows AI to generate content that resonates with project goals and organizational specifics, as explained earlier. Additionally, prompts help manage tone, complexity, and style, making them key to achieving useful, actionable, and consistent results. Thus, effective prompt design is essential for maximizing AI’s potential.

Providing Value

Business and enterprise architecture can enhance SAFe® project delivery a lot using generative AI. When planning and architecting from strategy to execution in an agile environment, I prefer to start with value streams, which are the cornerstone of the Scaled Agile Framework® (SAFe®). Figures 2 and 3 below illustrate the top 15 client-focused value streams for Boston Scientific, as suggested by generative AI—although many more could have been identified.

Figure 2 - Top 15 Client-Driven Value Streams for Boston Scientific Based on a ChatGPT Pro
Figure 3 - Top 15 Client-Driven Value Streams with their Value Stages for Boston Scientifi

The information in Figures 2 and 3 was generated by ChatGPT from Open AI, though other AI platforms could produce similar results. This table is a draft generated again in just a few minutes.

Next, let's take one of these value streams, ‘Develop Minimally Invasive Devices,’ and use generative AI to outline and describe its value stages. The results are presented in Figures 4 and 5 below.

Figure 4 – Value Stages of the “Develop Minimally Invasive Devices” Value Stream Based on
Figure 5 – “Develop Minimally Invasive Devices” Value Stream Diagram Based on a ChatGPT Pr

Again, these 7 value stages of the value stream “Develop Minimally Invasive Devices” were generated in a few minutes with strictly public content. This output would have been significantly enhanced using confidential information and documents. It now needs to be validated by subject matter experts, which can be anyone listed in the next section.

 

Participating Stakeholders

Using generative AI, identifying the internal and external stakeholders involved in each value stage of this value stream becomes a straightforward task. The results from my prompt are displayed in Figure 6 below for the first two value stages "Research Innovative Materials" and "Design Prototypes". Figure 7 shows 42 participating stakeholders for all value stages of this value stream.

Figure 6 – Some of the Participating Stakeholders of the “Develop Minimally Invasive Devic
Figure 7 – Participating Stakeholders of the “Develop Minimally Invasive Devices” Value St

This draft includes 12 internal and external stakeholders for the ‘Research Innovative Materials’ and ‘Design Prototypes’ value stages, with 6 stakeholders listed per stage. More stakeholders may be involved, and some may contribute across multiple value stages, though none are repeated here. Created in just minutes, this draft demonstrates a significant time-saving advantage over starting from scratch, though it remains a preliminary version open to further refinement.

Enabling Business Capabilities

Using generative AI to generate enabling capabilities for a value stream is challenging, as results often require more refinement. Substantial work is typically needed to polish these capabilities before presenting a draft to subject matter experts and business stakeholders. I used generative AI to identify enabling capabilities for the ‘Develop Minimally Invasive Devices’ value stream, resulting in a prompt with 42 capabilities. Figure 8 below highlights 12 of these for the ‘Research Innovative Materials’ and ‘Design Prototypes’ value stages. Figure 9 shows 42 enabling capabilities for all value stages of this value stream.

Figure 8 – Some of the Enabling Capabilities of the “Develop Minimally Invasive Devices” V
Figure 9 – Enabling Business Capabilities of the “Develop Minimally Invasive Devices” Valu

Significant improvements are needed before finalizing this draft for publication. The capabilities listed likely represent level-2, level-3, and level-4 business capabilities, which need to be accurately mapped to the organization’s current capability map. Some enabling capabilities may be missing, and others may require more precise naming. Additionally, certain capabilities may duplicate or closely resemble existing ones. Completing this task efficiently would benefit from a robust EA tool.

Furthermore, each of the 42 enabling capabilities in the 'Develop Minimally Invasive Devices' value stream, excluding duplicates, must be aligned with the organization's current applications, modules, or micro-services using APIs, where applicable. Advanced EA tools can now leverage generative AI to automate this alignment, achieving reasonable accuracy. Additionally, these enabling capabilities may need alignment with other important domains within the business architecture of the organization.

Required Information

Information and data architecture are vital to the success of any artificial intelligence project. Traditional information mapping is often too complex to effectively support digital transformation delivery teams. Identifying the required information types for a value stream offers a more accessible and practical alternative. Figure 10 illustrates 12 information types for the ‘Research Innovative Materials’ and ‘Design Prototypes’ value stages of the 'Develop Minimally Invasive Devices' value stream. Figure 11 shows 42 required information types for all value stages of this value stream.

Figure 10 – Some of the Required Information of the “Develop Minimally Invasive Devices” V
Figure 11 – Required Information of the “Develop Minimally Invasive Devices” Value Stream

Substantial refinement is necessary before this draft is ready for publication. The information types identified likely correspond to level-2, level-3, and level-4 types, which must be accurately aligned with the organization’s existing information map—a structure often less detailed than the organization’s capability map. Some required information types may be missing, while others may need more precise naming. Additionally, certain information types may overlap or closely resemble existing ones. Using a robust EA tool would streamline this process significantly.

 

Strategy to Roadmap

Specialized generative AI tools are now advancing the pace of digital transformation by enabling the rapid creation of multiple detailed iterative scenarios. These tools can support capability-based roadmap output drafts that can efficiently outline strategic requirements, paving the way for thorough planning in complex projects. The ability to generate such scenarios quickly allows organizations to adapt their digital transformation strategies with greater precision and responsiveness.

Beyond just strategic planning, specialized generative AI tools streamline the creation of user stories, which are often time-consuming but essential for guiding project execution. By automating in part this aspect, organizations can focus on delivering high-quality digital transformation initiatives, ensuring that roadmap requirements align seamlessly with user needs and overall business objectives.

 

In conclusion, generative AI is transforming the work of business and enterprise architects by enabling faster output and expanding the scope of scenario planning and iterations. With support from top management, planning from strategy to execution can now be more agile and efficient. This article explores how the architecture team of Boston Scientific, for example, could leverage generative AI, using only publicly available information, to achieve impactful, rapid results. Naturally, the impact would be even greater with access to confidential information.

_______________________________

[i] Boston Scientific is not a client of the author.

bottom of page