AI Readiness: It's Not What You Think

Charlie Dai, vice president and principal analyst at Forrester, started his keynote at the recent 7th Chief Digital & Data Officer Hong Kong Summit with an uncomfortable truth about enterprise AI. He noted that we're still grappling with the same hurdles we faced half a decade ago.

Data silos, exorbitant costs, geopolitical tensions, and the shadowy rise of "Bring Your Own AI" are all conspiring to keep real, transformative AI out of reach for many organizations.

But Dai also offered a lifeline: a four-step roadmap to navigate this treacherous landscape and finally unlock the power of AI. This isn't about chasing the latest shiny algorithm but building a solid foundation for sustainable innovation.

Step 1: Go foundational (But no, not that kind)

Forget large language models for a moment. Before you can even think about harnessing the power of AI, you need a solid foundation of AI readiness. This starts with a clear vision and a holistic approach to your data and AI ecosystem.

“Previously, data management and AI were two different worlds,” Dai explained. “Database administrators and data engineers worked on their stuff, while data scientists worried about model training and inferencing.” This siloed approach is a recipe for disaster.

Instead, Dai advocated for a federated organization that combines all the key players — from infrastructure engineers to application developers — under a unified data and AI strategy. This requires a full-stack approach, encompassing everything from AI infrastructure and platforms to security and governance. Think of it as building a robust nervous system for your organization capable of handling the complex flow of data and AI insights.

Modernizing your data management strategy is crucial. Dai highlights five key trends: global data management, end-to-end integration, AI-enabled processes, use-case-driven approaches, and converged data platforms. Global data management tackles the pervasive issue of data silos.

“Right now, we have data everywhere,” Dai warns. “We need a logical view of everything within your organization.” This calls for tools like data fabrics and lakehouses to create a unified data backbone across your entire enterprise, spanning public clouds, private data centers, and even the edge.

End-to-end data management requires integrating disparate data management functions into a cohesive whole. “In the past, you might have had many different data management functions within your enterprise,” Dai says. “Now, you need to pull them all together.” This requires integrated policies, metadata management, and an underlying architecture supporting end-to-end functionality.

Finally, converged data platforms are essential for taming the complexity of the modern data landscape. “Every enterprise has at least 20 different data management tools,” Dai observes. This proliferation of tools creates a management nightmare. Converged platforms offer a solution by consolidating functionalities, such as combining document databases with graph databases for multi-model capabilities or merging data lake and data warehouse functionalities into a lakehouse architecture.

Step 2: Understand the impact (beyond the buzzwords)

Once you have a solid foundation, it's time to look beyond the AI hype and understand its real impact on your business and technology. Dai identifies several top-of-mind use cases for organizations worldwide: content creation, customer service augmentation, knowledge management, document automation, developer workflow optimization, natural language interfaces, and industrial automation.

However, Dai cautions against a one-size-fits-all approach. “You also have to think about what is most relevant to your business context,” he emphasizes. For instance, while customer service augmentation might be a universal need, its specific implementation will vary drastically between a financial institution and a healthcare provider.

Furthermore, the impact of AI extends beyond data and algorithms. “Your data architecture and AI architecture can be dependent upon your cloud architecture,” Dai points out. This underscores the need for a cloud-native, and ultimately AI-native, approach to infrastructure.

Step 3: Learn the tech

Now comes the fun part: diving into the technical intricacies of AI. But Dai warns against getting fixated on the latest buzzwords. “It's not about adopting one single model,” he clarifies, referring to the current obsession with foundation models. “It's about the adoption of a whole ecosystem.”

This ecosystem encompasses a range of critical considerations, from model performance and modality support to cost-effectiveness, licensing models, and local availability. Dai stresses the importance of evaluating foundation models based on a comprehensive framework that goes beyond simple benchmarks.

Moreover, mastering the AI model lifecycle is crucial. This lifecycle, Dai explains, is more nuanced than traditional AI development, encompassing stages like data preparation, pre-training, fine-tuning, and application development. Each stage demands specific technical expertise and a deep understanding of the underlying processes.

Dai also highlights the importance of prompt engineering, a critical skill for effectively interacting with and guiding AI models. “Your data scientists don't always know how to speak human language,” he points out, “and your business stakeholders don't care what SQL means.” Prompt engineering bridges this gap, facilitating collaboration between technical and business teams.

Retrieval augmented generation (RAG) is another essential technique for grounding AI models in real-world knowledge. “RAG has become the most critical area if you really want to do something in real business production environments,” Dai asserts. By integrating authoritative knowledge sources, RAG helps mitigate the risk of hallucinations and ensures that AI models produce accurate and reliable outputs.

Finally, Dai touches on the rise of "TuringBots," AI-powered developer tools that automate various aspects of the software development lifecycle. These tools, Dai explains, go beyond simple code generation, assisting with tasks like application design, deployment, testing, and productivity analysis.

Step 4: Get your hands dirty

You've built the foundation, understood the impact, and mastered the technology. Now, it's time to combine it and apply AI to your business context. Dai emphasizes the importance of platforms, practices, and partners in this crucial step.

The AI platform architecture should encompass both foundation models and traditional AI models. “You must retain the models you have been investing in for the past decades,” Dai advises. This requires a platform capable of managing both types of models, along with their associated data, across the entire lifecycle. Furthermore, a full-stack approach, encompassing infrastructure, development tools, and applications, is essential.

On the practice front, Dai highlights three key areas: micro-augmentation, business process re-engineering, and addressing shadow AI. Micro-augmentation focuses on embedding AI capabilities into every facet of your business, empowering all lines of business with intelligent automation. Business process re-engineering is equally crucial. “If you have all the automation capabilities in the AI agent but still have to go through a very complex, antiquated business process, you won't be able to maximize the business outcome,” Dai warns.

Addressing shadow AI requires robust security, privacy, and compliance measures. Dai stresses the importance of evaluating partners based on their security posture and ability to protect sensitive data throughout the AI lifecycle. He also emphasizes the need to stay ahead of evolving privacy regulations and ensure compliance throughout your AI initiatives.

Finally, partnering with trustworthy vendors is crucial for navigating the complex AI landscape. Dai urges organizations to look beyond technical expertise and seek partners who can act as true business advisors, providing guidance and support throughout the AI journey.

AI FOMO is warranted

Dai's four-step roadmap offers a pragmatic and refreshingly hype-free approach to AI readiness. It's a call to action for organizations to move beyond pilot projects and embrace AI as a core driver of business transformation.

It also requires a fundamental shift in mindset, a willingness to invest in foundational capabilities, and a commitment to continuous learning and adaptation. The AI revolution is not a spectator sport. It's time to get in the game.

Image credit: iStockphoto/Bogdan Malizkiy