When MIT Sloan researchers examined why so many companies were failing to derive value from generative AI pilots, the problem wasn’t the models; it was the data. The study found that most enterprise AI agents lacked the necessary context to make sound decisions, often because the institutional knowledge they required was scattered, outdated or simply missing.
“Organizations risk falling into the ‘AI learning gap’ where tools don’t learn from feedback or fit with existing processes,” Paul McDonagh-Smith, senior lecturer at MIT Sloan School of Management, recently told Newsweek. “Many pilots use generic AI models that work in isolation but don’t readily adapt to existing and emerging workflows, resulting in employees abandoning AI when outputs prove unreliable or unworkable in their daily operations.”
A report from Bloomfire, The Value of Enterprise Intelligence, puts numbers to that blind spot. After surveying more than 10,000 professionals across 115 companies, the firm concluded that inefficient knowledge management affects a quarter of annual revenue, roughly $2.4 billion for a typical Fortune 500 company. That’s money lost not to bad algorithms, but to bad information hygiene.
“Despite massive investments in digital transformation, many organizations still struggle with the ‘last mile problem’: connecting employees to the knowledge they need, when they need it,” Philip Brittan, CEO of Bloomfire, told Newsweek. “Information remains scattered across siloed systems, slowing decisions and frustrating teams.”
In other words, the hidden asset is knowledge itself—the connective tissue linking people, data and decisions.
Turning Knowledge Into Capital
For decades, companies have tracked financial and physical assets down to the penny while letting intellectual capital drift unmeasured. Brittan calls it a costly paradox.
“Consider this paradox: while CFOs meticulously track financial assets … knowledge, often worth billions, remains entirely absent from balance sheets,” he said. “No executive stands up during earnings calls to report on ‘knowledge debt’ or ‘intellectual capital returns.’ This oversight has real consequences.”
Bloomfire’s data underscores the point: Organizations with strong knowledge-management programs realize measurable lifts across revenue, productivity and customer satisfaction.
The report found that employees reclaim 3.9 hours per week—nearly 10 percent of team capacity—once information flows freely. Customer-facing teams see a 40 percent drop in resolution time, a 35 percent decrease in escalations and a 28 percent rise in CSAT scores.
These aren’t abstract gains. A national insurance carrier using Bloomfire’s platform saved $121 million annually, the equivalent of 1,867 full-time employees, simply by reducing search time and duplicated work.
The Economics of Enterprise Intelligence
The report frames “enterprise intelligence” as the next stage of corporate infrastructure, the convergence of knowledge management, business intelligence, enterprise search and AI. When those systems operate in sync, data stops being static documentation and becomes an active, learning network.
From an accounting perspective, that shift is overdue. “Knowledge remains one of the enterprise’s most undervalued, undertracked and underutilized assets,” the study noted.
Companies spend millions developing expertise and insights, yet much of it “gets trapped in silos, buried in emails, or lost as employees walk out the door.”
By treating knowledge like any other depreciable asset, maintaining it, auditing it and assigning it measurable value, leaders can begin to quantify its return.
Bloomfire’s research suggests that each 100 employees generate $27.1 million in productivity gains and $2.69 million in cost avoidance when robust knowledge-management practices are in place.
When Knowledge Finds You
Brittan argues that the real breakthrough isn’t just storing knowledge, it’s activating it. “We worked with a client whose customer-support teams faced growing backlogs,” he recalled. “Agents wasted hours searching through disconnected systems … The knowledge existed; it just couldn’t reach the people who needed it when they needed it.”
After deploying Bloomfire’s enterprise-intelligence layer, the dynamic flipped. “Rather than agents hunting through systems, relevant knowledge began finding them through contextual recommendations,” he said.
“When a customer described an issue, the system immediately surfaced potential solutions based on similar past cases. But the real power emerged in the bidirectional flow—when agents discovered new solutions, they contributed that knowledge back … Each interaction improved the system’s effectiveness for everyone.”
That feedback loop, Brittan says, exemplifies how AI should function inside the enterprise: “continuously sensing, learning and adapting,” much like a nervous system. It’s a direct response to the shortcomings highlighted in MIT’s research, where disconnected data left AI agents without the context to make reliable decisions.
As McDonagh-Smith noted, “When data isn’t managed for cross functional use, AI systems are starved of their oxygen. Data is the hydrogen of AI. If at least 80 percent of our organizational data isn’t accessible when needed, how can we breathe AI into our products, services and experiences?”
From Stand-Alone Tools to Intelligent Foundations
The early wave of generative AI adoption followed a predictable arc: rapid experimentation, lofty expectations, uneven results. Brittan sees a maturation underway.
“The initial wave brought a rush of stand-alone AI applications … that grabbed headlines but often operated in a silo from core business systems,” he said. “These tools showed promise but frequently disappointed when deployed against complex organizational problems.”
Now, he believes, enterprises are building something more integrated: “What’s taking shape is essentially a new intelligence layer that sits between raw data and decision-making. It’s not just better search or smarter dashboards; enterprise intelligence orchestrates how knowledge flows throughout an organization.”
The distinction matters. Rather than asking which AI tool to buy, leading companies now ask, ‘How do we build an intelligent foundation that amplifies human judgment?’ That question, Brittan says, “will ultimately separate those who capture AI’s value from those who simply chase its promise.”
Governance, Not Guesswork
AI may generate excitement, but knowledge governance determines success. Brittan’s advice to executives is practical: start with a knowledge audit, establish federated governance (central standards with distributed ownership) and make knowledge capture part of everyday workflows. “Unsuccessful AI projects typically fail because of knowledge foundation issues, not technical problems,” he said. “The technology itself rarely causes failure.”
He warns that the greater risk isn’t AI malfunction but unmanaged information. “While many executives worry about AI implementation risks, the more serious threat comes from allowing fragmented, inconsistent knowledge to spread through their organization unchecked,” he said.
“Address your knowledge foundation first, and the AI implementation becomes dramatically more likely to succeed.”
That view echoes the MIT findings: Organizations aren’t suffering from a shortage of algorithms; they’re suffering from a shortage of coherence.
Or, as McDonagh-Smith put it, “AI is a team sport. It’s both that simple and that complex. AI-forward organizations are adept at operating with dual-speed teams: one experimenting rapidly, the other focused on enterprise-grade integration with disciplined and deliberate compliance rigor.”
The Measured Future of AI
For all the enthusiasm around generative AI, Brittan is careful to ground expectations. “AI is a very powerful tool, but it is important to remember that it is a tool, not magic,” he said. “Using it inappropriately is likely to be disappointing and potentially misleading or dangerous.”
He believes the winners will be those who pair curiosity with discipline. “When implemented properly, AI can be safe and very valuable and will give those companies that have the right expectations, the right mindset, and the right skills a definite edge over the long term.”
That edge, according to Bloomfire’s data, comes not from chasing novelty but from mastering fundamentals: clean data, connected teams and contextual knowledge that flows as freely as information itself.
The future of AI in business may depend less on what machines can imagine and more on how well humans remember what they already know.
