Extracting value from AI? Surely you are joking

Welcome to a new edition of The Board: Distillation Aftershots ^(*).

This is an online of my newsletter where I share curious and interesting insights and data points distilled from enterprise technology to identify what’s notable. If you are interested in getting this newsletter and don't miss any of my content, please subscribe.

In this issue, we explore why AI delivers value unevenly across the enterprise, what is actually working today, and how organizations should approach extracting value over the long term. The current wave of generative and agentic AI has sped up investments and expectations, but the path to lasting enterprise impact is proving more structured and slower than expected.

First, my take.

There is value in AI, but it doesn't come from widespread use of general models or standalone generative or agentic AI. Instead, it emerges through three intentional paths:

1. Treating AI as a capability built on top of a well-orchestrated hybrid infrastructure with executive ownership
2. Redesigning processes to align with how AI systems operate, and
3. Positioning AI as part of a broader move toward end-to-end automation.

Organizations that approach AI this way are beginning to see measurable results. Those that do not are gathering experimentation without achieving scale. This is not a short-term process; for most enterprises, fully realizing AI's value will require nearly a decade.

Recent research shows a consistent pattern: the main barriers to AI value are structural. Most organizations still try to add AI onto existing systems and processes, which limits their impact. The successful companies are instead aligning architecture, governance, and capital investment around AI as a long-term capability, often with direct executive ownership to ensure deployment decisions support business goals rather than isolated use cases.

At the operational level, value is driven by process redesign. Current AI systems don’t perform well when used in fixed workflows without adjustments. Organizations that see results are reorganizing how work is done—streamlining steps, redefining decision points, and using human oversight more strategically. This change reflects a broader shift in how enterprise processes are viewed, moving from linear models to more flexible, data-driven systems.

Strategically, AI is proving most effective when positioned within a broader automation agenda. Software is evolving toward service-oriented, automated delivery models, and AI is acting as an enabling layer rather than the endpoint. Generative and agentic AI, while valuable, are not yet capable of delivering consistent enterprise outcomes at scale without significant constraint, customization, and integration into controlled environments.

The resulting dynamic is a mismatch between expectation and execution. The technology is advancing quickly, but enterprise environments need predictability, repeatability, and governance. Closing that gap takes longer than expected, due to infrastructure complexity, cost factors (tokenomics and inference optimization are becoming popular topics), and the necessity for organizational change. AI adoption is happening more as a gradual restructuring rather than a quick transformation, with significant value building up over time as enterprises align systems, processes, and strategy.