The Knowledge Transfer Problem AI Can't Solve for You
Companies are investing heavily in AI, but many are failing to see a return on their investment. The problem lies not with the technology, but with the knowledge that companies have failed to capture and transfer effectively. Experts argue that the key to successful AI adoption is to extract and structure the knowledge of the best performers in an organization, rather than relying on generic instructions or assumptions.
The knowledge transfer problem is a long-standing issue that has been exacerbated by the adoption of AI technology. Companies are investing heavily in AI, but many are failing to see a return on their investment. The problem lies not with the technology, but with the knowledge that companies have failed to capture and transfer effectively.
## The Adoption Problem Isn't Where You Think
Companies have poured enormous sums into AI, and most have yet to see a return. According to Gartner, 88% of HR leaders saw no meaningful business value from AI tools in 2025. A BCG study of 1250 companies found that only 5% are capturing significant value from AI at enterprise scale, while 60% are seeing no material return at all.
| Company | % Capturing Significant Value | % Seeing No Material Return | | --- | --- | --- | | Gartner | 88% no meaningful business value | - | | BCG | 5% capturing significant value | 60% no material return |
The default reaction is to blame the technology: wrong model, weak governance, disconnected tools. However, this is not the root of the problem. Even perfectly integrated systems will only pass along the knowledge a company has managed to capture.
## We've Seen This Before AI
This is an old and expensive problem. APQC data shows that 85% of senior executives are concerned about knowledge loss when experienced employees leave, yet only 8% of organizations consistently preserve that knowledge.
| Concern | % of Senior Executives | % of Organizations Consistently Preserving Knowledge | | --- | --- | --- | | Knowledge loss when experienced employees leave | 85% | 8% |
The time lost searching for knowledge didn't start with AI either. According to APQC findings, professionals spend roughly 8 hours a week hunting for the information they need or re-explaining things they've already explained before.
| Time Spent | Hours per Week | | --- | --- | | Hunting for information | 8 |
Onboarding, which depends directly on effective knowledge transfer, tells the same story. Gallup reports that only 12% of employees strongly agree their company does a good job onboarding new hires, and reaching peak productivity typically takes a new person about a year.
## Extract Before You Automate
Many companies already have everything they need for successful AI adoption-except knowledge of how their best people actually work. To overcome this, companies need to extract and structure the knowledge of their best performers.
### Find the Knowledge Holder
This isn't necessarily the veteran with decades of experience. It could just as easily be someone who figured out a working technique with a new tool in their first month. Knowledge doesn't have to be old to be valuable.
### Make the Knowledge Explicit
This requires a series of in-depth interviews with the expert. Ideally, the interviewer is someone trained in asking the right questions: an Instructional Designer, a knowledge engineer, a learning specialist. Their job is to capture how a strong performer makes decisions. The logic behind those decisions then becomes material for the AI tool-the rules, examples, and context the model receives.
### Don't Ask AI to Replace the Expert-Ask it to Help Interview Them
At the extraction stage, AI is genuinely powerful: transcribing conversations, clustering similar cases, turning explanations into draft scenarios, flagging gaps in the expert's logic. What matters is where you point it. An analysis of Nonaka's model as applied to AI makes an apt observation: AI delivers the least value when it reorganizes what's already known, and the most when it helps surface what has never been articulated.
### Test for Real Uptake-in People and in the Model
You can write instructions and sit someone down to watch training videos, but that doesn't mean they know how to apply what they've learned. Adults convert knowledge from passive to active only through practice and feedback. The same goes for the tool: give it real cases from the expert's work and check whether its output matches what the specialist would have done. When the answers diverge, it usually doesn't mean "the model is weak"-it means a rule, an exception, or an example is missing.
## Ask This Before You Buy The Next Tool
If your last AI tool didn't meet expectations, don't rush to test the next one. Ask yourself a few questions instead:
- Who on your team is already getting disproportionately good results from AI-and have you documented exactly how they do it? - What are you teaching the AI model-how work is described in the handbook, or how it's actually done? - If your best specialist leaves tomorrow-what will walk out the door with them, filed under "they just know"?
There's a good chance you won't like the answers. But if you don't ask yourself these questions, the invoice for the AI tool that never paid for itself will ask them for you.
