Ninety-four percent of the information stored in a corporate wiki is destined to be read exactly once, usually by the person who wrote it, before it becomes a digital fossil. This is the quiet tragedy of the modern enterprise: the belief that if we simply archive enough data, we can eventually replace the people who understand it.

We treat knowledge like a pile of dry wood, assuming that as long as we have the lumber, we have the fire, forgetting that the fire is a living process that requires the oxygen of human context to stay lit.

The Soul of the Strawberry

This is a distinction I’ve learned to appreciate in my own line of work as an ice cream flavor developer. When you are trying to recreate the specific emotional resonance of a “Grandma’s Kitchen” vanilla, you can look up the chemical components of vanillin and the exact fat content of high-grade cream, but the database won’t tell you that the secret is actually a microscopic hint of salt and the smell of a pilot light.

This disconnect between the data and the soul of the thing is which is also how a strawberry flavor can contain forty chemicals but zero fruit. In the lab, we can document the molecular structure of a flavor until the servers groan, but if the developer who knows the “lean” of the batch retires, the recipe becomes a dead language.

When I accidentally laughed at my cousin’s funeral last year, it wasn’t because the service was funny; it was because the priest tripped over a flower arrangement and my brain, overwhelmed by the somber rigidity of the room, chose the only exit strategy it had left.

It was a human error, a failure of decorum that revealed a deeper truth about the pressure of the moment. Knowledge bases are incapable of this kind of “inappropriate” but deeply human response. They are decorum personified. They provide the “correct” answer regardless of whether that answer is going to sink the ship or save the crew.

The Oracle’s Wall of Noise

Consider Miller, a systems administrator who found himself staring at the “Oracle,” the company’s brand-new, six-figure licensing knowledge base. Miller was tasked with provisioning a new remote environment for a hybrid team that was scaling from 40 to 200 users over a holiday weekend.

Miller needed to know whether to pull the trigger on User CALs or Device CALs for their specific Windows Server 2022 deployment, especially since some users were sharing ruggedized tablets in the warehouse while others were strictly remote-deskbound.

Although the Oracle was searchable and sleek, it presented Miller with three different articles that all technically applied to his situation. Each article was a perfect piece of data, yet together they formed a wall of noise.

Miller realized then that the “Old Man Greg” he had been told to replace-the guy who had been the licensing lead for twelve years-was the only one who knew that for this specific warehouse contract, the tablets were actually being replaced by personal handhelds in three months.

Greg would have ignored the articles and told Miller to buy the User CALs because the long-term flexibility outweighed the short-term cost-per-device. Organizations keep trading living expertise for dead reference because the reference is easier to put on a balance sheet. You can’t “own” Greg’s intuition, so you try to strip-mine it into a wiki, only to find that the gold turns to lead the moment it’s removed from the vein of his daily experience.

The “Frank” Percentage

In the world of flavor chemistry, we have a saying: eighty-seven percent of the code that runs the world is just a digital version of “because Frank said so,” which means that when you fire Frank, you’re actually deleting the operating system of the office. We pretend we are building automated, resilient systems, but we are mostly just building faster ways to find the wrong answer.

The Tiebreaker Principle

When Miller reached the point of genuine frustration, he didn’t need another PDF; he needed a tiebreaker. He needed the judgment that sits in the gap between “what is written” and “what is happening.” This is the primary failure of the digital transition: we have optimized for the retrieval of facts while discarding the wisdom required to apply them.

Because the database was designed to be exhaustive, it became exhausting, a map the size of the territory it was supposed to simplify. This expansion of data mirrors the way ice cream flavors are developed, where more ingredients often mask the lack of a coherent base.

If you have to explain a flavor with a paragraph of text, you’ve probably missed the mark. The same applies to licensing or system architecture. If you have to read seventeen wiki entries to decide how to license a remote desktop environment, the system isn’t helping you; it’s hiding the answer behind a curtain of its own making.

The Judgment Tax

The “judgment tax” is what companies pay when they realize their knowledge base can answer the easy questions but none of the ones that actually matter. The easy questions-“What is a CAL?” or “How do I install the license?”-are commodities. You can find those answers in fifteen seconds on any forum.

The hard questions-“How do I structure this so I’m not paying for 500 licenses I don’t need in two years?” or “Why is the 2019 license pack showing as incompatible with my 2025 CAL pack?”-require a human who has felt the heat of the server room.

We are currently in a cycle where “efficiency” is measured by how few people we need to talk to. We want the “Buy Now” button to do all the heavy lifting. But in my lab, if I stop talking to the old-timers who remember the “bad batches” of ’98, I start making the same mistakes they did, just with more expensive equipment.

Every IT department has its own “bad batch.” Every MSP has a story about the time they miscalculated the licensing for a law firm and ended up eating the cost of twenty User CALs. The knowledge base records the loss, but it doesn’t record the lesson.

Judgment is Non-Transferable

The lesson is that judgment is a non-transferable asset. You can record a person’s voice, but you can’t record their ears. You can document a person’s decision, but you can’t document the invisible “no” that preceded the visible “yes.”

“Here is the one rule you can ignore, and here is the one you can’t.”

– The Mentor’s Tiebreaker

When Miller finally got the help he needed, it wasn’t from the Oracle. It was from a person who looked at his messy, contradictory environment and said those exact words.

Because we live in an era of infinite information, the most valuable thing in the room isn’t the person with the most facts; it’s the person with the best filter. We have enough data. We have enough wikis. What we are starving for is the mentor who can stand over the shoulder of the admin and point to the screen, not to show them where to click, but to tell them why they are clicking there in the first place.

When you finally stop searching the archive and start looking for the answerer, the complexity of things like RDS licensing begins to dissolve. You realize that the “unwritten” parts of the manual were the only parts that were keeping the lights on.

We can keep building our shiny new knowledge bases, and we can keep retiring the people who know what isn’t written down, but eventually, we will find ourselves in a library where all the books are in a language we’ve forgotten how to speak.

Until then, we’d better hope there’s still someone left who remembers where the “Grandma’s Kitchen” salt is hidden.