KleioBase vs Hiring a Professional: When to Pay a Person

There is a version of this page that tells you software has made professional transcribers unnecessary. That version would be wrong, and if you acted on it in the wrong circumstance it would cost you a citizenship application or a court filing.

So this page is about the decision rule instead. Not which is cheaper, which the cost breakdown article already covers in detail with current figures, but which documents genuinely require a person.

The short answer. Pay a human when the output needs to be attested, when the stakes of a wrong reading are irreversible, when the document has defeated general models, or when what you actually need is judgement and archive access rather than reading. Use software for the other ninety per cent, which is bulk, routine, and mostly boring. The two are sequential, not competing: triage with software, then spend your professional budget on the handful of documents that earn it.

Certification is a legal artefact, not a quality level

This is the distinction people most often miss, so it goes first.

A certified or sworn translation is not merely a better translation. It is a document carrying a translator's signed attestation that the translation is complete and accurate, and in some jurisdictions the translator must be registered with a court or ministry. The institution receiving it is relying on that signature and on the professional liability behind it.

No AI output has this. Not ours, not anyone's. There is no signature and no liable professional, so the attestation does not exist and the filing will be rejected.

You need certification for, at minimum: ancestry-based citizenship and nationality applications, immigration filings, probate and estate proceedings, court submissions, and most official registry corrections. If your document is destined for a government office, stop reading comparison pages and find a certified translator for the language pair.

Software is still useful before that point. Knowing what a document says tells you whether it is the document worth paying to certify, which is a much cheaper mistake to avoid than it sounds.

Some hands defeat general models

Being straightforward about capability rather than marketing it.

General-purpose models, ours included, are strong on ordinary period handwriting and weak in predictable places: heavy abbreviation in early Latin, badly faded or water-damaged ink, idiosyncratic personal hands with invented contractions, palimpsests and overwriting, and marginalia crammed in at angles by someone who assumed nobody would ever need to read it.

A trained palaeographer reads these by combining letterforms with knowledge of the formula the document is following, what a scribe in that place and decade would have written, and what the phrase has to be given the surrounding text. That is domain reasoning applied to an ambiguous image, and it remains genuinely better than pattern recognition on hard cases.

The practical test: if you have looked at the output, compared it to the image, and it is still incoherent in the specific place that matters, that document has earned a human. If the output is coherent and only a name spelling is uncertain, it has not.

Reading a document is an input, not the product

Worth stating plainly, because the shape of this comparison hides it: KleioBase is not a transcription tool. Transcription is one of two ways to get data in. The other is a GEDCOM import, and a researcher who imports a tree and never uploads a single scan still gets the whole product - duplicate detection across the tree, conflicting dates surfaced as conflicts rather than silently overwritten, missing generations flagged as gaps, a tree health score, and a companion that can be questioned about the whole archive in plain language.

That is the product. The reading step exists so that documents can join the same analysis, not because text is the deliverable.

What you are often actually buying is judgement

Much of what a professional genealogist provides is not reading at all, and framing it as a transcription purchase undervalues it badly.

A professional knows which record sets exist for a given parish in a given decade, which were destroyed and when, which are held locally and have never been filmed, and which archive requires an in-person visit with a written request in the national language. They know that an absence of records in one series is meaningless and an absence in another is strong negative evidence. They can evaluate conflicting evidence and write a conclusion that meets professional standards.

None of that is a reading problem. No software replaces it, and ours does not try to.

Side by side

When hiring a person is the better answer

The output must be accepted by an institution. Certification, every time, no exceptions and no amount of accuracy substitutes.

The document is pivotal and contested. If one reading decides which of two families is yours, and the ink is ambiguous, pay someone whose judgement you can cite and who carries liability for it.

The record is physically difficult. Severe damage, palimpsest, dense abbreviation in an early hand. Software degrades on these in ways that are not always obvious from the output.

You need someone in the archive. Records that were never filmed require a person in the building. This is common for eastern European parish records and for many regional archives worldwide.

You need a professional report. A written conclusion meeting professional genealogical standards is a deliverable with a methodology behind it, not a transcript.

What KleioBase does instead, and what it does not

We are a tool in this space, so read this as disclosure.

KleioBase is not a cheap transcriber and does not want to be read as one. It is an analysis layer: it reconciles people across every source you hold, raises contradictions between them, and names the gaps, whether your data arrived as a GEDCOM or as scans. Reading and translating a document is how a document joins that analysis. Across a large pile that is hours of work per document that you do not do, and it makes triage possible: you can see what all 200 documents say before deciding which six deserve a professional's time and your money. For professionals who want to work this way, workspaces keep each client's research separate.

Now the honest limits, which on this page are the whole point.

No certification, ever. We produce no attestation and carry no professional liability. For any official filing, this is disqualifying by design, not by current capability.

Hard documents are hard. Faded, damaged or heavily abbreviated hands produce degraded readings. Every result is shown to you for review before it enters your archive precisely because this is true.

No archive access and no local knowledge. We do not know what exists in a parish we have never seen. We only know what you add.

We are not a research conclusion. The archive is evidence, organised. Arguing from it to a defensible conclusion is still your work, or a professional's.

Where to start with the budget you have

Sort your documents into three piles.

The pile destined for a government office goes to a certified translator now, and nothing on this page applies to it.

The pile you have already tried to read and still cannot, where the illegible part is the part that matters, goes to a palaeographer or a specialist translator. That pile is usually much smaller than people expect, often two or three documents out of a hundred.

Everything else is the bulk: legible enough, routine in form, tedious in volume, and valuable mainly because of what it connects to. That is what software is for, and processing it first is what tells you how small the other two piles really are.