Navigating the GenAI Landscape

In the rush to harness the power of AI, companies are heavily investing in evaluating and probing AI solutions to grapple with the transformative nature of GenAI. AI has the potential to enhance operational efficiencies and act as a springboard for new products and services. However, its integration can be disruptive, often necessitating organizational transformations as it impacts everyone from developers to end-users. Consequently, organizations should adopt a systematic approach to implementing AI capabilities, measuring and adjusting their strategies based on its adoption and its influence on business outcomes.

In this extended blog post, we discuss how to systematically tackle AI transformation, reshaping the operational model, enhancing efficiency, and unlocking additional value propositions. This serves as a thought exercise to assist in formulating an AI strategy that aligns with industry trends and adoption levels. Given that AI is a rapidly evolving domain, the insights here should be viewed as current reflections, subject to change as the landscape shifts.

Use-case based approach

Executing a company’s vision for AI requires a framework that, at its core, drives innovation through a mechanism to create ideas, prioritize them, test or fail proof of concepts, and execute select minimum viable products (MVP) towards a programme of transformation and industrialization.

On the other hand, AI is not simply a solution in search of a problem. Instead of trying to force AI services and models into your organization without a clear plan, it’s important to first identify your business requirements and then match them with the right AI solution. The use-case based approach shown in the figure below has been drawn with this strategic intent in mind.

In broad strokes, AI use cases can be categorized into four primary categories: using AI on isolated systems, automating existing capabilities with AI, expanding your system to incorporate AI-driven innovations, and completely reinventing to build an ecosystem powered by AI. As you move from the bottom to the top of the table above, both disruption and transformation efforts intensify. Yet, the higher you climb on the table, the larger the value proposition becomes, allowing you to maximize the benefit.

Similarly, these use-cases can be addressed through four types of AI solutions: off-the-shelf foundational models, fine-tuned foundational models, task/domain-specific models, and custom-built models. As you move further to the right, the complexity and cost of the solution rise. However, the accuracy and efficiency of your solution also improve.

This section isn’t intended to create an exhaustive use-case matrix, but rather to guide your thinking and assist you in developing a use-case based approach to AI tailored to your organization’s needs.

Roadmapping AI Adoption

With limited funding and resources at hand, it’s vital to prioritize effectively. Categorize your use-cases into short-term, medium-term, and long-term timeframes. Then for each use-case, identify the required capabilities to support the solution and allocate the needed resources and funding. The graph below depicts these use-cases based on Geoffrey Moore’s zoning strategy.

For those unfamiliar with the zoning strategy, here’s a brief overview:

• Performance Zone — This aligns with your current capabilities or those that will be available shortly.
• Transformation Zone — This pertains to your medium-term goals. You should be actively developing these capabilities, with the aim to shift them to the Performance Zone soon.
• Incubation Zone — This represents your long-term objectives. Use cases in this zone may require a longer development phase, maturing first to the Transformation Zone before reaching the Performance Zone.
• Productivity Zone — This zone houses the capabilities necessary to support the use cases in the Performance Zone.

The goal is to continually building your capabilities in the Productivity Zone, enabling more AI use cases to move from Transformation and Incubation Zone to the Performance Zone. As such, you need to continually invest the right amount for each of these zones to stay ahead.

In this example, established capabilities (evaluation of vendor AI, prompt engineering techniques, AI skills and Training) in the “productivity zone” will help the organization to adopt their choice of AI services like GitHub Co-Pilot, CodyAI, and SalesForce Pardot.

Key Capabilities

Start early to establish the following key capabilities to sustain your AI adoption. It’s highly likely that the use of AI will become regulated across all industries to ensure fairness, transparency, accuracy, and ethics. Even if you are a small organization, having these frameworks will give you a head-start and place you ahead of the curve. Each of these capabilities is vast and requires in-depth reading and research to tailor them to your organization’s needs.

Remember that the primary goal of the data science and AI organization is to deliver business value quickly, ensuring solutions optimize processes, enhance decision-making, and free skilled resources to focus on more valuable activities. Your capabilities and processes should strive to achieve this primary objective for the company.

• AI Operating Model — Offers the overall strategy and management direction, detailing the insights needed and their significance. This can be set as a company-wide directive.
• AI Delivery Model — Focuses on how AI capabilities will be delivered and utilized. This can remain flexible, varying per project or business units to better suit their needs. Defining your processes, technologies, and tools at the delivery model level will facilitate scalability.
• Model Risk Management Framework —Focus on defining the model risk management activities ahead of your AI solution development. You should evaluate and produce AI Fact sheets to clarify the model use cases. This stage is pivotal for instilling fairness, transparency, accuracy, and bias detection.
• Data & AI Governance Framework — A crucial function that ensures the quality, integrity, and security of the data are upheld across various data stores and during the data processing. Rigorous management, and potentially tracking of data accessibility and usage, will be essential. Incorporate capabilities to assess the ethics and explainability of the solutions.
• Data Stewardship & Auditing Framework — The availability, timeliness, quality, privacy, and completeness of data are fundamental considerations that will determine the value delivered to the business.

In the fast-paced realm of GenAI, the challenge isn’t just about harnessing AI’s transformative potential, but doing so in a manner that’s strategic, practical, and sustainable. AI, as emphasized throughout this discourse, is not just a tool but an evolving ecosystem. Diving headfirst without a well-thought-out strategy is akin to setting sail without a compass. Thus, the frameworks and strategies discussed here aim to be that compass, guiding organizations towards maximizing AI’s promise while being adaptable to its ever-changing nuances. With the right approach, rooted in use-cases and backed by robust capabilities, AI’s horizon isn’t just promising; it’s achievable. Organizations, big or small, need only to be methodical, proactive, and aligned with their core objectives to make the most of the AI-driven future.