When I work with organizations on AI initiatives, I often hear: "We've implemented ChatGPT" or "We're using Claude" or "We're rolling out Copilot." Then comes the frustration: "But it's giving us generic outputs that don't really fit our business."
The problem isn't the AI. It's that the AI doesn't know what you know.
Before you can create meaningful business value with AI - whether that's boosting productivity, creating new customer value, or driving disruption - you need to solve a more fundamental challenge: capturing and encoding the knowledge that makes your organization unique. It's a prerequisite for any AI business value.
Research shows that tacit knowledge - the unwritten, experience-based expertise that lives in people's heads - accounts for over 80% of an organization's intellectual capital. This is the knowledge that matters most:
How your best salespeople qualify prospects and tailor their approach
Why certain customer implementations succeed while others struggle
Which shortcuts experienced operators take and which rules they never break
The context that turns raw data into meaningful insights
When customers actually need versus what they say they need
The AI tools you're using - no matter how sophisticated - don't have access to this knowledge. They have access to public internet data, their training sets, and whatever you happen to type into a prompt.
So when you ask Claude to draft a customer proposal, it writes something generic. When you ask ChatGPT to analyze your sales pipeline, it misses the nuances your experienced reps would catch immediately. When you deploy an AI chatbot for customer support, it gives technically correct answers that miss your customers' real concerns.
The AI isn't failing. It's working with incomplete knowledge.
Why knowledge capture is foundational, not optional
In the first issue of this newsletter, I discussed the three buckets of AI business value: Boosting Productivity, Creating New Value, and Driving Disruption.
Knowledge capture isn't a fourth bucket. It's the foundation that determines whether you can actually achieve value in any of the three.
For Boosting Productivity: If you want AI to help your team work faster, it needs to know how your team actually works. Not generic "best practices," but your specific workflows, your terminology, your quality standards, your decision criteria. Without capturing this knowledge, AI becomes a generic assistant that speeds up the wrong things or produces outputs requiring extensive revision.
For Creating New Value: If you want to enhance customer experiences with AI, the AI needs deep knowledge about your customers, your products, and your value proposition. A customer service AI needs to understand not just your product specs, but why customers buy from you, what frustrates them, what delights them. Without this captured knowledge, you're deploying generic AI that treats your unique customers like everyone else.
For Driving Disruption: If you want to reimagine your business model or operations, you first need to capture how things currently work and why. You can't disrupt what you don't understand. The most valuable insights about potential disruption often come from understanding the hidden knowledge - the workarounds, the pain points, the opportunities your experienced people see but haven't articulated.
The difference between "we're using AI" and "we're creating value with AI" is whether the AI has access to your organization's actual knowledge.
This week, identify one area where valuable knowledge is concentrated in a few people's heads. Pick something that matters to your business and where losing that expertise would hurt.
Schedule a one-hour conversation with one of those experts. Don't ask them to write documentation. Instead, have them walk through a real recent example of applying that expertise. Record it (with their permission). Have AI generate a transcript.
Look for patterns (again, you can use AI to ask these questions):
What questions did they ask that others wouldn't?
What did they notice that seemed unremarkable to them but wouldn't be obvious to others?
What decision criteria did they use?
What shortcuts or rules of thumb emerged?
That's the start of knowledge capture. You don't need a massive initiative. You need to start surfacing tacit knowledge and encoding it in ways that create value.
Because the organizations winning with AI aren't just using better models. They're teaching AI what they know - and making that knowledge available at the moment of need.