14/11/2026

Kibbutz and Community Archives: From Boxes to a Catalogue

Written for kibbutz and moshav archivists, volunteers in local and municipal heritage collections, and the people who keep synagogue and landsmanshaft records.

A small community archive usually looks the same from the inside. Decades of minutes from general meetings. A weekly newsletter, bound by year or loose in folders. Thousands of photographs, many with nothing written on the back. Member files. Film reels and audio cassettes nobody has played in years. Letters from relatives abroad in Yiddish, German, Russian or Polish, next to Hebrew typescript and handwritten notes.

And usually one part-time archivist, a few volunteers, and very little budget.

This post is a practical path from that situation to a catalogue people can search. It does not assume a grant, a new system, or a large team.

Description, not scanning, is the bottleneck

Most community archives already scan, or could. A flatbed scanner and a careful volunteer can image a box of newsletters in a few sessions. The hard part comes afterwards: a folder of image files that nobody can search, because nobody has written down what each one is.

A scan without a description is a photograph of a document, not a catalogue record. To find the 1958 meeting where the dining hall was debated, or every photograph of the first harvest, someone needs a title, a date, places and names attached to each item. That writing is the work that never gets finished, and it is where AI now helps most.

Start with a survey, not with "everything"

Before any scanning or AI, spend a few weeks on a survey. It does not need to be detailed. For each group of material, record:

  • what it is (minutes, newsletter, photographs, member files, film, audio, correspondence),
  • roughly how much (shelf metres, boxes, or an estimate of items),
  • the years covered and the languages,
  • condition and whether any of it is already scanned,
  • how sensitive it is, and who is allowed to see it today.

Then choose one well-bounded series to finish first. Good first series have a clear beginning and end, a uniform format, and real demand from the community. The weekly newsletter is often ideal: printed or typed, dated, and full of names, events and decisions. The photograph collection is the other strong candidate, because it is what members ask about most, and because the people who can identify the faces are getting older.

The backlog sprint post explains the logic in more detail: one series done completely is worth more than every series advanced a little.

A minimal description that will still import later

Small archives often describe material in a spreadsheet, a Word file or a card index. That is fine, as long as the fields are chosen with a future catalogue in mind. A minimal set that maps cleanly to common standards:

  1. Identifier. Collection, box, file and item number, matching the scan file name.
  2. Title. A short factual description: "General meeting: budget and new members".
  3. Date. As written, plus a normalized form (1958-03-14, or a range when uncertain).
  4. People. Names as they appear, plus the form the community uses today.
  5. Places. As they appear in the source.
  6. Type and language.
  7. Access. Open, members only, or restricted.
  8. Notes on identification. Who identified a person or place, and how certain.

These fields map to Dublin Core and can later feed a hierarchical finding aid. The metadata standards post shows the mapping and how to record uncertain dates. Names deserve special care in community archives, because many members Hebraized their surnames and appear under different names in different decades. The name variants post covers why one person ends up with many spellings.

Splitting the work: AI drafts, the community confirms

The scarce resource in a community archive is not typing time. It is local knowledge: the long-time member who knows that the man on the tractor is the founding treasurer, that "the hill" in the minutes is a specific field, and that a nickname in the newsletter belonged to a particular family.

Do not spend that knowledge on data entry. A sensible division of labour:

  • AI reads and drafts. It runs recognition on typed and handwritten pages, including Hebrew, Yiddish and Russian, and drafts a title, a date, places and names for each item. Each identification carries a mark: confirmed by the source, probable, or tentative. When the source says nothing, the field stays empty instead of being guessed.
  • The archivist checks the structure. Order, identifiers, access level, and whether the draft matches the item.
  • Volunteers and long-time members confirm identifications. They look at tentative names and places, correct them, and add what only they know. The record notes who identified whom.

For photographs, this works well as a regular session: a screen, a group of veteran members, the AI draft for each image, and someone recording the answers. Grouping similar photographs first, from the same event or the same roll, makes these sessions far more productive. The photo cataloguing guide explains where AI should stop, in particular why naming faces remains a human decision.

Member files and living people need restricted handling

Not everything in a community archive belongs in a searchable catalogue. Member files can hold applications for membership, health matters, family circumstances, financial requests, and committee decisions about individuals. Minutes of general meetings sometimes record personal disputes. Many of the people involved are alive, and where they are not, their children often are.

Some practical rules:

  • Leave member files out of the first series. Start with material the community already treats as shared.
  • Decide where each category may be processed before sending anything to an outside system, and ask what is kept and who can see it.
  • Separate description from access. A restricted file can still have an internal record; it does not have to appear in public search.
  • Flag sensitive content for review, rather than relying on automatic redaction.

The sensitive archives post sets out the questions to ask and how to tier a collection by sensitivity.

The memory of veteran members

Documents record decisions; they rarely record why. Veteran members carry the context: what the argument in a meeting was really about, who took a photograph, what an event meant at the time. That memory is part of the archive, and it does not wait.

Photo identification sessions are a natural starting point for recorded interviews. A photograph prompts specific memories, and the interview can then be linked to the items it explains. The oral history guide covers preparation, recording and transcription.

A short checklist

  • Survey everything briefly before starting anything.
  • Choose one bounded series: the newsletter or the photographs.
  • Fix a minimal field set that maps to a standard.
  • AI drafts title, date, places and names, with confidence marks; empty when the source is silent.
  • Volunteers confirm identifications; record who identified what.
  • Member files and personal matters stay restricted.
  • Link interviews with veteran members to the items they explain.

If your archive has a series ready to start with, our archival cataloguing service drafts the title, dating, places and names for documents, photographs and films, marks each identification as confirmed, probable or tentative, and leaves every record for a person on your staff to approve, before export to Dublin Core, EAD, MARC21 or the spreadsheet you already use. Organizations can start with a free pilot of 100 documents, 10 photographs and 5 films. Get in touch with a description of the series you have in mind.

Frequently Asked Questions

Where should a small community archive start digitizing?

With a short survey of everything it holds, followed by one well-bounded series, such as the weekly newsletter or the photograph collection. Finishing one series completely, described and searchable, teaches more than advancing everything a little.

What is the best use of volunteers in a kibbutz or community archive?

Identification, not typing. Long-time members can name the people, places and occasions in a photograph or a set of minutes, which no model and no outside archivist can do. AI drafts the routine fields so their time goes to confirming and correcting identifications.

Can member files be processed with AI?

Only with restricted handling. Member files often contain health, family and financial matters about people who are alive or whose children are. Leave them out of the first series, decide where they may be processed, and keep their descriptions out of public search.

Does AI-drafted description import into an existing catalogue?

It can, if the fields are decided first. Keep a small set of fields that map to Dublin Core or the spreadsheet the archive already uses, store dates in a normalized form, and record who confirmed each identification.