What if you could just keep writing, but custom to the wider topic or context, you get help in specifying things exactly, thereby adding to a body of knowledge with reasoning power?
One Example: Worldwide there are thousands of makerspaces, hackerspaces, tech shops and fablabs where a wide range of people go to MAKE things. Often at least partly by digital fabrication means: laser cutters, 3D-printers, etc. Sometimes writing a little report about it that others can learn from. These reports are however often scattered, hard to find and because of that not very useful, which is not very motivating to put in the time to make a proper one in the first place.
Often specific form fields are used to improve the situation a bit, but more can be done: The editor could provide photos of local machines or popular materials as buttons to 'stamp out' a standardized naming. Also while writing the editor could offer "Did you mean ...?" suggestions (possibly by DBpedia Spotlight) to do the same. The standardized naming also has an 'invisible code' to it, which allows machines to reason about it and provide smart search: One could for example ask for all projects in which a 3D-printer was used and a material with a melting point below 200 Celsius even when the material property was never entered into such a report (the material was named in such a way that the search engine could find it on Wikipedia, wikidata or other bodies of knowledge). Many examples of these bodies and the relationships in and among them are shown in the Linked Open Data cloud. Search engines like Google look for these 'invisible codes' (RDFa) and 'understand' them when they use the schema.org vocabulary. (sidenote: for this specific example use case, years ago http://fablab.metamatter.nl/ was designed.)
To relate content to SDGs: https://aurora-network.global/project/sdg-analysis-bibliometrics-relevance/ Or to COVID-19 which was mentioned a good idea or whatever societal very relevant, also see wikidata and how it works: https://github.com/egonw/fvtworkshop/blob/master/README.md
Maybe a nice combination with: Deep Spreadsheets and TrainOfThought
Do it in a ((R)markdown) notebook context?
As a Word/GoogleDocs/Authorea plug-in?
Would PIDINST (Persistent IDentifier for INSTruments) have use in this?