The Biggest AI Mistake is Treating Knowledge Like Data | Opinion
Newsweek · C · trust 30/100
0 Share Newsweek is a Trust Project member See more of our trusted coverage when you search. Prefer Newsweek on Google to see more of our trusted coverage when you search. Companies have spent billions racing to adopt artificial intelligence. Many assume the next breakthrough depends on larger models, faster infrastructure or more data. However, they are solving the wrong problem. The greatest obstacle to enterprise AI is not technology. It is that businesses still treat knowledge as if it were just another form of data.
That assumption is quietly undermining AI strategies across industries. Companies are investing in increasingly sophisticated systems while feeding them information that was never designed, governed or maintained to support intelligent decision-making at scale. Bigger models cannot compensate for unreliable knowledge. They can only process it faster.
A recent conversation reinforced this point for me. The chief executive of a rapidly growing public company reached out to me. His team was investing heavily in data ingestion for AI development and wanted to understand how organizations could prepare information for these systems. It was an important question because it reflects the conversation happening inside companies everywhere: how much information can we connect, how quickly can we ingest it, and how soon can we unlock AI-driven value?
But the question underneath that question is the one most companies are avoiding: does the organization actually understand, govern and trust the knowledge it is giving to AI?
For years, businesses have treated data governance as a strategic priority. They know financial data requires controls. Customer data requires security measures. Operational data requires consistency. These disciplines exist because organizations understand that unreliable inputs create unreliable outcomes.
Written knowledge has rarely received the same level of scrutiny. That is the mistake. Product specifications, support articles, internal procedures, training materials and technical documentation often exist across multiple systems with different owners, inconsistent updates and no clear source of authority.
This distinction deserves far more attention than it receives. Data enters enterprise systems through rigid validation rules, defined fields, standardized formats and carefully managed processes. Language follows a completely different path. Documentation is written by different teams, revised over many years, copied into multiple repositories and interpreted in contexts that often exist only in the minds of subject matter experts.
Companies do not fail with AI because the models are weak. They fail because they have spent decades neglecting the quality of the knowledge those models depend on.
A technical manual, a customer support article, an internal process document and a product roadmap all communicate meaning in different ways. Human readers understand the assumptions between the lines. AI attempts to interpret those same signals. When that content is inconsistent or poorly governed, uncertainty spreads throughout every response the system generates.
Evidence suggests this challenge extends far beyond isolated examples. An industry report shows that 23 percent of organizations still have no formal knowledge management structure, while nearly half rely on hybrid models that combine centralized and decentralized approaches. The report also describes a fragmented landscape filled with overlapping repositories and disconnected systems that complicate how knowledge is created, maintained and discovered. For executives investing heavily in enterprise AI, this should be a warning: fragmented knowledge inevitably produces fragmented intelligence.
Not all organizational knowledge deserves equal treatment. Companies should identify and isolate the high-value content that defines the unique value they bring to market. This is the knowledge that should become the foundation of AI: product documentation, user guides, specifications, training materials, internal playbooks, policies and processes. This is the infrastructure your AI will rely on to help run your business. Invest there first.
The same pattern appears in broader AI adoption. The organizations pulling ahead are not simply buying better technology. They are building the operating disciplines that make AI useful: clear governance, accountable ownership, coordinated execution and workforce readiness. AI agents are proving highly effective at automating workflows, yet many organizations still struggle to coordinate decisions across functions.
That gap is predictable. AI cannot consistently coordinate decisions when the knowledge behind those decisions is inconsistent. Automation moves quickly. Organizational understanding requires ownership, discipline and trust.
I have spent much of my career watching organizations solve increasingly complex information challenges. One lesson continues to repeat itself across every technology cycle. Businesses rarely suffer from a shortage of information. They suffer from an inability to turn expertise into durable knowledge that remains trustworthy after the expert has left the room. AI has not created this weakness. It has exposed it.
Highly regulated industries have already demonstrated that another path is available. Pharmaceutical companies, life sciences organizations, aerospace manufacturers and other heavily regulated sectors have spent years building disciplined content supply chains because the consequences of inaccurate information are immediate and severe. They know where critical information originates. They assign ownership, establish review cycles and document approvals. Moreover, they maintain version history and update knowledge as regulations, products and procedures evolve.
None of these practices emerged because AI demanded them. They became essential because reliable knowledge became a business requirement. Enterprise AI now faces…
Read the original at Newsweek →
Open in TruthVane →