Not every gap is a course. When someone is mid-task, they don’t want a thirty-minute module — they want the right answer, now. I design performance support around how people actually search, so the answer they need is findable in plain language at the moment of work.
Every operation has the same failure: the information is technically documented, but it’s buried in a folder tree only its author understands, written in the language of the expert instead of the searcher. So the same questions get re-asked, work stalls, and when a knowledgeable person leaves, their knowledge walks out with them. Reaching for a course here is the wrong tool — the person doesn’t have a skill gap, they have a findability gap.
Mosher & Gottfredson’s framework. Training serves the first two moments — learning something new or more. The last three happen at the point of work, where a course is too slow and performance support wins. Knowing which moment you’re designing for decides the whole intervention.
Learning something for the first time.
TrainingDeepening what you already know.
TrainingDoing it while working — needing a nudge or reference.
Performance supportWhen the process or rules change under you.
Performance supportWhen something breaks and you need the fix now.
Performance supportThe design work is turning messy source material into something a non-expert can actually retrieve. The transformation:
A raw paste — an email thread, a Slack exchange, a subject-matter expert’s scattered notes on the new returns exception.
A title written as the question a searcher would type; plain-language body; a taxonomy tag set built on how non-experts think, not the org chart; synonyms so the wrong words still find it.
One governed, current article, returned by natural-language search in seconds — not a folder hunt, and not five stale copies.
Two design decisions carry the load: an information architecture organized around the searcher’s intent, and governance — one owner, one current answer, so trust in the source never erodes.
AI takes the raw paste and drafts the structure — it categorizes, tags, and shapes an article, then serves it back through natural-language search. That collapses the capture work from hours to minutes. What stays mine is the design judgment: the taxonomy itself, the question-shaped titles, and the governance that keeps the answer current. The AI structures; I decide how it’s found.
Designing for the moment of need means fewer repeat questions, faster work, and tribal knowledge captured before it leaves. It also sharpens the training that remains: once performance support absorbs the “apply, change, solve” moments, courses can focus on what genuinely needs to be learned, not memorized for reference.
Frameworks in play: the 5 Moments of Need, performance support, and information architecture / taxonomy design. When the gap really is a skill and a course is warranted: From a sentence to a full course →