The Grammar is Wrong, Which Means the Thinking is Wrong
Every knowledge capture initiative I've seen has the same flaw. The grammar is wrong, which means the thinking is wrong. Knowledge capture is not a noun but a gerund.
This is not pedantry. A gerund is a verb form ending in -ing that functions as a noun while retaining its verb properties. The -ing suffix signals ongoing action rather than completed state. Reconciling is not the same as reconciliation. One is an action. The other implies an endpoint.
Organizations treat knowledge capture like a project with a finish line. Someone schedules a kickoff meeting. A project manager creates a timeline with milestones. The goal is always articulated as completion: "Capture the tribal knowledge by Q2." "Document the processes by year-end." "Build the wiki before the VP retires."
Six months later, the documentation is stale. The experts have moved on or evolved their thinking. The processes have changed. The wiki is a fossil.
This is not a failure of execution but a failure of framing. The wiki was built as a noun, a finished artifact, but knowledge doesn't hold still.
Accounting figured this out before tech did. Twenty years ago, reconciliation was a monthly event. You closed the books, you reconciled accounts, you were done until next month. Compare two ledgers at month-end and fix discrepancies. If they matched, you were done. If they didn't, you found the error and corrected it.
This worked when transactions moved slowly enough that monthly snapshots captured reality. But as transaction volumes increased and business moved faster, monthly reconciliation became an archaeological dig. By the time you closed the books, the discrepancies you were fixing were already obsolete.
Continuous reconciliation emerged not from better technology but from a conceptual shift. The question changed from "Do these match?" to "How quickly can we detect and correct drift?"
Finance teams stopped treating reconciliation as an event and started treating it as a process. Real-time matching. Daily validation. Anomaly detection running in the background. The month-end close didn't disappear, but it became confirmation rather than discovery.
In 2026, leading finance teams reconcile high-volume accounts daily or weekly, not monthly. Cash accounts, accounts receivable, accounts payable get continuous matching to prevent month-end bottlenecks and catch errors while they're still fixable. The shift wasn't about working harder. It was about acknowledging that reconciliation is not an event but a state you maintain. You don't reconcile once. You keep reconciling.
The accounting profession moved from reconciliation (noun) to reconciling (gerund) because the alternative was always being wrong about something that happened three weeks ago.
Knowledge work hasn't made the shift. Most organizations still treat knowledge capture like monthly close accounting. Launch a documentation project. Interview the experts. Capture the tribal knowledge. Lock it in a wiki or a SharePoint site or a Confluence page. Mark the project complete.
The problem is not that the documentation was done poorly. The problem is that it was treated as something that could be "done" at all.
Knowledge is not static. Expertise evolves. Processes drift. Context shifts. Treating knowledge capture as a project with a completion date is like reconciling your accounts once and assuming they'll stay balanced forever.
What would continuous reconciliation look like for knowledge? In accounting, it means matching transactions in real-time, detecting anomalies as they occur, and updating records incrementally rather than in batch.
Applied to knowledge, this means not interviewing the expert once, writing it down, and locking it. Instead, create a process where expertise is continuously elicited, refined, and validated against reality. Don't document the current state and assume it's durable. Track how understanding changes over time and surface where drift is happening. Don't aim for complete capture. Aim for improving approximation, knowing the answer will shift before you finish.
There's a calculus concept that clarifies this. Some functions cannot be integrated analytically. The exact answer doesn't exist in closed form. Functions like e^(-x²) or sin(x³) lack elementary antiderivatives. So you approximate using numerical methods: the trapezoidal rule, Simpson's rule, Monte Carlo integration.
Each method divides the area under the curve into shapes you can calculate. Rectangles. Trapezoids. Parabolic segments. You add them up. The answer is never exact, but with each refinement, you get closer.
The error shrinks but never reaches zero. Not because the methods are bad, but because that's the nature of approximation. You're reconstructing something continuous from discrete samples. More samples, better approximation. But "better" is not the same as "true."
This is not a bug in the math. This is the structure of knowing when perfect information is unavailable.
Imagine you're trying to calculate the area under a curve but you can't solve the integral analytically. The trapezoidal rule says: divide the domain into intervals, approximate each segment as a trapezoid, sum the areas.
The approximation improves as you add more trapezoids. Double the number of intervals and the error drops by a factor of four. But even with a million intervals, you're still approximating.
This is knowledge capture. Each interview is an interval. Each conversation adds a trapezoid. You get closer to the true shape of expertise, but you never capture it exactly because the curve itself is changing while you're measuring it.
The numerical integration metaphor holds. Each conversation is another sample point. Each refinement adds precision. The answer gets better but never final.
When you say "We are reconciling the knowledge base," you acknowledge that the work is continuous. When you say "We completed the reconciliation," you're claiming an endpoint that doesn't exist.
The grammar encodes the philosophy. Treat knowledge work as a gerund and you design for continuous approximation. Treat it as a noun and you design for a finish line that keeps moving.
Accounting made the shift from periodic to continuous reconciliation because the cost of being wrong grew faster than the cost of staying current. Knowledge work is hitting the same inflection point.
The organizations that figure this out first will stop planning knowledge capture projects and start building reconciliation processes. Not wikis that ossify. Not documentation that goes stale. But systems designed for continuous approximation, where each pass refines the model and the expectation is refinement, not completion.
The answer is a gerund, not a noun. Reconciling, not reconciliation. The process never ends, and that's not a bug. That's the architecture.
Next time: Why the most valuable people in your organization can't describe what they do, and why that's not a communication problem.