Why AI Isn't Autonomous (Yet), Part I
Why Data Alone Isn't Enough
Introduction
Agentic Artificial Intelligence is upon us— or so they say. But what does that really mean? Agentic AI refers to systems that can perceive their environment, reason about goals, plan and execute multi-step actions, and learn from feedback, adapting their behavior without constant human direction. In other words, these agents possess agency: the ability to make decisions in dynamic, real-world contexts-—a significant step beyond today’s largely reactive, prompt-driven models.
According to IBM, ”Agentic AI is focused on decisions as opposed to creating actual new content, and doesn’t solely rely on human prompts nor require human oversight. Early-stage Agentic AI examples include things like autonomous vehicles, virtual assistants, and copilots with task-oriented goals.”
The new AI race, driven by a fever-pitch marketing bonanza, is hyper-focused on staking the prodigal golden spike into the ground to stake claim victory in achieving Agentic AI. Yet most of the industry is still trying to figure out how to structure and prepare their data just to find any success with generative AI. The likelihood of the typical company achieving effective Agentic AI in the foreseeable future remains low..
AI is a knowledge tool, not a data tool. Until we face this reality head-on, Agentic AI will remain a dream, as companies continue to wrestle with the atomic units of knowledge: data.
AI: A Knowledge Tool
AI, at its core, is a knowledge tool, whether generative or agentic. Yet, while knowledge workers toil to untangle data infrastructures and define semantic layers, vast swaths of corporate and institutional information remain untouched, unstructured, and inert.
Ontologies and knowledge graphs are emerging as tools that promise to enrich data ecosystems with information and knowledge. But the premise of this plan is flawed. If data is the atomic molecule of information, and if information is the foundation of knowledge, then much of the tech industry is simply stirring the pot— hoping that, with enough heat, something will coalesce and magically take form as knowledge.
If AI systems require knowledge to function meaningfully, why are we so reluctant or afraid to build true knowledge infrastructures? And what does knowledge infrastructure even look like?
The Knowledge Infrastructure
The Culture of Knowledge
The concept of knowledge infrastructure is hardly new. Cultural institutions like libraries and archives have long relied on both digital and physical knowledge systems to organize, preserve, and share knowledge. That these institutions operate across both realms highlights the complexity of capturing and modeling knowledge in ways that make data and information findable, interoperable and machine-readable.
So why not learn from the masters? Librarians and archivists live and breathe knowledge infrastructure. Trained in knowledge management frameworks and methodologies, they approach the work with a service-minded approach— supporting both human and machine access to knowledge.
While library and enterprise digital ecosystems share some of the same core elements, their structures and priorities often diverge significantly.
At the heart of any knowledge infrastructure lies a culture of sharing. The defining feature of knowledge management is just that: knowledge sharing— the true engine of innovation. But the success of these initiatives depend on an organization’s ability to foster participation, break down silos and discourage the hoarding of critical data, knowledge and insights. Achieving this requires open, interdisciplinary collaboration, where teams across departments contribute their collective expertise to a shared, evolving knowledge base.
Building and sustaining a knowledge infrastructure requires more than goodwill; it demands the right blend of tools, processes, systems, and cultural enablers. Intuitive collaboration platforms, formalized taxonomies, metadata-driven repositories, ontologies and clear governance frameworks streamline the capture, transfer, and reuse of knowledge. By embedding these sharing mechanisms into daily workflows, organizations create continuous, cumulative learning loops that diffuse expertise broadly— enhancing intellectual capital and securing a lasting competitive edge in an evolving AI landscape.
A knowledge infrastructure cannot subsist on data alone. Information and knowledge are layered, requiring dynamic and descriptive structures. Unlike raw data points, knowledge includes context, relationship, interpretations, and evolving insights that require flexible systems to capture and connect these elements meaningfully. Current data infrastructures are rigid by design, built to manage copious amounts of data quickly. For example, a customer address in a database may include just a street, city and zipcode. But rich context would link that address to a customer's purchase history, preferences, social connections, regional regulations, and even local market trends— turning static data into actionable data. Without this depth and connectivity, organizations risk missing critical insights and stalling innovation.
All Hail Wiki Knowledge Bases & Linked Data
The magic of AI lies in its ability to generate rich, descriptive responses to prompts. These responses depend upon training data and reasoning engines, which are primarily built from structured information and knowledge sources such as Wikipedia and its affiliated sites. Wikidata has become the backbone of Wikipedia, serving as the authority source for verifying facts and claims.
