Do you have to digitise the entire paper archive?

No, and you should not. It makes sense to scan the active and frequently used records first – such as current contracts and ongoing project files. Old records that nobody needs any more only cost scanning effort without any benefit and can be left where they are.

How long does it take to digitise a paper archive?

That depends on the volume and the route you choose. A prioritised core collection can often be done on the side within a few weeks using an office device; a scanning service handles large volumes of old records considerably faster. What matters is not to wait for complete digitisation but to start with the core collection.

What does text recognition pick up from old or poor originals?

Modern OCR usually reads even yellowed pages, carbon copies and old image-only PDFs reliably. There are limits with handwriting, very faint copies, stamps over text and damaged pages – there, recognition remains patchy and important documents should be spot-checked.

Can paper documents be thrown away after scanning?

Not across the board. Statutory retention obligations apply to many business records, and some documents, such as notarised deeds, only retain their full evidential value in the original. Clarify the destruction of originals beforehand with your tax adviser or legal adviser.

Do you need a special scanner for digitisation?

To start with, the existing multifunction device with a document feeder is usually enough. What matters is sufficient resolution, the PDF format and a feeder that processes stacks. Only for very large volumes is a dedicated document scanner or a scanning service worthwhile.

Who should tag the scanned documents?

Ideally nobody by hand. Automatic tagging reads the recognised text and fills in attributes such as customer, project, date and topic itself. Only the document type is confirmed by a person in the review area, because it decides who may see the document – for clear-cut scans that is one click per document; naming and sorting by hand is no longer needed.

Digitising a paper archive

The best way to digitise a paper archive is in four steps: first review and prioritise, then scan, then make it searchable with text recognition (OCR) and finally tag it largely automatically. The most important principle: it is not the entire archive that should be scanned, but the part that is actually needed in everyday work. This way, within a manageable time, the filing cabinet becomes a collection an AI can answer questions about in normal language.

In brief

Review comes before scanning: digitise only active and frequently used records first, and add older records when needed.
There are three routes for scanning – your own office device, a scanning service or gradual digitisation in everyday work – which combine well.
Text recognition (OCR) turns the scanned image into searchable text and today works surprisingly well even on old originals; handwriting remains its biggest weakness.
Automatic tagging saves weeks of manual naming and sorting; only the document type has to be confirmed, because visibility within the company depends on it.
Do not dispose of originals too hastily: retention obligations are a separate topic for tax or legal advice.

Why should you not scan the whole archive?

The reflex in digitisation projects is: everything has to go in. This is exactly where many projects fail, because an archive that has grown over decades consists largely of documents nobody will ever ask about again. If you scan them anyway, you pay with weeks of working time for zero benefit – and on top of that, you dilute later searches with irrelevant hits.

It makes more sense to do an honest review based on one simple question: what has anyone actually looked for in the cabinet in the last two years? Typically, these are current contracts, ongoing project and customer files, technical documents on installations that are still in operation, and personnel files of active employees. This core collection is usually considerably smaller than the entire archive – and it delivers almost all of the benefit.

A three-way split has proven useful for prioritisation:

Category	Examples	Approach
Actively used	Current contracts, open projects, installation documentation	Scan first
Rarely used but relevant	Completed projects from recent years, warranty cases	Second wave or when needed
Retention only	Old accounting records, closed old cases	Leave in the cabinet, do not scan

What scanning options are there – and which suits your company?

For the actual scanning, there are three routes that do not exclude each other. The first is your own office device: the existing multifunction device with a document feeder is usually enough for the core collection. An office worker works through the prioritised stack on the side over a few weeks – low cost, full control, but limited speed.

The second route is a scanning service. It collects the folders, scans in high quality with industrial devices and returns the files in a structured form. This is worthwhile for large volumes and when it needs to be quick; in return, you temporarily hand over the documents, which needs to be considered for sensitive files.

The third route is gradual digitisation in everyday work: every paper document that is picked up anyway is scanned or photographed with a smartphone on that occasion. In this way, the collection that is actually used digitises itself – without a separate project. In practice, most companies combine all three: the core collection on the office device, old records to the service provider when needed, new arrivals immediately in everyday work.

What does text recognition get out of old originals – and where are the limits?

A scan is initially just a photo of the page; it only becomes searchable through text recognition, known in technical jargon as OCR (optical character recognition). It reads the image and converts it into machine-readable text. Only this text allows an AI to answer the question “What was agreed with the Berger company in 2016?” with the right passage from the scanned contract.

Modern text recognition is considerably better than its reputation from earlier years. Yellowed pages, carbon copies, faxes, sheets fed in at an angle and old PDF files that exist only as images are, as a rule, recognised reliably. Tables and multi-column layouts are usually no longer an obstacle either.

Honestly stated, the limits are part of the picture too. Handwriting is the biggest: a cleanly printed form with a few handwritten entries is often only partly recognised, and a completely handwritten file note is often not usable at all. Very faint thermal paper receipts, stamps and signatures across the text, and damaged or heavily soiled pages also remain difficult. For such individual items: spot-check important documents after recognition and, if necessary, briefly record the content again as a note.

How do structure and tags get into the scanned documents?

After scanning, you quickly have a mountain of files with names like “Scan_0233.pdf”. The classic route – open each file, name it, sort it into folders – is exactly the drudgery on which digitisation projects in SMEs regularly get stuck.

Automatic tagging takes the tedious part of this off your hands: the software reads the text recognised by OCR and derives the attributes from it – customer, project, date, topic. It fills in these field values itself. Only the document type is presented to a person in the review area, because it determines who gets to see the document later – for clear-cut scans that is one click, and longer rework is reserved for special cases. “Scan_0233.pdf” thus becomes a findable maintenance contract with an assignment, without anyone typing file names or choosing folders.

In an AI knowledge base, the scanned archive is then one source among several, alongside network folders and email inbox – the article Connecting documents and systems gives the overall overview of all sources and the recommended order. The result: anyone who asks about an old agreement gets the answer together with a reference to the scanned original page – even if nobody remembers which folder it was in.

What happens to the paper originals after scanning?

The temptation to switch on the shredder after scanning is great. Be warned: statutory retention periods apply to many business records, and certain documents – such as notarised deeds or contracts with original signatures – only have their full evidential value in the original. Whether and when an original may be destroyed is a legal question that you should clarify with your tax adviser or legal adviser; this article expressly does not replace that advice.

A middle way is practical: after scanning, the originals go into a compact, chronologically ordered storage archive – without elaborate folder logic, because from now on searching is done digitally. The paper only serves for retention, no longer for daily work. This reduces the space required without losing anything legally.

What typical mistakes should you avoid?

Some recurring mistakes can be derived from failed and successful projects:

Wanting to scan everything at once. The project becomes a mammoth, motivation collapses, and after three weeks the stack is left lying. Better: core collection first, the rest when needed.
Starting without a review. If you scan unfiltered, you also digitise duplicates, outdated versions and irrelevant material – and thus worsen the quality of answers later.
Scan quality too low. Scans that are too heavily compressed or too coarse in resolution ruin text recognition. Better to set it up properly once than to redo a thousand pages.
Relying on manual tagging. Naming by hand is the step that fails most reliably in practice. Plan automatic tagging from the start – the document type is then confirmed and no longer typed.
Disposing of originals too hastily. Clarify retention obligations and evidential value beforehand, rather than regretting it afterwards.
Losing sight of the goal. You do not scan to own files but so that the team finds answers. Ask test questions straight after the first batch – this shows whether the selection and quality are right.

The last point is also the best way to start: in a structured introduction project, the initial loading is deliberately followed by a phase with test questions from real working life before the team is trained – what this looks like step by step is shown on the Process page. If you proceed in this way, you notice early on whether the digitised archive delivers what it should – and from the second wave onwards you scan exactly what is really in demand.

Digitising a paper archive: from the filing cabinet to AI queries