In the rapidly evolving digital landscape, generative AI (GenAI) has emerged as a disruptive technology poised to fundamentally redefine how organizations generate customer value, drive innovation, and establish a competitive advantage. Data now fuels the GenAI transformation, making it imperative for forward-thinking organizations to develop a holistic and unified approach to data and GenAI.

In a recent survey of global data and AI leaders, 76.1% of leading companies stated that they were at an “early stage” when it comes to the implementation of GenAI, with 23.9% indicating that they had implemented GenAI in production at scale within their organizations. At the same time, 89% of survey respondents said that AI would be the most transformative technology in a generation, with 61% stating that GenAI would be the most transformative form of AI.

Given the vast transformative potential of GenAI, there is a general lack of information and direction on how organizations can undertake the design and implementation of data and GenAI use cases. Current literature often presents a fragmented view of data and GenAI, either focusing on abstract theoretical frameworks or delving deep into complex technical implementations.

This disconnected approach has created a significant gap, leaving business leaders without clear guidance for creating a holistic, actionable, and integrated strategic perspective.

To bridge this gap, we propose a two-pronged framework that is designed to transform a Data and GenAI strategic vision into practical reality. The first prong of this framework focuses on a Data and GenAI Strategy Canvas, which serves as a comprehensive strategic mapping approach that empowers business leaders across all industries to articulate and visualize their strategy with precision and clarity. The second prong is an architectural blueprint for a modern data and GenAI Platform, translating the strategy canvas into a tangible implementation.

Together, this two-pronged approach enables organizations to navigate the complex landscape of data and GenAI with greater confidence and coherence, ensuring alignment between strategic vision and practical implementation to directly deliver business value from a company’s data and GenAI investments.

Developing a data and GenAI strategy

A modern data and GenAI strategy systematically explores six pivotal domains:

1. Business objectives
2. Governance
3. Access and Management
4. Producer and Consumer Needs
5. Organization and Processes
6. Infrastructure and Technology

To help business leaders translate these domains into an actionable transformation plan of aligned actions, we describe the Data and GenAI Strategy Canvas shown in Figure 1 below. This canvas addresses each domain individually and provides lists of the most important guiding questions business leaders should answer to articulate their data and GenAI strategy:

1. Business objectives focus on defining concrete and tangible business outcomes and mapping out concrete data and GenAI use cases to drive them. Many manufacturing companies, for example, seek to reduce their production line downtime by implementing a GenAI-based maintenance assistant that translates complex error images into step-by-step repair instructions for maintenance engineers to improve operational efficiency and productivity.

In this example, the reduction of the production line downtime serves as the key performance indicator (KPI) to measure the business impact of the GenAI maintenance assistant use case. This KPI allows business leaders to monitor progress while ensuring that financial investments in this use case directly contribute to measurable business value.

KPIs in general not only allow for monitoring progress but also for establishing effective reporting and escalation mechanisms, to prevent development teams from creating aimless use cases that drift without strategic purpose and measurable business impact.

Another important question that business leaders should address as part of the business objectives domain of their Data and GenAI Strategy Canvas concerns risks and challenges. Identifying and proactively mitigating risks and challenges associated with each use case is pivotal to ensure use case success. In the case of the GenAI maintenance assistant, business leaders may face risks associated with the quality of training data.

For example, if the data used to train the maintenance assistant does not cover the most common maintenance cases, the assistant will likely fail to provide methodologically correct step-by-step repair instructions to maintenance engineers, which increases the production line downtime and thereby directly impacts use case success.

2. Governance encompasses developing clear policies and guidelines for security, quality, privacy, and compliance. In this domain, business leaders leverage Ethical AI and Responsible AI principles as well as regulatory compliance frameworks, such as ISO 27001 (Information security, cybersecurity, and privacy protection) or HIPAA (Health Insurance Portability and Accountability Act), to facilitate responsible use of data and GenAI assets across their organization and its different jurisdictions.

Governing the development of varied enterprise-level data and GenAI use cases is particularly challenging for large-scale organizations as the large number of disparate use case development teams and variety of inconsistent legacy frameworks promote the development of “shadow-GenAI” solutions.

To address these complexities, forward-looking business leaders adopt systematic use case assessment protocols like the EU AI Act Compliance Checker owned by a Chief Security Officer. Such assessment protocols allow for systematically identifying risks inherent in all data and GenAI use cases and deriving appropriate guidelines to mitigate them.

When establishing policies around security, quality, privacy, and compliance of data and GenAI assets, a careful balance between risk mitigation and freedom of innovation is critical to ensuring organizational agility and unlocking the full transformative potential of data and GenAI.

3. Access and Management addresses the critical challenge of data and GenAI democratization and ensures that all assets are shareable, properly maintained, and readily available to authorized users. To establish intelligent access controls that enforce governance and regulatory compliance, business leaders need to know who needs access to which assets, such as data products or GenAI models.

The guiding principle in this context should be the principle of least privilege (PoLP), in which a user is given the minimum level of access needed to perform his or her job-to-be-done. This also involves a systematic analysis of associated access patterns, focusing on key metrics like read and write frequency, access volume, and timing, and then strategically structuring access management to balance regional and global considerations with use-case-specific requirements.

For example, an engineer using the GenAI maintenance assistant needs frequent access to the chatbot itself, but not to the data used to train the underlying GenAI model. To implement such access management, two important models have emerged in the past differing in their organizational implementation.

Centralized access management consolidates permissions, credentials, and authentication keys for all use cases within a single team, which can become a bottleneck when scaling the data and GenAI platform to many users.

Decentralized access management with self-service access, on the other hand, allows organizations to overcome these limitations. In the case of the GenAI maintenance assistant, an engineer can request access to the chatbot directly from the chatbot development or operations team, who is solely responsible for this particular use case.

4. Producer and Consumer Needs explores the nuanced requirements of internal and external end users. Internal and external users generally fall into three categories:

• Producers that create and share assets
• Consumers who want to use assets
• Platform team that runs the data and GenAI platform on a daily basis

The GenAI maintenance assistant, for example, is provided by producers, i.e. the associated development team. Consumers, such as maintenance engineers, can log into the platform and request access to the assistant on demand, while the platform team operates the underlying compute and storage infrastructure.

Having a clear picture of this process and the associated producer and consumer needs is pivotal for translating user requirements into technical capabilities that support the different use cases developed on the data and GenAI platform. Technical capabilities can either be general or use-case-specific as we will see below.

5. Organization and Processes recognize the critical role of organizational structure, effective processes, and cultural change necessary to execute the data and GenAI strategy. In this domain, business leaders define ownership and responsibilities across teams, which is paramount to maintaining strategic integrity and operational efficiency.

In our experience, the most efficient approach to implementing a data and GenAI platform involves two main teams:

• A central platform team that owns and manages general platform capabilities, such as compute and storage, that are shared across all use cases
• Decentralized use case development teams with end-to-end accountability for their respective use case – an approach that became known as “centralized governance and decentralized ownership.”

To maximize organizational agility and flexibility, both teams are structured into smaller and autonomous sub-teams that each own specific platform capabilities or use case components. This approach has been famously championed by Amazon and became known as “2-Pizza Teams.”

6. Technology and Infrastructure represents the final domain of our Data and GenAI Strategy Canvas, informing the overall architectural and technological design of the data and GenAI platform. In this domain, business leaders together with their technical experts define the concrete implementation roadmap including the overall platform design, list of general and use-case-specific capabilities as well as a (3rd-party) tool and vendor strategy.

By meticulously defining the vendor strategy, thought leaders provide clear guidance on the optimal mix of on-premise and cloud resources required to construct a robust, scalable, and cost-optimized data and GenAI platform.

In the Technology and Infrastructure domain, business leaders also explore the tool landscape, systematically evaluating and recommending both (licensed) proprietary and open-source tools and services. This holistic assessment enables business leaders to make informed decisions in technology selection and ensure alignment with strategic objectives, performance requirements, and budget constraints.

After the Data and GenAI Strategy Canvas is outlined, business leaders and their specialized teams — including data and GenAI strategists, platform and solution architects, and security consultants — can collaboratively incorporate the various requirements into a user-centric and scalable data and GenAI platform.

Implementing the data and GenAI strategy

A conceptual blueprint for a modern data and GenAI platform is shown in Figure 2 below. The left-hand side depicts three user groups of the platform introduced above. These users access the platform through a unified portal that provides a standardized user interface/user experience (UI/UX) and seamlessly integrates general and use-case-specific capabilities. By providing a central, intuitive access point, this portal has emerged as a key catalyst for user engagement and platform adoption, dramatically reducing complexity and improving user experience.

Through its core, the data and GenAI platform provides the following groups of general capabilities that consumers can use to build their use case, with modularity of capabilities and reusability of assets serving as two underlying principles for accelerating use case development and decreasing their time-to-market, also called time-to-production for use cases:

• Core infrastructure: General base capabilities such as compute, storage, and networking services required for every IT platform.
• Data capabilities: General capabilities needed to ingest, process, analyze, and enrich data as prerequisites to build data products.
• AI/ML capabilities: General capabilities related to AI/ML, such as AI/ML models, Foundation Models (FMs) trained on vast amounts of data that can be adapted to various use cases, and Large Language Models (LLMs) needed to build the GenAI maintenance assistant, for example.
• Marketplace (or “asset catalog”): Organizational knowledge hub, establishing a single source of truth for all assets. In this context, “assets” include not only raw data and data products but also application programming interfaces (APIs) for machine-to-machine communication and prompt templates.
  Prompt templates are essential for many GenAI use cases as they provide a structured, repeatable approach to generating consistent and high-quality outputs in alignment with Ethical AI and Responsible AI principles.
• Playground: Dedicated experimental (non-production) workspace that enables users to explore, prototype, test, and learn about AI/ML and other technologies. This controlled environment has proven essential for fostering agile innovation and cross-functional collaboration, allowing organizations to rapidly develop and refine AI/ML-driven solutions.

The overall platform design follows a service-oriented architecture based on APIs, where platform capabilities are implemented as highly decoupled services, sometimes called microservices. By breaking down complex services into smaller, independently deployable services, organizations can achieve greater flexibility, seamless scalability, and operational resilience.

This modular approach enables faster innovation, more frequent updates, and easier troubleshooting, as teams can focus on individual services without navigating complex, tightly coupled services based on monolithic codebases.

APIs not only enable the connection of different microservices, but also the implementation of Agentic AI, advanced AI/ML systems capable of interpreting the user input, setting goals autonomously, and taking proactive decisions by calling the APIs of other microservices needed to execute the user’s input or command.

This highly flexible and service-oriented data and GenAI platform approach allows organizations to progressively expand their general capabilities and incrementally add new use cases addressing evolving use and business cases in alignment with the overarching data and GenAI strategy.

Conclusion

The integration of data and GenAI requires a comprehensive strategic approach to drive innovation and competitive advantage. The proposed framework combines a Strategy Canvas covering six essential domains (business objectives, governance, access and management, producer and consumer needs, organization and processes, and infrastructure and technology) with a practical data and GenAI platform design blueprint.

This dual approach enables organizations to translate strategic vision into actionable implementation through user-centric, scalable, and service-oriented architecture with self-service capabilities. By following this framework, organizations can effectively navigate the data and GenAI landscape, unlocking unprecedented opportunities for innovation, efficiency, and competitive advantage.

References:

• A Practical Guide to Building Ethical AI
• How to Implement AI — Responsibly
• How to manage two types of generative artificial intelligence
• EU AI Act Compliance Checker
• Two Frameworks for Balancing AI Innovation and Risk
• Know Your Customers’ “Jobs to Be Done”
• Centralizing or Decentralizing Generative AI? The Answer: Both
• Powering Innovation and Speed with Amazon’s Two-Pizza Teams
• A Better Way to Put Your Data to Work
• AWS Summit Berlin 2023: The Amazon Software Development Process
• What Is Agentic AI, and How Will It Change Work?

About the Authors:

Randy Bean is the author of Fail Fast, Learn Faster: Lessons in Data-Driven Leadership in an Age of Disruption, Big Data, and AI, and a contributor to Forbes, Harvard Business Review, and MIT Sloan Management Review. He has been an advisor to Fortune 1000 organizations on data and AI leadership for 3+ decades.

Volker Lang is the author of Digital Fluency and is a Strategic Program Leader with Amazon Web Services in Munich. Lang holds a Doctor of Philosophy in Materials Science from Oxford University with a specialization in Quantum Physics. He previously held technology leadership positions with Audi and Volkswagen AG.