Satya Nadella argued recently that every company needs to build a continuous learning loop around AI. Tom Davenport, a pioneer of Knowledge Management, points to a shift from accumulating knowledge to putting knowledge into the flow of work where it creates value.
They describe two sides of the same transformation. Organizational knowledge plus human/AI collaborative reasoning, creates an opportunity for “compounding knowledge.”
Consider reasoning involved in a trans-pacific flight. A flight plan incorporates reasons for a particular flight path, including the start time, anticipated weather, performance of the jet, load assumptions, and other inputs. The plan is documented, and actual time and fuel consumption are recorded and evaluated for future optimization.
Let’s consider a more typical enterprise need. A quarterly analysis seeks to identify underperforming wholesalers. Knowledge and context are provided, and an AI model generates an analytic plan sourcing data from varied databases, and calculating performance. The sales team reviews and adds judgement that wholesalers impacted by a recent hurricane are exempt. The plan is updated and decisions are made, and the reasoning is saved. The next quarterly review will begin by updating the saved plan.
“Reasoning records” are a new form of compounding knowledge made possible by advanced reasoning AI models. Knowledge Management is the context for collaborative human and AI reasoning, producing reasoning that becomes the foundation for future improved reasoning.
Knowledge → Human + AI Reasoning → Validated Reasoning → New Knowledge → Better Reasoning → Improved Knowledge
That isn’t simply AI-enabled Knowledge Management.
It’s a compounding enterprise learning loop.
Knowledge Management Preserved the Products of Knowledge Work
Organizations are good at preserving knowledge work outputs, such as reports, presentations, policies, procedures, and other lessons learned. But consider what happens before one of those artifacts is created.
An analyst determines which evidence is relevant, and a finance executive challenges the definition of margin. An analyst explains why a historical comparison isn’t valid. Alternatives are considered, and eventually a conclusion is reached.
The final report preserves the answer, but much of the reasoning that produced it disappears. Historically, reasoning has been difficult to capture, as much of it resides in people’s heads, in meeting minutes, in email, and increasingly in chat sessions.
AI Reasoning
When people collaborate with AI, knowledge creation becomes explicit. The problem is described and evidence is evaluated, and assumptions are established. The analysis evolves, human expertise is applied, and conclusions are reached.
A valuable chain of reasoning is created, but is typically discarded once a conclusion is reached. AI makes advanced reasoning more accessible, making employees more capable, while the organization learns surprisingly little from the new intelligence.
Enterprise Reasoning as Compounding Knowledge
This is the opportunity behind Enterprise Reasoning: treating human/AI reasoning as a first-class enterprise artifact.
Instead of preserving only a final answer or report, an organization can retain a structured reasoning record, including the problem, evidence, assumptions, semantic definitions, analytical steps, calculations, alternatives, AI contributions, validation and provenance.
The organization preserves what it knows, and how it arrived at what it knows.
Knowledge Flows and Knowledge Stocks
Davenport and coauthors distinguish between knowledge stocks and knowledge flows. Human/AI collaboration creates a new knowledge flow. Every day, employees use AI to analyze problems, challenge ideas, interpret evidence and develop conclusions.
Enterprise Reasoning creates the possibility of durable knowledge of reasoning: validated, governed reasoning that can be discovered and reused by people, AI models, agents and software.
This suggests an expansion of Knowledge Management.
AI-enabled Knowledge Management asks: how did the organization arrive at what it knows, and is it compounding that knowledge?
This becomes increasingly important as AI participates in more consequential enterprise decisions.
From Knowledge Management to Organizational Learning
The opportunity grows when reasoning is connected to outcomes. An organization preserves the reasoning behind a pricing decision. Six months later, it knows what happened to margins, revenue, and customer retention. It can now evaluate whether the decision was successful, and whether the reasoning behind the decision was sound.
This creates a learning loop:
Problem → Human + AI Reasoning → Validation → Decision → Outcome → Evaluation → Improved Reasoning → Update + Reuse
The reasoning repository becomes infrastructure for organizational learning.
AI Can Make the Enterprise Smarter
Does the use of AI leave the organization more knowledgeable than it was before?
If thousands of employees and agents solve thousands of problems, but the reasoning disappears after each interaction, much of the intellectual capital disappears with it.
If validated reasoning is retained, evaluated against outcomes, improved and reused, the organization accumulates a body of knowledge about how it solves problems.
That may be one of AI’s most important contributions to Knowledge Management.
The next generation of Knowledge Management will help them preserve, improve and reuse how they reason over problems.