Paperless ECM Services: What You Are Buying and Where OCR Fits
People who search for paperless ECM services are usually past the question of whether to scan. They want someone, or something, to take over the whole document lifecycle: get paper and email attachments into a system, make them findable, keep them for the right number of years, and hand them to the right person when an auditor or customer asks. That is what enterprise content management (ECM) means. A paperless ECM service is any mix of software and people that runs those steps for you. OCR is one part of the first step, not the product, and the buying mistakes mostly come from confusing the two.
This guide explains what the service category covers, what to ask a provider, and where an OCR or extraction API such as OCRskill fits next to a document management system.
What ECM covers, in plain terms
AIIM, the industry association that standardized the term in the early 2000s, defines enterprise content management as the strategies, methods, and tools used to capture, manage, store, preserve, and deliver information through its lifecycle. Those five verbs are still the clearest way to read a vendor proposal:
- Capture gets content into the system. Scanning, email ingestion, uploads, and OCR all live here, along with the indexing that makes a file usable.
- Manage covers metadata, permissions, versioning, and check-in or check-out, so the right people can find and change the right file.
- Store is where the file sits day to day: a repository, a NAS, a cloud bucket.
- Preserve is long-term retention and archiving, including keeping records unaltered for as long as rules require.
- Deliver is search, sharing, export, and integration with the systems that need the content.
An overview in Business & Information Systems Engineering uses the same model and adds integration as its own function, because ERP, CRM, and HR systems sit outside the ECM repository and still need to read from it and write to it. That is a useful reminder when a proposal treats “connects to your ERP” as a checkbox.
ECM, content services, and why the label keeps changing
If you compare vendors, you will also see “content services platform” (CSP). In 2017 Gartner dropped the ECM name for its market report and described content services as services and microservices that share APIs and repositories, rather than one self-contained system. AIIM agreed the newer tools are more modular but argues in its ECM vs. content services piece that ECM also names a practice, not just software.
For a buyer, the practical difference is architecture. An older ECM suite expects every document to live in its repository. A content services approach accepts that invoices sit in the accounting system, contracts in a CLM tool, and scans in a document archive, and tries to manage them across those places. Neither label tells you whether a provider will scan your backlog, tag it correctly, or delete it on schedule. Ask about the five functions instead of the category name.
What “paperless ECM services” can mean when you get a quote
The phrase covers very different offers. Sort quotes into one of these before you compare prices:
- Software only. A licensed or hosted ECM or document management platform. You run capture, indexing, and retention rules yourself.
- Scanning and conversion bureau. A provider picks up boxes, preps and scans pages, indexes them to a spec you approve, and returns images plus metadata. Ongoing intake is usually out of scope.
- Managed capture. A provider operates intake for you, often a mailroom or an email inbox, classifies documents, keys or extracts index fields, and loads them into your repository.
- Implementation services. Consultants configure the platform, design the metadata model and retention schedule, migrate legacy shares, and train staff.
- Self-hosted stack. An open-source document management system such as Paperless-ngx on your own server, plus OCR or extraction APIs where built-in recognition is not enough. No outside provider touches the paper after the scanner.
Most real projects combine two of these, for example a one-time bureau scan of the backlog followed by self-run intake for new mail. Write down which combination you want before vendor calls, or every demo will steer you to the vendor’s own slice.
Questions that separate a real service from a scanning job
Use these as a shortlist when reviewing a provider or your own design:
- Who owns the metadata model? Ask for the list of document types and index fields the provider will populate, with examples. “Full-text searchable” is not an index.
- How is quality checked? Ask what percentage of pages or index fields gets human verification, and what happens to pages the OCR could not read.
- What happens to originals? Destruction certificates, return of boxes, or offsite storage should be explicit, and should follow your retention rules, not the provider’s convenience.
- Which file format is the archive copy? For long-term preservation, many teams keep a PDF/A copy with a text layer next to the original file. Confirm you can get both.
- How are retention and deletion enforced? ISO 15489-1:2016 sets out records management concepts such as records controls and metadata for records. A provider does not need to cite it, but they should be able to show how a retention period is attached to a document type and what triggers disposal.
- How do you leave? Ask for an export of files plus metadata in an open format, and test a sample export before signing. Lock-in is the most expensive line item nobody prices.
- What integrates, and how? Get the actual mechanism: API, watched folder, scheduled export, or manual download. Then ask who maintains it when one side upgrades.
If a provider cannot answer the metadata and exit questions concretely, you are buying a scanning job with ECM branding.
Where OCR fits inside ECM capture
OCR turns page images into text. Inside an ECM process it does two jobs:
- Search. A text layer behind each scan lets staff find a document by a phrase on the page. Most document management systems, including Paperless-ngx, run OCR on intake for this reason.
- Indexing. Pulling specific values, such as a supplier name, a document date, or an address, into metadata fields so documents can be filtered, routed, and retained by type.
The first job is a commodity. The second is where projects slow down, because generic full-text OCR does not tell the system which date on the page is the document date. That is the extraction layer, sometimes sold as intelligent document processing.
OCRskill sits in that extraction layer. It offers two endpoints:
POST https://api.ocrskill.com/ocrreturns readable Markdown of the page, which is useful as the searchable body and for anything that is not a fixed field.POST https://api.ocrskill.com/ocr.json?fields=...returns typed JSON for the fields you name. Add a trailing?to mark a field optional. If a required field cannot be found, the call returns400with the extracted text, so you can send the document to review instead of filing it under bad metadata.
Both accept images and documents, including PDF, up to 20 MB per request, with Authorization: Bearer auth. A free key from get-key.json is enough to test.
A minimal indexing call for an incoming supplier document:
curl https://api.ocrskill.com/get-key.json
export API_KEY="sk-your-key-here"
curl "https://api.ocrskill.com/ocr.json?fields=company_name,invoice_date?,town_or_city?,country?" \
-H "Authorization: Bearer $API_KEY" \
-F "file=@incoming-scan.pdf"
Map the returned fields to your repository’s metadata (correspondent, date, location) and keep the Markdown from /ocr as the body text. The field catalog and schema rules are covered in the OCR JSON API reference and the form data extraction API guide. Pick only the fields you will actually search or retain by. Every extra field is one more thing a clerk has to check.
What OCRskill is not in an ECM stack
Be clear about the boundary so nobody buys the wrong thing:
- OCRskill does not store your documents long term. It is not a repository, records system, or archive.
- It does not enforce retention, legal holds, permissions, or deletion schedules.
- It does not watch inboxes or scanners on its own. Your capture step, or a document management system’s consume folder, has to send files to it.
- It does not replace human review for low-confidence pages. Design the review queue first.
In ECM terms, OCRskill helps with capture and indexing. Manage, store, preserve, and deliver belong to the platform you choose.
A self-hosted paperless ECM pattern
Smaller teams often do not need a full enterprise suite. A workable pattern for an office or a single department looks like this:
- Repository and search. Run Paperless-ngx on a server or NAS. It handles a consume folder, OCR text layer, tags, document types, correspondents, storage paths, and browser search. The Paperless-ngx documentation covers setup, and our electronic archive tutorial walks through a Docker deployment.
- Metadata rules. Define document types and matching rules before the archive grows. The Paperless-ngx tips post covers inbox tags, workflows, and storage paths that hold up.
- Extraction where built-in OCR stops. Send document families that need typed fields, such as supplier paperwork or forms, to
/ocr.json, then write the values back as custom fields or tags. - Retention. Keep a written schedule per document type and review it with whoever owns compliance. A self-hosted system will not decide this for you.
- Backups and exit. Use the exporter, store copies off the box, and test a restore. This also answers the exit question you would ask any vendor.
This pattern covers capture, manage, store, and deliver well for a small team. Preservation and formal records controls take more discipline because nobody else is accountable for them. If you operate in a regulated sector with strict records obligations, a managed platform or records consultant may be worth paying for even if the scanning stays in house.
Choosing between a provider and doing it yourself
A rough way to decide:
- Large paper backlog, small team. Pay a scanning bureau for the backlog with a strict indexing spec, then run new intake yourself.
- High daily volume across many departments. A managed capture service or full content services platform usually earns its cost through staffing and support, not features.
- One department, moderate volume, technical staff available. A self-hosted document management system plus an extraction API is often enough, and you keep full control of the files.
- Heavy regulation or litigation exposure. Prioritize retention, audit trails, and legal holds over capture features, and involve your records or legal owner before choosing.
Industry-specific versions of this design, with intake channels and document types already mapped, are in our workflow guides for accounting and healthcare.
Conclusion
Paperless ECM services are worth buying when you know which of the five functions you are handing off. Capture is the easiest to demo and the least likely to be your real problem. The parts that decide whether a paperless project lasts are the metadata model, the review path for unreadable pages, the retention rules, and a tested way to export everything and leave. Put OCR and structured extraction where they belong, inside capture and indexing, and judge every provider on the rest.
