Knowledge management has long been one of those fuzzily-defined areas of the business that is essential, often with a foot in information technology and a foot in business management, but never really in the spotlight of either. Its proponents and overseers represent a range of job roles, from librarians to content management systems specialists to knowledge managers by title.
These professionals handle documents such as software documentation, user manuals, repair guides, contracts, and much more. But, think about it: data scientists are the hot job for today’s economy, but what do we hear about “content scientists?”
In an era when unstructured data — content — is seen as the most critical fuel for the artificial era ahead, it’s time knowledge management had a seat at the table.
Yes, AI can boost knowledge management itself to new levels, helping to automatically capture and analyze unstructured information from across the enterprise. But knowledge management needs to be a key piece of overall AI strategy as well.
“AI is only as powerful as the knowledge it’s built on, said Vele Galovski, vice president of TSIA. “Even the most advanced AI tools can fall short without a strong knowledge management foundation.”
Foundation is a key word here, and that view has been reinforced in a recent analysis of enterprise intelligence systems, published by Bloomfire. Part of the challenge with AI that it’s proliferation is divided between three separate technology categories – knowledge management, enterprise search, and business intelligence, the study’s author, Anthony J. Rhem, PhD, points out. “Knowledge management platforms captures what people know, enterprise search helps them find it, and business intelligence systems makes sense of the numbers. Each was built to solve a different problem, governed by a different team, and measured against a different set of outcomes. The assumption that they’d converge into something coherent on their own was never realistic.”
This disconnect is resulting in off-kilter AI initiatives. “Most enterprise AI is underperforming, and the cause is rarely the underlying language models,” Rhem states. “Organizations are turning to countless software tools to break through and achieve the return on their AI investments, but are often left wanting.”
The best approach for well-rounded AI is a to seek a fusion, or tight integration between knowledge management and IT areas – not easily done, as these represent separate roles or departments, but desperately needed. “The primary risk zone comes from a failure to appropriately embed the knowledge foundation with a data and AI backbone,” says Rhem.
The result of such disconnect? More than half of employees bypassed their company’s AI tools in the past 30 days, completing work manually instead, the Bloomfire study suggests. While AI budgets have been increasing a clip of 38% a year, 40% of that spend is underperforming, executives said. Three in four executives even admit their AI strategy is “more for show than substance,” the study states.
Bloomfire’s own analysis of its content showed 31 concepts explored, with zero documents that actually define them. “Ten documents referenced it, but none explained it.”
What’s at issue? Rhem outlines how and where organizations are missing the boat:
- Underinvestment in unstructured data and knowledge management. “Most organizations have focused technology investments on structured data and analytics while underinvesting in the tools needed to capture, organize, govern, and retrieve unstructured content and tacit knowledge,” Rhem writes.
- Over-reliance on search. Enhanced inability does not translate into decision support.
- Over-reliance on dashboards. Dashboards show what happened, but they rarely explain why or connect insight to the knowledge needed to act,” Rhem observes.
- Over-indexing on AI alone. “Copilots and agents built on weak content, weak metadata, and weak governance produce fast answers with inconsistent trustworthiness.”
- Treating aggregation as intelligence. “Intelligence requires relevance, explanation, governance, and actionability,” Rhem urges.
- Ignoring tacit knowledge. “Many tools handle explicit content better than they capture know-how, decision rationale, and institutional memory.”
- Selecting point tools before defining the target operating model. “This creates a disconnected architecture and a patchwork user experience.”
Rhem’s report also provides some recommendations for getting an intelligence platform in order:
- Make knowledge your foundation. A well-governed knowledge foundation is a good place to start. Connect solutions ”into the collaboration, service, operational, and decision workflows where employees work,” Rhem urges. “Build the knowledge foundation before layering on search, analytics, copilots, and agents. Trusted content and accountable governance are prerequisites, not afterthoughts.”
- Be selective. Go with platforms “based on how well they work together across the full operating model, not on which single vendor appears strongest in a product demonstration or one evaluation category,” Rhem advises. “Prioritize platforms with strong governance, metadata, and integration over highly specialized point tools that solve narrow problems without contributing to the broader architecture.”
- Rethink how AI investments are planned, deployed, and measured. Emphasize decision effectiveness as the primary objective. Aim for “context-aware insights and recommendations at the moment decisions are made.”
