Document processing and OCR companies in South Africa: the types of provider compared

Companies that turn documents into data in South Africa fall into seven types: extraction software you run yourself, cloud OCR services your developers build on, scanning and digitisation bureaus, records management and storage firms, outsourced data capture teams, managed document processing services, and implementation partners for document management platforms. They are not interchangeable. Some sell software, some sell a service, and some sell both.

Start from your documents, not from a list of companies. Typed forms in volume suit software. Handwriting, poor scans and documents that differ every time still need people. Paper in boxes needs a bureau before anything else can happen. Once you know which of those you have, the type of provider follows, and you can shortlist two or three of that type and test them on your own documents.

This guide names no company and ranks none, because a fair ranking would need accuracy and turnaround figures measured on the same documents, which we cannot check from outside. Instead it describes the types, with written questions whose answers you can compare across providers and test in a pilot.

Four of the types turn documents into data: extraction software, cloud OCR services, data capture teams and managed services. The other three handle what surrounds that work: bureaus and storage firms deal with the paper, and implementation partners with the platform.

• Extraction software is sold as software and suits firms with steady volume, consistent layouts and their own people to run it.
• Cloud OCR services are sold as software building blocks and suit firms with developers who want to build their own flow.
• Scanning and digitisation bureaus sell a service and suit firms with paper archives or paper still arriving.
• Records management and storage firms sell a service and suit firms that must keep paper and want it stored, retrieved and destroyed on schedule.
• Outsourced data capture teams sell a service and suit handwriting, poor scans and documents that need judgement.
• Managed document processing sells software and people together, run for you, and suits firms that want the outcome without running the system.
• Implementation partners sell a project, sometimes with support after it, and suit firms that have chosen a platform and need it fitted into their systems.

What it does. Intelligent document processing software reads a document image, finds the fields that matter, checks them against rules, and gives each field a confidence score. Your own staff review the fields it is unsure of and correct its mistakes.

Who it suits. Firms with volume, a limited number of document layouts, and an operations or IT team that can own the set-up: adding new document types, tuning the rules and watching the error rate.

How it sells. As software: a licence or subscription. Ask whether a set-up project is extra. You do the work; the vendor supplies the tool.

What to ask before signing:

• Which of our document types has it read before, and can we test it on a sample of ours, including the worst scans?
• Is the accuracy figure measured by character, by field or by document, and what share of fields goes to a person?
• Does it run in our environment or the vendor's, and in which country is the data processed?
• What does adding a new document type involve, and who does it?
• If we leave, do we keep the rules and field definitions built on our documents?

What it does. The large cloud platforms offer text recognition and form extraction as services that a developer calls from code. They return the text and fields; everything around them, such as intake, checks, review screens and the audit trail, is yours to build.

Who it suits. Firms with developers and an unusual flow that no packaged product fits, or firms that already run most of their systems on one cloud.

How it sells. As software, a building block charged by use. The service is a component, not a finished product, so most of the cost is your own build and upkeep.

What to ask before signing:

• Where is each document processed and stored? Under section 72 of POPIA, sending personal information to a third party in another country needs a listed ground, such as a binding agreement giving adequate protection.
• Is our data used to train the service, and can that be switched off?
• Who in our team will build and maintain the review screens, the checks and the audit trail?

A bureau turns paper into images: it collects boxes, prepares documents (removing staples, repairing tears), scans them, indexes each file so it can be found, and returns or stores the images. Many also capture a few index fields, such as a policy number or a date.

Who it suits. Firms with paper archives, branches that still receive paper, or a backlog that must be digitised before any automation can start.

How it sells. As a service, by volume, with collection, preparation and indexing often quoted as separate lines.

What to ask before signing:

• How are boxes collected, tracked and secured in transit and on site?
• What resolution and file format will the images be, and which index fields are captured?
• How is quality checked: what share of pages is inspected, and what happens to a page that fails?
• What happens to the paper afterwards? Under section 16 of the Electronic Communications and Transactions Act, where a law requires information to be retained, keeping it electronically meets that requirement if the information is accessible for later reference, is in its original format or one shown to represent the information accurately, and shows its origin, destination, and date and time of sending or receipt. Take legal advice on your own documents before any paper is destroyed.

Storage firms keep paper off site, find and deliver files on request (often by scanning a file on demand), and destroy records when their keeping period ends, with a certificate of destruction.

Who it suits. Firms that must keep paper originals, or that hold years of records they rarely need but cannot yet destroy.

How it sells. As a service: storage by the box, retrieval by the request, and destruction by volume.

What to ask before signing:

• How quickly can a file be retrieved, and in what form?
• Can you apply our retention schedule and tell us when each box is due for destruction? Section 14 of POPIA says personal information may not be kept longer than its purpose needs, unless a law requires or authorises it, it is reasonably required for your lawful functions, a contract requires it, or the person consented. Once keeping it is no longer authorised, it must be destroyed, deleted or de-identified as soon as reasonably practicable.
• If we are an accountable institution under FICA, can you meet the access terms? A third party may keep FICA records on the institution's behalf as long as the institution has free and easy access to the records and they are readily available to the Financial Intelligence Centre and the supervisory body. The institution stays liable if the third party fails.

What it does. Trained people read each document and key the fields into your system or theirs. Some teams double-key important fields: two people capture the same field independently and any difference goes to a third. Ask whether yours will, and which fields.

Who it suits. Handwriting, phone photos, poor scans, long reports and documents that need judgement. Also low volumes, where software would cost more to set up than it saves.

How it sells. As a service: by the document, page, field or record, or as a dedicated team.

What to ask before signing:

• Who does the work, in which country, and are they dedicated to our account?
• How is accuracy measured, by field, and how often is it checked?
• Will you sign our operator agreement? Section 21(1) of POPIA requires you, in terms of a written contract, to ensure that the operator establishes and maintains the security measures in section 19.
• What stops staff copying data: no downloads, no printing, no personal phones at the desk, masked identity and bank numbers?
• How do you cover month-end peaks and staff leave?

Here the provider runs the whole flow for you: intake from email, uploads or shared drives, extraction by software, checks against your systems, people for the exceptions, and delivery of checked data into your system. You buy the result rather than the tool or the hours.

Who it suits. Firms that want automation but do not want to run software or hire a team for it, and firms whose document flow mixes clean forms with messy exceptions.

How it sells. As both: software and people bundled into one service. Ask how the set-up and the running fee are each charged.

What to ask before signing:

• Which parts are software and which are people, and who checks the people's work?
• What turnaround will you commit to by document type?
• Insurers only: the Policyholder Protection Rules set a claim-recording deadline. An insurer must record a claim by no later than the first business day after the date that the initial claim is received, and may not wait until all requirements for the claim have arrived. If a provider handles claim intake, the contract must let you meet that deadline.
• Which subcontractors and model providers touch our documents, and where?
• Is every decision and correction recorded in an audit trail we can read?
• If we leave, what do we get back: our data, our field definitions and the set-up built on our documents?

What it does. These firms resell and install a document or content management platform, connect it to your other systems, and configure the capture, workflow and archive around your processes.

Who it suits. Larger firms that have chosen, or must standardise on, one platform and need it fitted to their own systems and rules.

How it sells. As a project, with licences for the platform, and a support contract if you want one.

What to ask before signing:

• Who owns the configuration once the project ends, and can another partner support it?
• What is in the support contract, and how are changes charged?
• Which of the platform's capture features are included, and which are extra modules?

1. Sort your documents first. List each document type with its monthly volume, how it arrives (paper, email, upload, phone photo), whether it is typed or handwritten, and which fields matter. That list tells you which type of provider you need, and often that you need two.
2. Decide what you want to own. If you want to run the system, look at software or cloud services. If you want the outcome, look at a capture team or a managed service. If paper is the problem, start with a bureau.
3. Shortlist two or three providers of the right type, and put the same written questions to each.
4. Test on your own documents. Draw a sample from a normal month, including the worst ones, settle the correct answer for each field with your own people, and compare each provider's output field by field.
5. Compare on the cost of one correct, checked record in your system, including the exceptions your own staff still handle. A low fee with a high exception rate can cost more than a higher fee that sends fewer documents back to you.
6. Where the provider processes personal information for you, sign the operator agreement before any of it moves, and agree retention, deletion and the exit terms before you sign.

IAMX is one option. Its Document processing and data capture service is automation built first for insurers, brokers and financial services firms: it takes in your documents, reads what matters and acts on it. Each item is checked against your other systems, simple cases are handled automatically, and anything unusual is passed to a person for a decision. IAMX is one accountable front door to a vetted network of specialist partners.

• Which document processing companies operate in South Africa?

  They range from software vendors and cloud OCR services to scanning bureaus, records storage firms, data capture teams, managed services and implementation partners for document management platforms. Rather than start from a named list, decide which type fits your documents, then shortlist two or three of that type and test them on a sample of your own documents.

• Should we buy OCR software or outsource data capture?

  Buy software if your documents are typed, consistent and in volume, and you have people to run it. Outsource capture if they are handwritten, poorly scanned or varied, or the volume is too low to repay a set-up. Many firms need both, or a managed service that combines them.

• What is the difference between a scanning bureau and a document processing service?

  A bureau turns paper into images and indexes them so files can be found. A document processing service reads the fields inside each document, checks them and puts the data into your systems. Paper usually goes through a bureau first.

• What should a document processing contract include?

  How accuracy is measured and by which field, turnaround by document type, the fee basis and what counts as a page or document, the operator agreement that section 21(1) of POPIA requires where the provider processes personal information for you, where documents are processed and stored, how long images are kept, and what you get back if you leave.

• Can we destroy paper after it has been scanned?

  Take legal advice first. Under section 16 of the Electronic Communications and Transactions Act, where a law requires information to be retained, keeping it electronically meets that requirement if the information is accessible for later reference, is in its original format or one shown to represent the information accurately, and shows its origin, destination, and date and time of sending or receipt. Whether a particular paper original may then be destroyed is a question for your attorney.