Where does company knowledge hide?

In ten common places: people’s heads, conversations that were never recorded, recorded meetings and transcripts, your top performers’ processes, email, instant channels like Slack, Teams, and WhatsApp, completed client work, composite documents, CRM and project management tools, and shared drives. The biggest share of valuable knowledge sits in people’s heads. The smallest sits on shared drives, even though drives hold the most files.

What percentage of company data is unstructured?

According to IDC and Gartner, 80-90% of a company’s data is unstructured. It lives in email threads, meeting notes, call transcripts, chat messages, presentations, and documents. IDC’s Data Age 2025 report also found unstructured data growing at roughly 61% per year.

What is Scattered Knowledge?

Scattered Knowledge is the knowledge, experience, expertise, judgment, standards, and definitions a company runs on, spread across people, systems, documents, formats, and locations. That spread makes it hard or impossible for a new employee or an AI to find and use.

How much of what employees know is unique to them?

According to a 2018 Panopto study of 1,001 US workers, 42% of the knowledge an employee needs to do their job is unique to that person. Nobody else in the company holds it, so when that person is unavailable, colleagues can’t do that part of the work.

What’s the difference between implicit and tacit knowledge?

Implicit knowledge can be put into words when someone asks the right question, like your real value proposition or your view of the market. Tacit knowledge can’t be put into words even when asked, like a founder’s instinct about fit. Implicit knowledge comes out through interviews and decision-forcing sessions. Tacit knowledge only comes out through observation.

How do you capture knowledge before a key employee leaves?

Start with structured interviews and decision-forcing sessions for what they can explain, then observe what they can’t: call recordings reviewed for patterns, shadowing, and debriefs right after significant events. Reflect what you see back until they confirm or correct it. According to Deloitte and eGain (2026), 92% of organizations fail to consistently capture knowledge from people about to retire.

Do AI notetakers solve the problem of lost knowledge?

They solve capture, not extraction. According to Kolmogorov Law and Pollfish (July 2026), one in three US workers has had an AI notetaker in a work meeting. But a recording only becomes knowledge once you decide what to surface from it: decisions, objections, customer questions, and the patterns across many calls.

Can AI just read our shared drive?

It can read it, but it can’t tell which document is true. When a drive holds four versions of the onboarding process, AI treats all four as equally valid. A 2025 systematic review in ScienceDirect found that knowledge management systems can reinforce outdated practice because obsolete content keeps surfacing as if it were authoritative.

Where should a company start?

Start with what’s already captured: meeting transcripts, composite documents, CRM exports, and the shared drive. Then interview the people whose knowledge isn’t written down anywhere. Then decide, because the most important knowledge, like who you won’t serve and what you won’t claim, can’t be extracted. It can only be decided.