How do we choose the right AI project to start with?

Aligning AI with Business Strategy and Outcomes

The journey to successful AI implementation begins not with technology, but with a clear understanding of your business objectives. It is crucial to define a business- and customer-centric north-star for AI adoption and underpin it with an actionable strategy that moves step by step. This strategy should be based on tangible business impact, whether it's reducing business risks, growing revenue, increasing operational efficiency, or improving environmental, social, and governance (ESG) performance.

It is important to factor in both short-term and long-term impacts of adopting AI. Rather than asking "How can we use generative AI?", the more effective question is "What use cases do we need the most help with, and what role could different areas of technology and data analytics play?".

Work backward from existing business and customer problems and the effect that AI can have on them, defining your expected business outcomes over time. This approach minimizes the chances of building data capabilities for their own sake and ensures investments have the power to move the metrics that truly matter.

Identifying High-Value Use Cases

Once the strategic direction is clear, the next step is to identify and prioritize high-value AI products and initiatives that are feasible. Organizations tend to adopt use cases in a specific order, and starting with projects that are too challenging or offer only small, long-term benefits can stifle enthusiasm and slow down transformational changes. Instead, focus on use cases that are easier to achieve or have a proportionally faster or higher return on investment to create momentum for further investment.

Examples of high-impact applications include cloud pricing optimization, voice assistants and chatbots, predictive maintenance, and uptime/reliability optimization. These pilot initiatives should be highly impactful on the organization and the business, meaningfully benefiting from AI. Learning from these initial projects helps adjust the approach before scaling to full production, enabling broader, sustained value. As you consider these opportunities, also determine what data assets and sources these initiatives rely upon, working backward from opportunities toward data requirements.

Assessing Feasibility and Readiness

Before committing to an AI project, a thorough assessment of feasibility and readiness is essential. This involves evaluating if the necessary technological capability can be established and if existing talent can be enabled to use and adopt AI. Crucially, it means investigating the data requirements, considering aspects like volume, velocity, variety, and veracity (the 4 V’s of Data). It's important to ask whether you need to build, tune, or adopt an existing AI system. Many organizations, especially those with fewer years of AI experience, can jump-start their initiatives by acquiring AI as a product or service, rather than building every solution from scratch. This can involve consuming prebuilt off-the-shelf AI services, low-code/no-code functionalities, or customizing foundation models for specific needs if in-house capabilities for creating them from scratch are lacking. Involving business, data, executive, and machine learning stakeholders in this assessment is vital, as machine learning products fuse data, domain, and technology into one predictive system.

Managing Costs and ROI

AI initiatives, particularly in their early stages, require careful financial planning. It is important to plan for the cost structure of training and inference from the outset when budgeting for individual projects and the overall funding of AI. While proof-of-concept initiatives might be relatively low-cost compute-wise, the training of larger models or constant retraining for domain-specific models can quickly become costly. Organizations should connect their AI initiatives to an underlying business goal, translating expected business value into AI metrics. This means calculating what an incremental improvement in an machine learning metric is worth to optimize investment. Additionally, the often-underestimated cost of not recognizing the need for responsible use of AI should be factored in, as it can lead to significant mid-term damages if overlooked. Hybrid cloud architectures, for instance, can be essential for scaling AI goals at manageable costs by allowing organizations to use the most cost-effective infrastructure for every workload.

Ethical and Governance Considerations

AI introduces new risks and ethical dilemmas, making it imperative to embed responsible AI practices from the beginning of the AI journey and throughout its lifecycle. This includes risks related to data privacy, bias, and unexpected outcomes. Organizations should establish an AI governance board with representation from multiple business units, such as human resources, legal, and regulatory affairs, to oversee and guide the ethical development and deployment of AI technologies. This board ensures alignment with industry regulations and compliance with AI-focused legislation. Considering how the system affects individuals, subgroups, users, customers, and society at large is paramount. Embedding explainability by design, establishing practices to recognize and discover biases, and using tools to monitor the status quo are critical steps. This proactive approach helps mitigate damage and fosters continuous AI innovation through responsible practices.

Leveraging External Expertise and Partnerships

Companies do not need to possess all AI expertise in-house to embark on their AI journey. The talent gap for professionals who combine deep AI knowledge with specific business and cybersecurity expertise is significant, making these roles expensive and difficult to fill. This is where external resources become invaluable. Strategic partners, such as consultants and specialized technology providers, can offer advisory, consultation, and program management, bringing specialized knowledge in various AI use cases and best practices to accelerate adoption. For common AI use cases that do not offer competitive differentiation, outsourcing to suppliers with specialized AI software is a sensible strategy. This can include co-development engagements with vendors or utilizing Business Process Outsourcing (BPO) providers to automate non-core processes. Cloud service providers also play a crucial role, offering scalable resources and managed machine learning services that reduce the need for extensive in-house infrastructure management, allowing internal teams to focus on strategic initiatives.

Fostering an AI-Ready Culture

Ultimately, AI is largely about augmenting human labor, not replacing it. It enriches, supplements, and enhances human capabilities, freeing employees to focus on more complex, meaningful, and creative tasks. Organizations can transition existing talent into AI-related roles, leveraging their domain knowledge and fostering an "AI-first" culture. This involves empowering employees to experiment with AI systems and providing structured learning paths, from beginner to advanced, to build AI literacy across all functions. Low-code and no-code platforms enable even non-technical employees to create or modernize applications with AI, streamlining processes and fostering innovation. While deep technical expertise might be outsourced, internal leaders must bridge the gap between technology and business objectives, defining the AI vision and identifying high-value use cases. Their role is not to be AI scientists, but to understand the "art of the possible" with AI and identify its main risks, fostering cross-functional collaboration and managing expectations.

By embracing a pragmatic approach that combines clear strategic objectives, robust governance, a mix of internal upskilling, and external partnerships, companies can effectively adopt AI, drive innovation, and achieve measurable business value without needing to build an extensive in-house AI expert team from day one.

Talk to us to explore your options.