Writing

Most companies won't fail at AI because of the technology — they'll fail because they don't invest in their people. This piece makes the case that AI tools are only as effective as the human behaviors feeding them, and argues that Knowledge-Centered Service gives support organizations a concrete, proven framework for building exactly those habits before layering in AI. If workers aren't documenting consistently, categorizing accurately, and capturing knowledge in real time, no AI implementation can compensate for that gap.

KCS linking — attaching the correct knowledge article to a support incident — is a high-judgment task that directly affects resolution quality and knowledge health. This piece introduces the Link Validator model I built as a proof of concept: an ML system that predicts link accuracy using sentence embeddings and cosine similarity. It covers how the model works, where it can help, and — importantly — how to implement it without creating over-reliance or undermining analyst judgment.